Public report of huggingface/text-embeddings-inference, published 14 November 2026. It has the same score, grades and number of
findings as the full report. The details of security findings (where they are in the code, which known
vulnerabilities and which passwords or keys were found) are left out, so nobody can use them against the project.
Public
Executive reportMeasured under the Code Assurance Index, an open standard for code quality, using its published scoring rules (rubric version 2026.11.4). Assert signed this survey and filed it in the standard's public register, where anyone can look it up by its survey ID. Check this survey →
Signed and filed in the standard's public register, 14 November 2026, 19:26 UTC · Survey ID cd_2bb06c0cda384882b525e0cdfff1072b
huggingface/text-embeddings-inference
Measured 14 November 2026, 19:24 UTC, at code version da9a9ce
Knowledge freshness is not measured: Freshness signal contradicted by repo activity.
CAI scoreAt riskUp 1 since the last survey
The score is built from these lenses. Each lens grades one area of the code, such as how easy it is to change or how well it is protected. The weakest lens counts most, so one weak area pulls the score down instead of being averaged away.
A medium codebase: 26,547 lines of product code in 11 projects, tests not counted · Estimated rebuild effort: about 0.2 person-years · Weakest lens: Code Health, 55
Findings by how serious they are
61 critical471 serious61 minor
21 dimensions are left out of the score. For each one, we either couldn't measure it in this code, or found no evidence of the safeguard we look for, such as a tested backup restore. They count neither for nor against the score, and Limitations lists each one with its reason.
This survey was produced by
Canine Code Analyzer
Producer
Canine Development
Analyzer
Canine Code Analyzer engine 1.0.0
Measured
14 November 2026, 19:24 UTC
A measurement, not a certificate. The Code Assurance Index records a number
together with the evidence it was worked out from, so anyone can recalculate it, but it doesn't certify or guarantee
the code. The standard was written by Canine Development, which also builds the Canine Code Analyzer. It is open to any tool that
implements it, and today the Canine Code Analyzer is the only one that does.
A critical security finding is open, such as a library with a known critical vulnerability or a password or key found in the code. The score itself is unchanged, but its band shows At risk until the finding is fixed. See the Security and Compliance lens.
The score is deterministic, which means the same code always gets the same score. We work it out with tools that follow published rules, and an AI model never decides it, so anyone who applies the same rules to the same code gets the same number. How it is done is explained under Methodology ▸
41of 41 dimensions measured by tools.
581of 593 findings point to an exact file and line, so you can open each one in the code. The other 12 concern the repository as a whole, such as missing documentation.
41dimensions across 6 lenses, read from 26,547 lines of product code in 11 projects.
huggingface/text-embeddings-inference is rated At risk because critical security findings are open, most of them in the packages it uses. It scores 61 out of 100, which the At risk rating leaves unchanged.
Code Health, at 55, is the weakest lens and counts most in the score. The best first step is recording design decisions, worth about 4 points. Writing log messages as searchable templates would lift Code Health itself, also worth about 4. Raising Code Health to 70 would lift the score to about 64.
Forty-two known security holes in the packages the code depends on can be used by attackers. Seventeen settings leave the servers or containers unsafe, and one pipeline step installs packages that can change without warning. Updating the vulnerable packages, correcting the unsafe settings and pinning what the pipeline installs would close these risks. The place where the code runs text as code or as a command needs a direct call instead.
The product code is 26,547 lines, about €32,000 to rebuild.
A lens is capped at Adequate while one of its dimensions scores below 4 out of 10, however well the rest of the lens does. That way a serious problem in one place always shows in the lens's rating.
How the score is built: what each lens counts forThe lenses don't count equally. The weakest lens counts most, the next weakest a little less, and so on, so a weak area pulls the score down instead of being averaged away. In the bar below, each part's width shows how much that lens counts and its colour shows its rating. Improving the widest part moves the score most, and as that lens improves, it counts for less, because another lens becomes the weakest.
If Code Health rose from 55 to 70, the lowest score rated Strong, the score would rise from 61 to about 64.
Code composition: what the lines of code are
Product code 89%Tests 11%
New since the last survey: 160 findings
These 160 critical or serious findings weren't in the previous survey, on 28 October 2026. The first 100 are listed here, and the findings file in the download package lists them all.
Code Duplication · Near-duplicate member family (10 members, 56 shared lines) backends/candle/src/models/flash_bert.rs
Code Duplication · Edited copy of a member (30 corresponding lines) backends/candle/src/models/flash_modernbert.rs
Code Duplication · Members sharing a duplicated core (12 members, 50+ identical tokens) backends/candle/src/models/bert.rs
Code Duplication · Members sharing a duplicated core (10 members, 50+ identical tokens) router/src/grpc/server.rs
Code Duplication · Members sharing a duplicated core (5 members, 50+ identical tokens) backends/candle/src/models/flash_mistral.rs
Code Duplication · Edited copy of a member (77 corresponding lines) backends/candle/src/models/flash_qwen3.rs
Code Duplication · Edited copy of a member (72 corresponding lines) backends/candle/src/models/flash_modernbert.rs
Code Duplication · Edited copy of a member (75 corresponding lines) router/src/grpc/server.rs
Code Duplication · Edited copy of a member (63 corresponding lines) backends/candle/src/models/flash_nomic.rs
Code Duplication · Edited copy of a member (54 corresponding lines) backends/candle/src/models/distilbert.rs
Code Duplication · Edited copy of a member (39 corresponding lines) backends/python/server/text_embeddings_server/models/flash_bert.py
Code Duplication · Edited copy of a member (42 corresponding lines) core/src/infer.rs
Code Duplication · Edited copies of one member (4 members, 38+ corresponding lines) backends/candle/src/models/flash_gte.rs
Code Duplication · Edited copy of a member (36 corresponding lines) backends/candle/src/models/flash_bert.rs
Code Duplication · Edited copy of a member (28 corresponding lines) backends/python/src/lib.rs
Code Duplication · Edited copies of one member (6 members, 28+ corresponding lines) backends/candle/src/models/flash_gte.rs
Code Duplication · Edited copies of one member (3 members, 42+ corresponding lines) router/src/grpc/server.rs
Code Duplication · Edited copy of a member (24 corresponding lines) backends/python/server/text_embeddings_server/models/flash_mistral.py
Code Duplication · Edited copies of one member (3 members, 31+ corresponding lines) backends/candle/src/models/bert.rs
Code Duplication · Edited copy of a member (33 corresponding lines) backends/candle/src/models/flash_distilbert.rs
Code Duplication · Edited copies of one member (3 members, 29+ corresponding lines) backends/python/server/text_embeddings_server/models/classification_model.py
Code Duplication · Edited copy of a member (31 corresponding lines) backends/candle/src/models/bert.rs
Code Duplication · Edited copy of a member (23 corresponding lines) backends/candle/src/models/distilbert.rs
Code Duplication · Edited copy of a member (31 corresponding lines) backends/candle/src/models/flash_gte.rs
Code Duplication · Edited copy of a member (30 corresponding lines) backends/candle/src/layers/cublaslt.rs
Code Duplication · Edited copies of one member (4 members, 18+ corresponding lines) backends/candle/src/models/flash_jina.rs
Code Duplication · Edited copy of a member (21 corresponding lines) backends/candle/src/models/flash_modernbert.rs
Code Duplication · Edited copies of one member (4 members, 20+ corresponding lines) router/src/grpc/server.rs
Code Duplication · Edited copy of a member (13 corresponding lines) backends/python/server/text_embeddings_server/models/jinaBert_model.py
Code Duplication · Edited copy of a member (18 corresponding lines) backends/candle/src/models/debertav2.rs
Code Duplication · Edited copies of one member (3 members, 16+ corresponding lines) backends/candle/src/models/distilbert.rs
Code Duplication · Edited copy of a member (11 corresponding lines) backends/candle/src/models/flash_mistral.rs
Code Duplication · Edited copies of one member (4 members, 15+ corresponding lines) backends/python/server/text_embeddings_server/models/flash_bert.py
Code Duplication · Edited copy of a member (27 corresponding lines) backends/python/server/text_embeddings_server/models/flash_mistral.py
Code Duplication · Edited copy of a member (12 corresponding lines) backends/candle/src/models/flash_bert.rs
Code Duplication · Edited copy of a member (20 corresponding lines) backends/python/server/text_embeddings_server/models/flash_mistral.py
Code Duplication · Edited copy of a member (13 corresponding lines) backends/candle/src/models/bert.rs
Code Duplication · Edited copy of a member (10 corresponding lines) backends/candle/src/models/flash_jina.rs
Code Duplication · Edited copy of a member (14 corresponding lines) backends/candle/src/models/flash_modernbert.rs
Code Duplication · Edited copy of a member (14 corresponding lines) backends/candle/src/models/bert.rs
Code Duplication · Edited copy of a member (8 corresponding lines) backends/candle/src/models/flash_nomic.rs
Code Duplication · Edited copies of one member (3 members, 15+ corresponding lines) backends/candle/src/models/flash_gte.rs
Code Duplication · Edited copy of a member (8 corresponding lines) backends/candle/src/models/bert.rs
Code Duplication · Edited copy of a member (8 corresponding lines) backends/candle/src/models/distilbert.rs
Code Duplication · Edited copies of one member (3 members, 26+ corresponding lines) core/src/tokenization.rs
Code Duplication · Edited copies of one member (4 members, 13+ corresponding lines) backends/candle/src/models/distilbert.rs
Code Duplication · Edited copies of one member (4 members, 11+ corresponding lines) backends/candle/src/models/flash_qwen3.rs
Code Duplication · Edited copies of one member (4 members, 10+ corresponding lines) backends/candle/src/models/bert.rs
Code Duplication · Edited copy of a member (14 corresponding lines) backends/python/server/text_embeddings_server/utils/flash_attn.py
Code Duplication · Edited copy of a member (12 corresponding lines) backends/candle/src/models/bert.rs
Code Duplication · Edited copy of a member (11 corresponding lines) backends/candle/src/models/debertav2.rs
Code Duplication · Edited copy of a member (9 corresponding lines) backends/candle/src/models/debertav2.rs
Code Duplication · Edited copy of a member (11 corresponding lines) backends/candle/src/models/flash_distilbert.rs
Code Duplication · Edited copy of a member (4 corresponding lines) backends/python/server/text_embeddings_server/models/flash_mistral.py
Code Duplication · Edited copy of a member (6 corresponding lines) core/src/tokenization.rs
We compare every finding with the previous survey's. A finding that has only moved to another line in the same file counts as the same finding, so only new findings appear here. Advisory notes from the AI model are left out, because the model words them afresh in every survey.
Fixed since the last survey: 79 findings
These 79 critical or serious findings were in the previous survey, on 28 October 2026, and are gone from this one.
Code Duplication · Near-duplicate member family (10 members, 56 shared lines) backends/candle/src/models/flash_bert.rs
Code Duplication · Edited copy of a member (30 corresponding lines) backends/candle/src/models/flash_modernbert.rs
Code Duplication · Members sharing a duplicated core (12 members, 50+ identical tokens) backends/candle/src/models/bert.rs
Code Duplication · Members sharing a duplicated core (10 members, 50+ identical tokens) router/src/grpc/server.rs
Code Duplication · Members sharing a duplicated core (5 members, 50+ identical tokens) backends/candle/src/models/flash_mistral.rs
Dependency Vulnerabilities · Medium vulnerability: [GHSA redacted] REDACTED
Dependency Vulnerabilities · Medium vulnerability: RUSTSEC-2026-0367 REDACTED
Dependency Vulnerabilities · Medium CVE: [GHSA redacted] REDACTED
Dependency Vulnerabilities · Medium CVE: [GHSA redacted] REDACTED
Dependency Vulnerabilities · Medium CVE: [GHSA redacted] REDACTED
Dependency Vulnerabilities · Medium CVE: [GHSA redacted] REDACTED
Dependency Vulnerabilities · Medium CVE: [GHSA redacted] REDACTED
Dependency Vulnerabilities · Medium CVE: [GHSA redacted] REDACTED
Dependency Vulnerabilities · Medium CVE: [GHSA redacted] REDACTED
Supply-chain Provenance & Signing · CI installs an unverified third-party binary REDACTED
Supply-chain Provenance & Signing · Base-context workflow trigger runs with an unscoped token .github/workflows/upload_pr_documentation.yml
Supply-chain Provenance & Signing · Workflow holding a long-lived secret is unscoped .github/workflows/build_documentation.yml
Supply-chain Provenance & Signing · PR-triggered workflow without a permissions block .github/workflows/build_pr_documentation.yml
Supply-chain Provenance & Signing · Secret handed to the step that runs a pull request's code
Rebuild cost: about €32,000 · Estimate
Cost to rebuild
€11,000 to €54,000, for 179 to 569 hours of work, about 0.2 person-years for one engineer
Difficulty of the problem
Standard, judged from the kind of software and how it is built. Code that solves a harder problem, such as payments or medical bookings, costs more per line to write.
Quality factor
×0.8 at a score of 61. Better code takes more work to write, so the factor is ×1.0 at a score of 70, lower below it and higher above it.
Size
Medium: 26,547 lines of product code
What it is not
Not a quotation: nobody has offered to build it for that. Not a valuation: it says nothing about what the business is worth. Not a certified figure: no professional has signed it as one.
How the estimate is made: We price the hours a team would need to write this code again, at €60 to €95 an hour. Declarations and boilerplate are priced at a fast rate and logic at a slower one, and the logic is then multiplied by the difficulty of the problem and by the quality factor. The estimate can be about 30% off either way, and it moves most with the hourly rate and the difficulty. For this code, every line is priced as straight logic, which keeps the estimate on the low side.
Top priorities
The three actions that raise the score most for the work they take. The full ranked list is in the Roadmap.
1
Resolve the 1 No ADRs found finding(s) in ADR Quality — start with Cargo.toml.
Raises the score by about 4 points · Low effort · ADR Quality
Each conclusion below combines several dimensions into one point that no single dimension shows, and lists the measurements it rests on.
Where the investment is most exposed · Value at risk
This code would cost about €32,000 to build again, about 0.2 person-years of work. Its weakest lens is Code Health, at 55 (Adequate). Code Health is about how easy the code is to read and change safely, and code that is hard to read is slow and risky to change, so every future change costs more than it should. Because the weakest lens counts most in the score, work on Code Health also moves the score most.
Based on: the rebuild estimate · the Code Health grade · how much each lens counts
What to do: Start with the Code Health actions in Top priorities.
If you do one thing first · First action
Resolve the 1 No ADRs found finding(s) in ADR Quality. Of all the actions in this report, it raises the score most for the work it takes: about 4 points for low effort, because it clears the 1 ADR Quality findings. The next survey will show the change in the score.
Based on: the ranked actions in Top priorities, by score gain and effort
What hard-to-read code costs each year · Cost
Four dimensions measure how easy the code is to work in: two measure how tangled its logic is, one how much of it is copied, and one how well each part sticks to a single job. In this codebase they average 4.6 out of 10. Code at that level is slower to change than clean code, and we estimate the extra work at 9% to 19% on every change. The history of the code shows how much it changes: about 4,403 lines over the last 90 days, which is about 17,857 lines a year. At that pace, the extra work comes to an estimated 3.8 to 22.8 engineer-days a year, and it grows as more code changes.
How this is estimated: The number of changed lines is measured from the history of the code. The rest rests on two assumptions of our own: code scoring 8 out of 10 or more adds no extra work, and each point below 8 adds 2.5% to 5.6% to every change, never less than 2% to 5% and never more than 40% to 60% in all; and an engineer changes 150 to 400 lines in a day. The result is a range for planning, not a measurement.
Based on: Cyclomatic Complexity · Cognitive Complexity · Code Duplication · Cohesion (LCOM4) · the code's change history
What to do: Start with the files listed under Churn × Complexity Hotspots, which change often and are hard to read, because that is where the extra work is paid most often.
Architecture — module dependency graph
Project dependencies, layered top-to-bottom; arrows show direction. Any dashed red edge points upward or sideways, which only happens where the projects form a dependency cycle. A graph without cycles has none.
Architecture: which parts of the code depend on which
The code is divided into modules, groups of code that each handle one part of the product, and a module depends on another when it uses code from it. The table lists the modules both down the side and across the top, in an order where each module should only use modules listed above it. A number in a cell means the module on that row uses the module in that column, and it counts the pairs of building blocks (types) that connect them. The one thing to look for is a mark above the diagonal line. It means two or more modules use each other in a loop (a dependency cycle), so none of them can be changed, tested or released on its own.
79 modules, with 159 dependencies between them. 5 loops run through 10 of the modules, marked above the diagonal.
The table shows the 40 modules with the most connections; 39 more are left out to keep it readable.
Which modules use which. The module on each row uses the module in each column, and the number counts the pairs of types that connect them. A cell above the diagonal is part of a loop.
1 pair of types, where a type in backends.python.server.text_embeddings_server.models.classification_model uses a type in backends.python.server.text_embeddings_server.models.model.
backends.python.server.text_embeddings_server.models.classification_model uses backends.python.server.text_embeddings_server.models.model, so a change to backends.python.server.text_embeddings_server.models.model can break backends.python.server.text_embeddings_server.models.classification_model, while a change to backends.python.server.text_embeddings_server.models.classification_model can't break backends.python.server.text_embeddings_server.models.model.
Position
Below the diagonal: backends.python.server.text_embeddings_server.models.classification_model uses a module listed above it, which is the healthy direction.
1 pair of types, where a type in backends.python.server.text_embeddings_server.models.classification_model uses a type in backends.python.server.text_embeddings_server.models.types.
backends.python.server.text_embeddings_server.models.classification_model uses backends.python.server.text_embeddings_server.models.types, so a change to backends.python.server.text_embeddings_server.models.types can break backends.python.server.text_embeddings_server.models.classification_model, while a change to backends.python.server.text_embeddings_server.models.classification_model can't break backends.python.server.text_embeddings_server.models.types.
Position
Below the diagonal: backends.python.server.text_embeddings_server.models.classification_model uses a module listed above it, which is the healthy direction.
1 pair of types, where a type in backends.python.server.text_embeddings_server.models.default_model uses a type in backends.python.server.text_embeddings_server.models.model.
backends.python.server.text_embeddings_server.models.default_model uses backends.python.server.text_embeddings_server.models.model, so a change to backends.python.server.text_embeddings_server.models.model can break backends.python.server.text_embeddings_server.models.default_model, while a change to backends.python.server.text_embeddings_server.models.default_model can't break backends.python.server.text_embeddings_server.models.model.
Position
Below the diagonal: backends.python.server.text_embeddings_server.models.default_model uses a module listed above it, which is the healthy direction.
1 pair of types, where a type in backends.python.server.text_embeddings_server.models.default_model uses a type in backends.python.server.text_embeddings_server.models.types.
backends.python.server.text_embeddings_server.models.default_model uses backends.python.server.text_embeddings_server.models.types, so a change to backends.python.server.text_embeddings_server.models.types can break backends.python.server.text_embeddings_server.models.default_model, while a change to backends.python.server.text_embeddings_server.models.default_model can't break backends.python.server.text_embeddings_server.models.types.
Position
Below the diagonal: backends.python.server.text_embeddings_server.models.default_model uses a module listed above it, which is the healthy direction.
text_embeddings_backend uses text_embeddings_backend.dtype, so a change to text_embeddings_backend.dtype can break text_embeddings_backend, while a change to text_embeddings_backend can't break text_embeddings_backend.dtype.
Position
Below the diagonal: text_embeddings_backend uses a module listed above it, which is the healthy direction.
text_embeddings_backend uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend, while a change to text_embeddings_backend can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models, while a change to text_embeddings_backend_candle.models can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.mistral uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.mistral, while a change to text_embeddings_backend_candle.models.mistral can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.mistral uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.mistral uses text_embeddings_backend_candle.layers.rotary, so a change to text_embeddings_backend_candle.layers.rotary can break text_embeddings_backend_candle.models.mistral, while a change to text_embeddings_backend_candle.models.mistral can't break text_embeddings_backend_candle.layers.rotary.
Position
Below the diagonal: text_embeddings_backend_candle.models.mistral uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.qwen2 uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.qwen2, while a change to text_embeddings_backend_candle.models.qwen2 can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.qwen2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.qwen2 uses text_embeddings_backend_candle.layers.rotary, so a change to text_embeddings_backend_candle.layers.rotary can break text_embeddings_backend_candle.models.qwen2, while a change to text_embeddings_backend_candle.models.qwen2 can't break text_embeddings_backend_candle.layers.rotary.
Position
Below the diagonal: text_embeddings_backend_candle.models.qwen2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_python uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_python, while a change to text_embeddings_backend_python can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_python uses a module listed above it, which is the healthy direction.
text_embeddings_router uses text_embeddings_backend.dtype, so a change to text_embeddings_backend.dtype can break text_embeddings_router, while a change to text_embeddings_router can't break text_embeddings_backend.dtype.
Position
Below the diagonal: text_embeddings_router uses a module listed above it, which is the healthy direction.
text_embeddings_router uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_router, while a change to text_embeddings_router can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_router uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.bert uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.bert, while a change to text_embeddings_backend_candle.models.bert can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.bert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.bert uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.bert, while a change to text_embeddings_backend_candle.models.bert can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.bert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.bert uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.bert, while a change to text_embeddings_backend_candle.models.bert can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.bert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.bert uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.bert, while a change to text_embeddings_backend_candle.models.bert can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.bert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.debertav2 uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.debertav2, while a change to text_embeddings_backend_candle.models.debertav2 can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.debertav2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.debertav2 uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.debertav2, while a change to text_embeddings_backend_candle.models.debertav2 can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.debertav2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.debertav2 uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.debertav2, while a change to text_embeddings_backend_candle.models.debertav2 can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.debertav2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.debertav2 uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.debertav2, while a change to text_embeddings_backend_candle.models.debertav2 can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.debertav2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_mistral uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_mistral, while a change to text_embeddings_backend_candle.models.flash_mistral can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_mistral uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_mistral uses text_embeddings_backend_candle.layers.rms_norm, so a change to text_embeddings_backend_candle.layers.rms_norm can break text_embeddings_backend_candle.models.flash_mistral, while a change to text_embeddings_backend_candle.models.flash_mistral can't break text_embeddings_backend_candle.layers.rms_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_mistral uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_mistral uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_mistral, while a change to text_embeddings_backend_candle.models.flash_mistral can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_mistral uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_mistral uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_mistral, while a change to text_embeddings_backend_candle.models.flash_mistral can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_mistral uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_mistral uses text_embeddings_backend_candle.models.mistral, so a change to text_embeddings_backend_candle.models.mistral can break text_embeddings_backend_candle.models.flash_mistral, while a change to text_embeddings_backend_candle.models.flash_mistral can't break text_embeddings_backend_candle.models.mistral.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_mistral uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen2 uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_qwen2, while a change to text_embeddings_backend_candle.models.flash_qwen2 can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen2 uses text_embeddings_backend_candle.layers.rms_norm, so a change to text_embeddings_backend_candle.layers.rms_norm can break text_embeddings_backend_candle.models.flash_qwen2, while a change to text_embeddings_backend_candle.models.flash_qwen2 can't break text_embeddings_backend_candle.layers.rms_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen2 uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_qwen2, while a change to text_embeddings_backend_candle.models.flash_qwen2 can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen2 uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_qwen2, while a change to text_embeddings_backend_candle.models.flash_qwen2 can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen2 uses text_embeddings_backend_candle.models.qwen2, so a change to text_embeddings_backend_candle.models.qwen2 can break text_embeddings_backend_candle.models.flash_qwen2, while a change to text_embeddings_backend_candle.models.flash_qwen2 can't break text_embeddings_backend_candle.models.qwen2.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen2 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gemma3 uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.gemma3, while a change to text_embeddings_backend_candle.models.gemma3 can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.gemma3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gemma3 uses text_embeddings_backend_candle.layers.rotary, so a change to text_embeddings_backend_candle.layers.rotary can break text_embeddings_backend_candle.models.gemma3, while a change to text_embeddings_backend_candle.models.gemma3 can't break text_embeddings_backend_candle.layers.rotary.
Position
Below the diagonal: text_embeddings_backend_candle.models.gemma3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gemma3 uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.gemma3, while a change to text_embeddings_backend_candle.models.gemma3 can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.gemma3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gemma3 uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.gemma3, while a change to text_embeddings_backend_candle.models.gemma3 can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.gemma3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.mpnet uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.mpnet, while a change to text_embeddings_backend_candle.models.mpnet can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.mpnet uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.mpnet uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.mpnet, while a change to text_embeddings_backend_candle.models.mpnet can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.mpnet uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.mpnet uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.mpnet, while a change to text_embeddings_backend_candle.models.mpnet can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.mpnet uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.mpnet uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.mpnet, while a change to text_embeddings_backend_candle.models.mpnet can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.mpnet uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.nomic uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.nomic, while a change to text_embeddings_backend_candle.models.nomic can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.nomic uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.nomic uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.nomic, while a change to text_embeddings_backend_candle.models.nomic can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.nomic uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.nomic uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.nomic, while a change to text_embeddings_backend_candle.models.nomic can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.nomic uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.nomic uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.nomic, while a change to text_embeddings_backend_candle.models.nomic can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.nomic uses a module listed above it, which is the healthy direction.
23→28 text_embeddings_backend_candle.models.nomic uses text_embeddings_backend_candle.models.flash_nomicpart of a loop✕
Connections
3 pairs of types, where a type in text_embeddings_backend_candle.models.nomic uses a type in text_embeddings_backend_candle.models.flash_nomic.
text_embeddings_backend_candle.models.nomic uses text_embeddings_backend_candle.models.flash_nomic, so a change to text_embeddings_backend_candle.models.flash_nomic can break text_embeddings_backend_candle.models.nomic, while a change to text_embeddings_backend_candle.models.nomic can't break text_embeddings_backend_candle.models.flash_nomic.
Position
Above the diagonal, so it is part of a loop. Neither module can be changed, tested or released on its own until one of these dependencies is removed.
text_embeddings_backend_candle.models.qwen3 uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.qwen3, while a change to text_embeddings_backend_candle.models.qwen3 can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.qwen3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.qwen3 uses text_embeddings_backend_candle.layers.rms_norm, so a change to text_embeddings_backend_candle.layers.rms_norm can break text_embeddings_backend_candle.models.qwen3, while a change to text_embeddings_backend_candle.models.qwen3 can't break text_embeddings_backend_candle.layers.rms_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.qwen3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.qwen3 uses text_embeddings_backend_candle.layers.rotary, so a change to text_embeddings_backend_candle.layers.rotary can break text_embeddings_backend_candle.models.qwen3, while a change to text_embeddings_backend_candle.models.qwen3 can't break text_embeddings_backend_candle.layers.rotary.
Position
Below the diagonal: text_embeddings_backend_candle.models.qwen3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.qwen3 uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.qwen3, while a change to text_embeddings_backend_candle.models.qwen3 can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.qwen3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.qwen3 uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.qwen3, while a change to text_embeddings_backend_candle.models.qwen3 can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.qwen3 uses a module listed above it, which is the healthy direction.
24→29 text_embeddings_backend_candle.models.qwen3 uses text_embeddings_backend_candle.models.flash_qwen3part of a loop✕
Connections
3 pairs of types, where a type in text_embeddings_backend_candle.models.qwen3 uses a type in text_embeddings_backend_candle.models.flash_qwen3.
text_embeddings_backend_candle.models.qwen3 uses text_embeddings_backend_candle.models.flash_qwen3, so a change to text_embeddings_backend_candle.models.flash_qwen3 can break text_embeddings_backend_candle.models.qwen3, while a change to text_embeddings_backend_candle.models.qwen3 can't break text_embeddings_backend_candle.models.flash_qwen3.
Position
Above the diagonal, so it is part of a loop. Neither module can be changed, tested or released on its own until one of these dependencies is removed.
text_embeddings_core.infer uses text_embeddings_backend, so a change to text_embeddings_backend can break text_embeddings_core.infer, while a change to text_embeddings_core.infer can't break text_embeddings_backend.
Position
Below the diagonal: text_embeddings_core.infer uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.distilbert uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.distilbert, while a change to text_embeddings_backend_candle.models.distilbert can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.distilbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.distilbert uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.distilbert, while a change to text_embeddings_backend_candle.models.distilbert can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.distilbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.distilbert uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.distilbert, while a change to text_embeddings_backend_candle.models.distilbert can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.distilbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.distilbert uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.distilbert, while a change to text_embeddings_backend_candle.models.distilbert can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.distilbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.distilbert uses text_embeddings_backend_candle.models.bert, so a change to text_embeddings_backend_candle.models.bert can break text_embeddings_backend_candle.models.distilbert, while a change to text_embeddings_backend_candle.models.distilbert can't break text_embeddings_backend_candle.models.bert.
Position
Below the diagonal: text_embeddings_backend_candle.models.distilbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_bert uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.flash_bert, while a change to text_embeddings_backend_candle.models.flash_bert can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_bert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_bert uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_bert, while a change to text_embeddings_backend_candle.models.flash_bert can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_bert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_bert uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_bert, while a change to text_embeddings_backend_candle.models.flash_bert can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_bert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_bert uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_bert, while a change to text_embeddings_backend_candle.models.flash_bert can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_bert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_bert uses text_embeddings_backend_candle.models.bert, so a change to text_embeddings_backend_candle.models.bert can break text_embeddings_backend_candle.models.flash_bert, while a change to text_embeddings_backend_candle.models.flash_bert can't break text_embeddings_backend_candle.models.bert.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_bert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_nomic uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.flash_nomic, while a change to text_embeddings_backend_candle.models.flash_nomic can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_nomic uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_nomic uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_nomic, while a change to text_embeddings_backend_candle.models.flash_nomic can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_nomic uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_nomic uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_nomic, while a change to text_embeddings_backend_candle.models.flash_nomic can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_nomic uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_nomic uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_nomic, while a change to text_embeddings_backend_candle.models.flash_nomic can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_nomic uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_nomic uses text_embeddings_backend_candle.models.nomic, so a change to text_embeddings_backend_candle.models.nomic can break text_embeddings_backend_candle.models.flash_nomic, while a change to text_embeddings_backend_candle.models.flash_nomic can't break text_embeddings_backend_candle.models.nomic.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_nomic uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen3 uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_qwen3, while a change to text_embeddings_backend_candle.models.flash_qwen3 can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen3 uses text_embeddings_backend_candle.layers.rms_norm, so a change to text_embeddings_backend_candle.layers.rms_norm can break text_embeddings_backend_candle.models.flash_qwen3, while a change to text_embeddings_backend_candle.models.flash_qwen3 can't break text_embeddings_backend_candle.layers.rms_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen3 uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_qwen3, while a change to text_embeddings_backend_candle.models.flash_qwen3 can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen3 uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_qwen3, while a change to text_embeddings_backend_candle.models.flash_qwen3 can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_qwen3 uses text_embeddings_backend_candle.models.qwen3, so a change to text_embeddings_backend_candle.models.qwen3 can break text_embeddings_backend_candle.models.flash_qwen3, while a change to text_embeddings_backend_candle.models.flash_qwen3 can't break text_embeddings_backend_candle.models.qwen3.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_qwen3 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gte uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.gte, while a change to text_embeddings_backend_candle.models.gte can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gte uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.gte, while a change to text_embeddings_backend_candle.models.gte can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gte uses text_embeddings_backend_candle.layers.rotary, so a change to text_embeddings_backend_candle.layers.rotary can break text_embeddings_backend_candle.models.gte, while a change to text_embeddings_backend_candle.models.gte can't break text_embeddings_backend_candle.layers.rotary.
Position
Below the diagonal: text_embeddings_backend_candle.models.gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gte uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.gte, while a change to text_embeddings_backend_candle.models.gte can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gte uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.gte, while a change to text_embeddings_backend_candle.models.gte can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.gte uses text_embeddings_backend_candle.models.bert, so a change to text_embeddings_backend_candle.models.bert can break text_embeddings_backend_candle.models.gte, while a change to text_embeddings_backend_candle.models.gte can't break text_embeddings_backend_candle.models.bert.
Position
Below the diagonal: text_embeddings_backend_candle.models.gte uses a module listed above it, which is the healthy direction.
30→36 text_embeddings_backend_candle.models.gte uses text_embeddings_backend_candle.models.flash_gtepart of a loop✕
Connections
2 pairs of types, where a type in text_embeddings_backend_candle.models.gte uses a type in text_embeddings_backend_candle.models.flash_gte.
text_embeddings_backend_candle.models.gte uses text_embeddings_backend_candle.models.flash_gte, so a change to text_embeddings_backend_candle.models.flash_gte can break text_embeddings_backend_candle.models.gte, while a change to text_embeddings_backend_candle.models.gte can't break text_embeddings_backend_candle.models.flash_gte.
Position
Above the diagonal, so it is part of a loop. Neither module can be changed, tested or released on its own until one of these dependencies is removed.
text_embeddings_backend_candle.models.jina uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.jina, while a change to text_embeddings_backend_candle.models.jina can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.jina, while a change to text_embeddings_backend_candle.models.jina can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.jina, while a change to text_embeddings_backend_candle.models.jina can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.jina, while a change to text_embeddings_backend_candle.models.jina can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina uses text_embeddings_backend_candle.models.bert, so a change to text_embeddings_backend_candle.models.bert can break text_embeddings_backend_candle.models.jina, while a change to text_embeddings_backend_candle.models.jina can't break text_embeddings_backend_candle.models.bert.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina uses a module listed above it, which is the healthy direction.
31→37 text_embeddings_backend_candle.models.jina uses text_embeddings_backend_candle.models.flash_jinapart of a loop✕
Connections
3 pairs of types, where a type in text_embeddings_backend_candle.models.jina uses a type in text_embeddings_backend_candle.models.flash_jina.
text_embeddings_backend_candle.models.jina uses text_embeddings_backend_candle.models.flash_jina, so a change to text_embeddings_backend_candle.models.flash_jina can break text_embeddings_backend_candle.models.jina, while a change to text_embeddings_backend_candle.models.jina can't break text_embeddings_backend_candle.models.flash_jina.
Position
Above the diagonal, so it is part of a loop. Neither module can be changed, tested or released on its own until one of these dependencies is removed.
text_embeddings_backend_candle.models.modernbert uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.modernbert, while a change to text_embeddings_backend_candle.models.modernbert can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.modernbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.modernbert uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.modernbert, while a change to text_embeddings_backend_candle.models.modernbert can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.modernbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.modernbert uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.modernbert, while a change to text_embeddings_backend_candle.models.modernbert can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.modernbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.modernbert uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.modernbert, while a change to text_embeddings_backend_candle.models.modernbert can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.modernbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.modernbert uses text_embeddings_backend_candle.models.bert, so a change to text_embeddings_backend_candle.models.bert can break text_embeddings_backend_candle.models.modernbert, while a change to text_embeddings_backend_candle.models.modernbert can't break text_embeddings_backend_candle.models.bert.
Position
Below the diagonal: text_embeddings_backend_candle.models.modernbert uses a module listed above it, which is the healthy direction.
32→39 text_embeddings_backend_candle.models.modernbert uses text_embeddings_backend_candle.models.flash_modernbertpart of a loop✕
Connections
3 pairs of types, where a type in text_embeddings_backend_candle.models.modernbert uses a type in text_embeddings_backend_candle.models.flash_modernbert.
text_embeddings_backend_candle.models.modernbert uses text_embeddings_backend_candle.models.flash_modernbert, so a change to text_embeddings_backend_candle.models.flash_modernbert can break text_embeddings_backend_candle.models.modernbert, while a change to text_embeddings_backend_candle.models.modernbert can't break text_embeddings_backend_candle.models.flash_modernbert.
Position
Above the diagonal, so it is part of a loop. Neither module can be changed, tested or released on its own until one of these dependencies is removed.
text_embeddings_backend_candle.models.pplx1 uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.pplx1, while a change to text_embeddings_backend_candle.models.pplx1 can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.pplx1 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.pplx1 uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.pplx1, while a change to text_embeddings_backend_candle.models.pplx1 can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.pplx1 uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.pplx1 uses text_embeddings_backend_candle.models.qwen3, so a change to text_embeddings_backend_candle.models.qwen3 can break text_embeddings_backend_candle.models.pplx1, while a change to text_embeddings_backend_candle.models.pplx1 can't break text_embeddings_backend_candle.models.qwen3.
Position
Below the diagonal: text_embeddings_backend_candle.models.pplx1 uses a module listed above it, which is the healthy direction.
text_embeddings_router.grpc.server uses text_embeddings_router, so a change to text_embeddings_router can break text_embeddings_router.grpc.server, while a change to text_embeddings_router.grpc.server can't break text_embeddings_router.
Position
Below the diagonal: text_embeddings_router.grpc.server uses a module listed above it, which is the healthy direction.
text_embeddings_router.grpc.server uses text_embeddings_core.infer, so a change to text_embeddings_core.infer can break text_embeddings_router.grpc.server, while a change to text_embeddings_router.grpc.server can't break text_embeddings_core.infer.
Position
Below the diagonal: text_embeddings_router.grpc.server uses a module listed above it, which is the healthy direction.
1 pair of types, where a type in text_embeddings_backend_candle.models.flash_distilbert uses a type in text_embeddings_backend_candle.layers.layer_norm.
text_embeddings_backend_candle.models.flash_distilbert uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.flash_distilbert, while a change to text_embeddings_backend_candle.models.flash_distilbert can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_distilbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_distilbert uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_distilbert, while a change to text_embeddings_backend_candle.models.flash_distilbert can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_distilbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_distilbert uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_distilbert, while a change to text_embeddings_backend_candle.models.flash_distilbert can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_distilbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_distilbert uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_distilbert, while a change to text_embeddings_backend_candle.models.flash_distilbert can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_distilbert uses a module listed above it, which is the healthy direction.
10 pairs of types, where a type in text_embeddings_backend_candle.models.flash_distilbert uses a type in text_embeddings_backend_candle.models.distilbert.
text_embeddings_backend_candle.models.flash_distilbert uses text_embeddings_backend_candle.models.distilbert, so a change to text_embeddings_backend_candle.models.distilbert can break text_embeddings_backend_candle.models.flash_distilbert, while a change to text_embeddings_backend_candle.models.flash_distilbert can't break text_embeddings_backend_candle.models.distilbert.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_distilbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_gte uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.flash_gte, while a change to text_embeddings_backend_candle.models.flash_gte can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_gte uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_gte, while a change to text_embeddings_backend_candle.models.flash_gte can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_gte uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_gte, while a change to text_embeddings_backend_candle.models.flash_gte can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_gte uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_gte, while a change to text_embeddings_backend_candle.models.flash_gte can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_gte uses text_embeddings_backend_candle.models.gte, so a change to text_embeddings_backend_candle.models.gte can break text_embeddings_backend_candle.models.flash_gte, while a change to text_embeddings_backend_candle.models.flash_gte can't break text_embeddings_backend_candle.models.gte.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_gte uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.flash_jina, while a change to text_embeddings_backend_candle.models.flash_jina can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_jina, while a change to text_embeddings_backend_candle.models.flash_jina can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_jina, while a change to text_embeddings_backend_candle.models.flash_jina can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_jina, while a change to text_embeddings_backend_candle.models.flash_jina can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina uses text_embeddings_backend_candle.models.bert, so a change to text_embeddings_backend_candle.models.bert can break text_embeddings_backend_candle.models.flash_jina, while a change to text_embeddings_backend_candle.models.flash_jina can't break text_embeddings_backend_candle.models.bert.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina uses text_embeddings_backend_candle.models.jina, so a change to text_embeddings_backend_candle.models.jina can break text_embeddings_backend_candle.models.flash_jina, while a change to text_embeddings_backend_candle.models.flash_jina can't break text_embeddings_backend_candle.models.jina.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina uses a module listed above it, which is the healthy direction.
2 pairs of types, where a type in text_embeddings_backend_candle.models.flash_jina_code uses a type in text_embeddings_backend_candle.layers.layer_norm.
text_embeddings_backend_candle.models.flash_jina_code uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.flash_jina_code, while a change to text_embeddings_backend_candle.models.flash_jina_code can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina_code uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_jina_code, while a change to text_embeddings_backend_candle.models.flash_jina_code can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina_code uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_jina_code, while a change to text_embeddings_backend_candle.models.flash_jina_code can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina_code uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_jina_code, while a change to text_embeddings_backend_candle.models.flash_jina_code can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina_code uses text_embeddings_backend_candle.models.bert, so a change to text_embeddings_backend_candle.models.bert can break text_embeddings_backend_candle.models.flash_jina_code, while a change to text_embeddings_backend_candle.models.flash_jina_code can't break text_embeddings_backend_candle.models.bert.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_jina_code uses text_embeddings_backend_candle.models.jina, so a change to text_embeddings_backend_candle.models.jina can break text_embeddings_backend_candle.models.flash_jina_code, while a change to text_embeddings_backend_candle.models.flash_jina_code can't break text_embeddings_backend_candle.models.jina.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_jina_code uses a module listed above it, which is the healthy direction.
2 pairs of types, where a type in text_embeddings_backend_candle.models.flash_modernbert uses a type in text_embeddings_backend_candle.layers.layer_norm.
text_embeddings_backend_candle.models.flash_modernbert uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.flash_modernbert, while a change to text_embeddings_backend_candle.models.flash_modernbert can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_modernbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_modernbert uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.flash_modernbert, while a change to text_embeddings_backend_candle.models.flash_modernbert can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_modernbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_modernbert uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.flash_modernbert, while a change to text_embeddings_backend_candle.models.flash_modernbert can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_modernbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.flash_modernbert uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.flash_modernbert, while a change to text_embeddings_backend_candle.models.flash_modernbert can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_modernbert uses a module listed above it, which is the healthy direction.
6 pairs of types, where a type in text_embeddings_backend_candle.models.flash_modernbert uses a type in text_embeddings_backend_candle.models.modernbert.
text_embeddings_backend_candle.models.flash_modernbert uses text_embeddings_backend_candle.models.modernbert, so a change to text_embeddings_backend_candle.models.modernbert can break text_embeddings_backend_candle.models.flash_modernbert, while a change to text_embeddings_backend_candle.models.flash_modernbert can't break text_embeddings_backend_candle.models.modernbert.
Position
Below the diagonal: text_embeddings_backend_candle.models.flash_modernbert uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina_code uses text_embeddings_backend_candle.layers.layer_norm, so a change to text_embeddings_backend_candle.layers.layer_norm can break text_embeddings_backend_candle.models.jina_code, while a change to text_embeddings_backend_candle.models.jina_code can't break text_embeddings_backend_candle.layers.layer_norm.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina_code uses text_embeddings_backend_candle.layers.linear, so a change to text_embeddings_backend_candle.layers.linear can break text_embeddings_backend_candle.models.jina_code, while a change to text_embeddings_backend_candle.models.jina_code can't break text_embeddings_backend_candle.layers.linear.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina_code uses text_embeddings_backend_core, so a change to text_embeddings_backend_core can break text_embeddings_backend_candle.models.jina_code, while a change to text_embeddings_backend_candle.models.jina_code can't break text_embeddings_backend_core.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina_code uses text_embeddings_backend_candle.models, so a change to text_embeddings_backend_candle.models can break text_embeddings_backend_candle.models.jina_code, while a change to text_embeddings_backend_candle.models.jina_code can't break text_embeddings_backend_candle.models.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina_code uses text_embeddings_backend_candle.models.bert, so a change to text_embeddings_backend_candle.models.bert can break text_embeddings_backend_candle.models.jina_code, while a change to text_embeddings_backend_candle.models.jina_code can't break text_embeddings_backend_candle.models.bert.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina_code uses a module listed above it, which is the healthy direction.
text_embeddings_backend_candle.models.jina_code uses text_embeddings_backend_candle.models.jina, so a change to text_embeddings_backend_candle.models.jina can break text_embeddings_backend_candle.models.jina_code, while a change to text_embeddings_backend_candle.models.jina_code can't break text_embeddings_backend_candle.models.jina.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina_code uses a module listed above it, which is the healthy direction.
3 pairs of types, where a type in text_embeddings_backend_candle.models.jina_code uses a type in text_embeddings_backend_candle.models.flash_jina_code.
text_embeddings_backend_candle.models.jina_code uses text_embeddings_backend_candle.models.flash_jina_code, so a change to text_embeddings_backend_candle.models.flash_jina_code can break text_embeddings_backend_candle.models.jina_code, while a change to text_embeddings_backend_candle.models.jina_code can't break text_embeddings_backend_candle.models.flash_jina_code.
Position
Below the diagonal: text_embeddings_backend_candle.models.jina_code uses a module listed above it, which is the healthy direction.
At a glance — Code Health · 55% · Adequate · gated by D2, D4, X4 ·
Findings mapped to OWASP categories; the specific CVEs/secrets are in the Security dimension cards below and findings.md (redacted only on the public version of this report).
OWASP category
Findings
Severity
A03:2025 — Software Supply Chain Failures
100
High / Critical
A02:2025 — Security Misconfiguration
32
High / Critical
A05:2025 — Injection
3
High / Critical
Roadmap
Top priorities: Interpolated log message defeats structured logging; Resolve the 1 No ADRs found finding(s) in ADR Quality — start with Cargo.toml; Record your significant design decisions. Write one short document per decision, giving its date, the situation, the decision and its consequences, and keep them together with your other design documents. A docs/adr folder with each file named by number and title is the easiest place to find them.
Ranked by impact ÷ effort. "Helps" is the estimated gain on the 0–100 health score.
Do this
Helps
Effort
Dimension
Resolve the 1 No ADRs found finding(s) in ADR Quality — start with Cargo.toml.
Record your significant design decisions. Write one short document per decision, giving its date, the situation, the decision and its consequences, and keep them together with your other design documents. A docs/adr folder with each file named by number and title is the easiest place to find them.
Every finding carries one of four grades. Three say how serious it is. The fourth says this
survey could not settle it — and it is a grade, not a gap.
Critical — 61
A definite problem that already costs you something and drags the score down: a
missing authorisation check, a dependency with a known exploit, a build that does not reproduce. Failure here
tends to cause failures elsewhere.
Serious — 471
Likely wrong, but not failing yet. It degrades
the codebase over a longer horizon and can cause failures elsewhere — not urgent this week, not something to
carry for two years either.
Minor — 61
Recorded, with no effect on how the codebase functions.
Present so the survey is complete, not because it needs doing.
Could not be resolved — 21
Something this survey could not settle
from the outside, and which could be critical or serious. Either a control was required and no
positive evidence of it exists in the repository — a backup job that nothing shows was ever restored from proves
nothing about restores — or our own analysis could not run over that part of the tree. This is not a clean
result. These are excluded from the score rather than awarded a pass, so the number on the cover neither
rewards nor penalises them: if you act on this survey without resolving them, you carry that risk yourself. Each
one is named under Limitations.
Methodology & how to trust this report
This survey is a deep, periodic assessment — run each sprint, monthly, or quarterly, taking the time to go wider and deeper than a quick check and surfacing in one coherent report what you'd otherwise piece together from a dozen separate tools. It scores deterministically: the same commit yields the same score, every run. All 41 evaluated dimensions are computed by tools and static analysis; no language model sets or judges any of them. Overall confidence is 0.9, the weighted average across measured dimensions.
Every figure here is one of two kinds, and we label which: ✓ Measured — a deterministic fact (LoC, complexity, coverage); ~ Modeled — an estimate from a stated model (cost, effort, value-at-risk), always a range with its assumptions, never a precise fact. We never present a modelled estimate as if it were measured. Perfect or absent scores carry their provenance too (ADR-0011): ✓ Tool-verified means the property itself was measured across the surface; ○ Nothing flagged means the probes came back clean — a claim bounded by what a repository can show; ⊘ Not evidenced means a working control (a tested restore, an automated rollback) showed no positive evidence — absence of evidence is not evidence of a control, so it's excluded from the score rather than awarded a spurious 10.
What we checked — 41 dimensions across the health lenses
Each chip is a dimension scored from real signals across architecture, testing, dependencies, security & compliance, documentation, git-history and code quality — in one coherent pass. A surface report typically covers a handful.
How to trust any code-health report — three questions
Can you open the finding? Real findings cite a repo-relative file and line you can open at the cited line — never an absolute scratch path. Here, 581 of 593 do; the remainder are repo-wide signals — a dimension-level measurement, not a single line. (Every path in this report is repo-relative by construction: paths are normalized at the producer and the report is rejected if any rooted path leaks through.)
Is there a tool behind the number? Every score below names the method that produced it: Roslyn, git or a scanner, not a narrative.
Does re-running give the same result? Run it again on the same commit and the score — and this report, byte for byte — is identical. A report whose numbers move between runs is describing the run, not the code.
This report answers yes to all three. That's the bar to hold any assessment to.
Tools & methods
The actual versions used this run (captured at analysis time) — re-run on the same commit for the identical score.
Method
Backs
Version
Evaluator
Roslyn static analysis
Complexity, cohesion, coupling, dead code, API surface, layering
What ran differently this time — a tool absent, degraded, or that fell back to an estimate. Named openly, not folded silently into the scores. A degraded run also records its exact cause in diagnostics.md.
D8 Code Coverage — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. Coverage NOT MEASURED: the Rust half could not be measured — the Cargo suite in repository root did not build, so it was never run. Coverage is excluded from the score rather than counted as a near-zero. This is OUR limitation, not a defect in the repo.
D11 Test Reliability — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. Test source is present (.rs) and this repository declares a Cargo test suite (repository root, 38 test files), but it was not re-run: the analyzer environment could not run it. Not scored — this is a gap in the analyzer's language coverage, not a finding about this repository.
D14 License Compliance — measured, with a gap in what it reached — We measured this, but not all of it. What we did not reach is a gap on our side — a collector, parser or image we have not built yet — so the numbers on that dimension cover less than the repository, and the part left out is not evidence that it would have passed. This repository declares a Python pyproject.toml/requirements.txt (pip/uv/Poetry), but the licence verdict published here was taken over its crate dependencies. Nothing was read about its Python dependencies' licensing in either direction, and a clean score on this card must not be read as covering them.
D19 Documentation Quality — measured, with a gap in what it reached — We measured this, but not all of it. What we did not reach is a gap on our side — a collector, parser or image we have not built yet — so the numbers on that dimension cover less than the repository, and the part left out is not evidence that it would have passed. Each ecosystem's manifests (what makes a folder runnable or a published package) need their own proof before D19 judges its units. Until then those folders' READMEs are judged only as documents (boilerplate, structure), never counted as complete.
D21 Naming Consistency — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. D21's naming rules read C# through Roslyn and every other language through its sidecar model; these languages are left out (PHP's reading is its own change, a Roslyn-loaded VB or F# project has no population builder yet). Their declarations carry no naming verdict this run — this says nothing about whether their names are consistent.
D22 Internal API Consistency — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. D22's deterministic call-shape rules are gated on C# first (docs/plans/deterministic-engine.md, P5): each other language is switched on only after its own concept suite and A/B gate pass. This says nothing about whether that surface is consistent.
D24 Comment Value — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. Each source language's comment syntax, and the code a comment annotates in it, needs its own proof before D24 judges it. Until then those comments are left unread, not counted as clean.
D27 Navigability — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. Navigability is taken over call sites whose callee resolved. For C# that population includes calls into referenced assemblies; for this language only calls declared inside the repository count, so the abstention threshold reads the repository's use of its standard library as a resolution failure. That is a limit of this analyzer, not a build problem in this repository.
D34 Knowledge Freshness — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. File-level freshness contradicts repo activity: 34 commits in the last 90 days, yet 33 of 62 significant files carry no living knowledge. Those two readings cannot both be true, so the per-file recency signal is treated as unreliable here and freshness is not scored for this run.
AX6 Interface segregation — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. This check is computed over the public interfaces this run's compilations declare, and none was loaded, so it had nothing to read. That is a gap in this analyzer's language reach — not a finding that the repository is free of what this check looks for.
AXR1 Runtime accessibility — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. Render strategy: boot via ToolchainServer (router). the service did not build — `cargo build --locked --bin 'text-embeddings-router'` exited 1: info: syncing channel updates for 1.92.0-x86_64-unknown-linux-gnu
error: could not download file from 'https://static.rust-lang.org/dist/channel-rust-1.92.0.toml.sha256' to '/usr/local/rustup/tmp/2qkq…; runtime evidence skipped (launch chosen because 'axum' dependency in router/Cargo.toml, binary 'text-embeddings-router').
C1 Data Protection — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. These personal data controls are read from declarative annotations, request middleware, entity/column names and guard methods in a C# source model, and none was loaded on this run, so there was nothing to gather. That is a gap in this analyzer's language reach — not a finding that the repository lacks personal data controls.
C2 Access Controls — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. These authorization controls are read from declarative annotations, request middleware, entity/column names and guard methods in a C# source model, and none was loaded on this run, so there was nothing to gather. That is a gap in this analyzer's language reach — not a finding that the repository lacks authorization controls.
C3 Audit Trail — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. These audit controls are read from declarative annotations, request middleware, entity/column names and guard methods in a C# source model, and none was loaded on this run, so there was nothing to gather. That is a gap in this analyzer's language reach — not a finding that the repository lacks audit controls.
C4 Data Retention — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. These retention controls are read from declarative annotations, request middleware, entity/column names and guard methods in a C# source model, and none was loaded on this run, so there was nothing to gather. That is a gap in this analyzer's language reach — not a finding that the repository lacks retention controls.
C5 Data-Subject Rights — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. These data-subject rights controls are read from declarative annotations, request middleware, entity/column names and guard methods in a C# source model, and none was loaded on this run, so there was nothing to gather. That is a gap in this analyzer's language reach — not a finding that the repository lacks data-subject rights controls.
GD1 Unfinished & placeholder code — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. This check reads C# syntax, F#, JavaScript/TypeScript, Dart, Ruby and Scala source only, and no C#, F#, JavaScript/TypeScript, Dart, Ruby or Scala was found, and this repository's Python, Rust is not read yet, so it had nothing of this repository's product to read. That is a gap in this analyzer's language reach — not a finding that the repository is free of what this check looks for.
IC1 Incompleteness & stubs — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. This check reads C# syntax, JavaScript/TypeScript, Java, Dart, Ruby and Scala source only, and no C#, JavaScript/TypeScript, Java, Dart, Ruby or Scala was found, and this repository's Python, Rust is not read yet, so it had nothing of this repository's product to read. That is a gap in this analyzer's language reach — not a finding that the repository is free of what this check looks for.
P5 DR & Backup — not measured this run — This is a true statement about the repository that carries nothing for its owner to act on, so it is reported here rather than as a defect in their code. No backup/snapshot/replication config, RTO/RPO or restore-procedure documentation was found — and no production persistence was detected either (no data-access packages, no data-store services, no database resources), so there is nothing in this repository whose loss a DR control would recover. If this system's data lives in a platform or ops repo we can't see, that's where the DR evidence belongs.
S1 Web-Security Posture — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. These web-security controls are read from declarative annotations, request middleware, entity/column names and guard methods in a C# source model, and none was loaded on this run, so there was nothing to gather. That is a gap in this analyzer's language reach — not a finding that the repository lacks web-security controls.
X12 Unreachable branch — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. This check's reader does not cover most of the languages this repository's product is written in. What it read is kept, but it is too small a share to grade the product by. That is a gap in this analyzer's language reach — not a finding about this repository.
X16 Unfloored truncation loop — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. This check reads C# syntax, and JavaScript/TypeScript, Java, PHP, Dart, Ruby, Scala and Kotlin source only, and no C# was loaded and no JavaScript/TypeScript, Java, PHP, Dart, Ruby, Scala or Kotlin was found in this repository, so it had nothing of this repository's product to read. That is a gap in this analyzer's language reach — not a finding that the repository is free of what this check looks for.
X18 Disposal-pattern correctness — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. This check reads C# syntax, and JavaScript/TypeScript, Java, Kotlin, Dart, Ruby, Scala and PHP source only, and no C# was loaded and no JavaScript/TypeScript, Java, Kotlin, Dart, Ruby, Scala or PHP was found in this repository, so it had nothing of this repository's product to read. That is a gap in this analyzer's language reach — not a finding that the repository is free of what this check looks for.
X3 Exception handling — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. This check's reader does not cover most of the languages this repository's product is written in. What it read is kept, but it is too small a share to grade the product by. That is a gap in this analyzer's language reach — not a finding about this repository.
X4 Structured logging — measured, with a gap in what it reached — We measured this, but not all of it. What we did not reach is a gap on our side — a collector, parser or image we have not built yet — so the numbers on that dimension cover less than the repository, and the part left out is not evidence that it would have passed. 99 further occurrence(s) are not listed individually; the score already reflects all 139.
X5 Nullable reference types — not measured this run — We could not measure this here. That is a gap on our side — a collector, parser or image we have not built yet — and it is neither a defect in this repository nor evidence that the check would have passed. This check reads C# syntax and MSBuild project configuration, TypeScript source and its tsconfig files, Java source and its package and module declarations, PHP source and its PHPStan and Psalm configuration, Dart source and its pubspec language version, Ruby source and its Sorbet and Steep configuration, Scala source and its build's explicit-nulls, WartRemover and Scalafix settings, and Kotlin source only, and no C# was loaded, no TypeScript was found, and this repository's Python, Rust is not read yet, so it had nothing of this repository's product to read. That is a gap in this analyzer's language reach — not a finding that the repository is free of what this check looks for.
Repo exclusion declarations (.gitattributes linguist-generated/vendored, .editorconfig generated_code): none declared — every source file was scored.
Limitations & what we did not check
We assess the repository exactly as committed, and only the repository. By design we do not reach outside the source tree: the live cloud account, the running CI/CD pipeline, the host's branch-protection and approval rules, the production configuration, or a restore actually exercised against a backup are all out of scope. That boundary is a feature, not a gap — a repo-relative, deterministic scan re-runs identically on any commit and every finding opens at a real file and line, where a live audit can neither be reproduced nor traced. The visible consequence is that controls which leave no in-repo evidence are reported as "not evidenced" and excluded from the score rather than awarded a number a static scan cannot justify.
Per-dimension blind spots
For each dimension that was measured, what a static, repo-only scan structurally cannot see — the honest edge of the measurement, not a failure of it.
D1 Cyclomatic Complexity: Cyclomatic complexity counts branches statically — it cannot tell an essential decision tree from accidental tangle, nor see complexity that lives in data or configuration (large switch-case token tables, DSL lexers/parsers, data-as-code rule tables) rather than control flow: a tokenizer's many single-character cases read as high complexity though each branch is trivial.
D2 Cognitive Complexity: Cognitive-complexity heuristics approximate how hard code is to follow; genuine domain difficulty and well-named intent that eases reading are not captured.
D3 God Classes: "God class" is sized by members and responsibilities visible in the type — a deliberately broad facade over a coherent subsystem can read the same as an accidental grab-bag. For front-end JS the file-length check is cohesion-aware (a single-responsibility module — one class/IIFE — earns a 3× threshold), but cohesion is approximated from top-level declarations, not true dependency structure.
D4 Code Duplication: Duplication is token-similarity — an in-process token-stream comparison over sliding windows, with type-aware normalization — so it finds copy-paste, not semantic duplication expressed differently. Committed machine-written code (scaffolded migrations, designer/codegen output, protobuf/OpenAPI stubs, model snapshots) is EXCLUDED — its repetition is the tool's, not the team's — so the score reflects hand-written duplication only.
D5 Coupling: Coupling is measured between projects/assemblies — runtime coupling through DI, reflection, messaging or shared databases is invisible to a static reference graph.
D6 Cohesion (LCOM4): LCOM4 cohesion is syntactic — it infers connectivity from which methods touch which fields/methods by name, not from real runtime behaviour or intent.
D12 Dependency Hygiene: Dependency health reads manifests and lockfiles — a vulnerability in a vendored/copied dependency, or risk from how a dependency is actually used, is outside this view.
D13 Secret Scanning: Secret detection is signature- and entropy-based on the current tree — a secret that does not match a known pattern, or one already rotated, will not be flagged (a clean scan is "nothing matched", not "no secrets exist").
D14 License Compliance: License compatibility is checked against declared package metadata and a policy — mislabelled or missing license metadata, and obligations that depend on how you distribute, are not resolved here.
D15 Churn × Complexity Hotspots: Churn hotspots come from git history — a freshly imported or squashed repository has no churn signal, and recent rewrites can mask a historically risky file.
D16 Bus Factor: Bus-factor is a time-decayed model of commit attribution (who has recently, repeatedly worked a file), not comprehension — pairing, review and reading-without-committing spread knowledge it can't see; bot commits and shared accounts still distort it.
D17 Explicit Debt: Acknowledged-debt signals (TODO/FIXME, suppressions, dead code) are textual — undocumented debt that nobody marked, and debt that lives in design rather than annotations, is invisible. Committed machine-written code (scaffolded migrations, designer/codegen output, generated stubs) is excluded — it is never the team's dead code to delete.
D19 Documentation Quality: Documentation quality publishes two things only: a runnable or packaged unit with no example, and a multi-project root README that maps none of its parts. It does not judge clarity, and its other checks (overview, licence, boilerplate, structure, empty sections, XML-doc coverage) are withheld after adjudication: their candidates are counted in the card's gaps, not shown as rows. A false statement in a document is M4's; units are .NET projects only.
D20 ADR Quality: ADR quality checks the structure of the decision records present: it cannot know about decisions made and never recorded, and it does not judge how well a record argues its case. Records in a language the section lexicon does not cover are not evaluated.
D26 Project Cohesion: Project focus is sized from members/namespaces per project — a project that is broad by deliberate design reads the same as one that has sprawled.
D28 Secrets (history): Secret-history scanning sweeps the git log for known patterns — a secret that predates the available history, or never matched a signature, is not found (clean means "nothing matched in the history we can see").
D29 Static Analysis (SAST): SAST findings are pattern-based (semgrep) — it finds classes of bug it has rules for; logic flaws, auth/authorization gaps and issues needing runtime context are out of reach (and clean means "no rule matched").
D30 Dependency Vulnerabilities: CVE matching depends on accurate package/version metadata and on the advisory databases — a vulnerability with no published advisory, or in code not declared as a dependency, is not seen. Coverage needs a RESOLVED graph: an unpinned REDACTED, or a pom without a resolved build, yields partial coverage rather than a clean verdict. An ecosystem the analyzer cannot scan is reported as unmeasured, never as clean.
D31 IaC & Container Security: IaC scanning checks Dockerfiles/Terraform/Kubernetes against best-practice rules — it cannot see the live cloud account, runtime configuration, or drift between the committed config and what is actually deployed.
D35 Change Coupling: Change coupling is co-change in COMMITS — files split across separate commits, or coupled only through a shared config/build step, read as uncoupled, and a sweeping commit (rename/format) is excluded so it doesn't couple everything. It shows that files change together, not WHY: a high coupling can be a healthy cohesive pair as readily as a hidden leak.
D43 Malicious Dependencies: Only packages some vulnerability database has already NAMED as malicious are seen — a compromise published in the last hours, or never reported at all, is invisible here, and this dimension reading 10 is not evidence that a dependency is trustworthy. There is no typosquat or dependency-confusion analysis: a package nobody has reported is simply absent from the feeds. Coverage is the dependency scan's: an ecosystem that could not be scanned is disclosed as unmeasured, never as clean.
D44 Platform End-of-Life: The support table is FROZEN, so it goes out of date by losing RECALL: a release that ended support after the table was written is missed until the table is refreshed, and this dimension reading 10 is not evidence that a platform is current. Only platforms the repository DECLARES in a place this pass reads are seen — a runtime named only in a REDACTED (D31's subject), in a CI workflow (D29's), or in a file this pass does not parse (go.mod, a Gemfile ruby directive) is invisible here, which is why a repository declaring none of them abstains rather than scoring. Only frameworks with a PUBLISHED support policy are tracked: React, Flask and Express publish none, so their age cannot be judged and their absence from a report is not a statement that they are supported.
AX10 Code composition: Role is inferred from namespace/folder convention, not semantics — a domain concept living in a folder named "Services" reads as application, and the split is lines-of-code, not business value. The business-logic-share score is a SOFT, FLOORED signal: it contributes to the Architecture lens but is floored at the Critical gate, so an infrastructure-heavy design (a gateway, an ETL, a driver) is legitimately low without being nuked to zero.
AX9 CQS / query purity: Handlers are found by interface/name convention — a query handler using neither is not seen. Mutation is a resolved write/publish invocation (SaveChanges/repository/bus), so a write hidden behind a hand-rolled wrapper, reflection, or a string-keyed service locator resolves to a non-persistence type and isn't flagged; it detects that a query writes state, not whether the write is a legitimate read-side cache update. Clean means "no resolved write/publish in a query body", not a proof of CQS purity.
P4 Deployment & Rollback: Approval/branch-protection rules live in repository settings the scan cannot see — only their in-repo evidence (config files, workflows) is checked, so a control enforced purely in the host's settings reads as "not evidenced".
P6 Release Hygiene: Rollback/observability controls are inferred from repo artefacts (pipelines, dashboards-as-code) — controls configured in external tooling, with no in-repo trace, cannot be credited.
What it measures: How tangled the control flow is — methods with many branches are hard to test and change.
Method: Cyclomatic complexity per method (1 + decision points), computed exhaustively across production source; test projects separated by convention. Deterministic.
+ 21 more group(s) — more in Appendix A; the complete list is findings.md.
What to do
Bring the 2 bodies over 30 down to 30 or less in Cyclomatic Complexity — start with CandleBackend::new (cyclomatic 79), OrtBackend::embed (cyclomatic 33). — This score is capped by its worst body, so a finding fixed alone moves it by almost nothing — the next one down takes its place. Refactoring these 2 together lifts Cyclomatic Complexity from 5.0 to about 6.7/10, projected with the scoring formula itself and assuming each lands exactly at 30; a cleaner split scores higher.
Detailed fixes: d1_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: How hard the code is for a person to follow, beyond raw branching.
Method: Cognitive complexity per method (Sonar-style nesting-penalized score), computed exhaustively over production code, excluding test projects. Deterministic.
+ 23 more group(s) — more in Appendix A; the complete list is findings.md.
What to do
Bring the 3 bodies over 60 down to 60 or less in Cognitive Complexity — start with OrtBackend::embed (cognitive 112), CandleBackend::new (cognitive 95), Qwen3Model::forward (cognitive 62). — This score is capped by its worst body, so a finding fixed alone moves it by almost nothing — the next one down takes its place. Refactoring these 3 together lifts Cognitive Complexity from 3.4 to about 4.8/10, projected with the scoring formula itself and assuming each lands exactly at 60; a cleaner split scores higher.
Detailed fixes: d2_recommendation.md · top locations in Appendix A, every location in findings.md.
Do you agree with this assessment?
D3 · God Classes6.8 / 10Adequate✓ Tool-verified
What it measures: Over-large classes that try to do too much ("god classes").
Method: God-class detection by line and method-count thresholds per logical type (partial classes unified), filtered for generated code and registration/contract false positives. Deterministic.
Resolve the 19 MethodTooLong finding(s) in God Classes — start with lib.rs (3), server.rs (2), gemma3.rs (2). — One of this dimension's main actionable groups (19 warning-level).
Resolve the 13 FileTooLong finding(s) in God Classes — start with lib.rs (4), server.rs (2), debertav2.rs. — One of this dimension's main actionable groups (13 warning-level).
Resolve the 9 FunctionTooLong finding(s) in God Classes — start with server.rs (7), lib.rs (2). — One of this dimension's main actionable groups (9 warning-level).
Detailed fixes: d3_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: How much of the production code is copy-pasted: the share of production lines inside a reported copy (every copy of a duplicated block or of an edited member, each line once), 10 at none and 0 at a quarter of the lines.
Method: Code duplication via token-stream sliding windows with type-aware normalization (locals masked, type names preserved), scored on the share of production lines inside a reported copy (each line once; 10 at none, 0 at a quarter). Deterministic.
218 duplicated block group(s) detected. 34.2 % of the production lines (9,092 of 26,547) are in a reported copy. A further 64 rows report members as variants of one another; the share counts each copied line once, however many rows cite it.
+ 164 more group(s) — more in Appendix A; the complete list is findings.md.
What to do
Resolve the 11 Duplicated block (10 lines × 2) finding(s) in Code Duplication — start with lib.rs (2), bert.rs (2), flash_gte.rs. — One of this dimension's main actionable groups (11 warning-level).
Resolve the 8 Duplicated block (17 lines × 2) finding(s) in Code Duplication — start with server.rs (4), rms_norm.rs, flash_modernbert.rs. — One of this dimension's main actionable groups (8 warning-level).
Resolve the 8 Duplicated block (9 lines × 2) finding(s) in Code Duplication — start with lib.rs (2), bert.rs, flash_jina_code.rs. — One of this dimension's main actionable groups (8 warning-level).
Detailed fixes: d4_recommendation.md · top locations in Appendix A, every location in findings.md.
Do you agree with this assessment?
D5 · Coupling10.0 / 10Exemplary✓ Tool-verified
What it measures: Whether volatile projects sit underneath others that depend on them (so their churn ripples upward), and whether project dependencies form cycles. A widely-depended-on but stable shared/kernel project is healthy, not penalised.
Method: Dependency cycles via elementary-DFS over real .csproj references, plus Martin instability (afferent/efferent) per project. Exhaustive over the reference graph, deterministic.
Coverage: Exhaustive · type-level: afferent/efferent coupling + cycles computed over every production type — the population is all types, not a name convention.
10 production modules (Cargo+Python), 0 dependency cycle(s), 0 unstable depended-on module(s). Read from the build's own module declarations; 0 module(s) off the main sequence.
What it measures: Whether a class's methods are focused on a single responsibility.
Method: LCOM4 cohesion per production class with at least two methods: connected components of methods sharing state or calls, computed syntactically. Deterministic, not a proxy.
Coverage: Exhaustive · type-level: LCOM4 cohesion computed over every production class — the population is all types, not a name convention.
What it measures: Whether dependencies are current, not deprecated, and not bloated. Known vulnerabilities are D30's.
Method: Manifest scan via dotnet list package across all projects; worst-signal-per-package deduction (capped-linear for deprecation, prerelease and outdated) per KLoC. Known-vulnerable packages are D30's and are not charged here (rubric-2026.11.2). Exhaustive.
5 outdated, 0 yanked direct Cargo dependencies. 58 of 59 direct crates were graded against crates.io (0 not published there, 1 not resolved by a committed REDACTED). Only DIRECT edges are graded: a transitive crate cannot be moved past what its parent's requirement admits, so reporting one would be advice its owner cannot take. A newer release is reported only where this repository's OWN requirement already admits it, so the remedy is `cargo update` and never a manifest edit — which means a release outside the declared range is deliberately NOT charged, because a written-down constraint is a decision rather than a defect. Note that a bare requirement is a CARET, and for a 0.x crate its ceiling is the minor. Whether any crate is UNMAINTAINED is not graded — crates.io publishes no maintenance status, and release age does not stand in for one. Known CVEs in this dependency graph are D30's question, read from REDACTED there.
Outdated: clap · ×5
✓ On the Gold path — maintain.
Detailed fixes: d12_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Whether any secrets (keys, tokens, passwords) have leaked into the code.
Method: In-process native secret scanner (entropy plus signature patterns) across all tracked files; no external tool. A clean result is a measured 10, not no-data zero. Deterministic.
What it measures: Whether the licenses of third-party packages are compatible with your policy.
Method: Third-party package licenses resolved from declared package metadata and checked against the configured policy (allow/deny/copyleft). Deterministic; clean = no incompatible license found at metadata depth.
0 of 581 shipped crate(s) use a banned license. Licences were resolved from crates.io over the crates a consumer compiles — this repository's REDACTED closed over its manifests' `[dependencies]` and `[build-dependencies]`. Crates it asks for ONLY under `[dev-dependencies]` are excluded: they are not compiled by anything that depends on this repository. A further 7 crate(s) in the closure resolve from git, a path or a private registry and publish no licence this pass can read; they are outside this verdict. ★ COVERAGE OF THIS VERDICT: it grades this repository's crate dependencies and nothing else. The repository also declares a Python pyproject.toml/requirements.txt (pip/uv/Poetry), and the licences of those dependencies were NOT read by this pass — a gap in this engine's coverage, not a statement about them. So this result says the graded closure carries no banned licence; it does NOT say this repository's licensing is clear.
What it measures: Files that change often and are also complex — the riskiest hotspots.
Method: Per production file churn times cyclomatic complexity over a rolling window, computed from git and Roslyn/JS/Razor analysis. Exhaustive, deterministic per commit date.
Resolve the 3 Hotspot finding(s) in Churn × Complexity Hotspots — start with lib.rs, server.rs, qwen3.rs. — One of this dimension's main actionable groups (3 warning-level).
Detailed fixes: d15_recommendation.md · top locations in Appendix A, every location in findings.md.
Do you agree with this assessment?
D16 · Bus Factor9.7 / 10Exemplary✓ Tool-verified
What it measures: Whether knowledge is concentrated in too few people (the "bus factor").
Method: Living knowledge per author via time-decayed commit attribution (6-month half-life, focus weighting) across largest source files. Deterministic, avoids blame's mechanical-refactor false positives.
2 source file(s) have their living knowledge concentrated in one author (≥90% of recent, decayed contribution). The largest is router/src/prometheus.rs. Counted over 62 of the 84 production source files in this repository: the rest are under the ~2,400-byte size floor this dimension measures over.
Off-boarding risk: anonymized user #1
Further sole-owners (lower concentration)
✓ On the Gold path — maintain.
Detailed fixes: d16_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Acknowledged debt left in the code — TODOs, dead code, suppressed warnings.
Method: Roslyn syntactic debt markers (suppressions/TODO/FIXME/HACK/empty-catch/commented-code/Obsolete) plus SymbolFinder dead-code analysis; weighted-debt-per-KLoC density deducted 2.0x per unit. Deterministic, exhaustive.
10 deducted task-comment markers across 26547 LoC (0.0/KLoC) → score 9.9. Task comments only: this repository's language is read without a compiler, so D17's suppression, dead-code and commented-out-code arms did not run and this score counts fewer marker kinds than a .NET repository's would.
TodoComment · ×9backends/src/lib.rs:527
HackCommentrouter/src/main.rs:237
What to do
Resolve the 9 TodoComment finding(s) in Explicit Debt — start with lib.rs (6), qwen3.rs (2), debertav2.rs. — One of this dimension's main actionable groups (9 warning-level).
Resolve the 1 HackComment finding(s) in Explicit Debt — start with main.rs. — One of this dimension's main actionable groups (1 warning-level).
Detailed fixes: d17_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Whether the project's documentation is clear, complete, and useful.
Method: Every README, docs/ document and root guide parsed by its markup and measured in Unicode letters. Published: a runnable or packaged .NET project whose README (or a document it links, or a root guide) shows no example, and a root README naming none of three or more projects. Overview, licence, boilerplate, wall-of-text, empty-section and XML-doc rows are measured and withheld (counted in gaps) after adjudication. Deterministic, exhaustive, advisory.
1 of 1 documentation facets pass in full. 17 document(s) read across 1 unit(s).
✓ On the Gold path — maintain.
Detailed fixes: d19_recommendation.md.
Do you agree with this assessment?
D20 · ADR Quality0.0 / 10Critical✓ Tool-verified
What it measures: Whether architecture decisions are recorded well (context, decision, consequences).
Method: Every ADR read by its section structure: nine located rules (a missing Decision, Context or Consequences section, a stale status, two records deciding the same question, a dangling supersede link, a number-only title, a single option, no named cost, no status where the log uses one, no enforcement field where the log adopted one), each a fixed deduction from 10; the score is the mean over evaluated records. Deterministic, no model. Advisory.
Resolve the 1 No ADRs found finding(s) in ADR Quality — start with Cargo.toml. — One of this dimension's main actionable groups (1 recommendation-level).
Detailed fixes: d20_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Whether any secrets were ever committed — scanned across the full git history, not just now.
Method: Secret scan via TWO gitleaks detect passes in an isolated checkout — the full git history, then a second --no-git pass over the working tree as it stands — merged and de-duplicated by (rule, file, line); each match flagged High. Both invocations are recorded in the audit trail. Exhaustive; when the tool is absent, or when its output cannot be parsed into the expected shape, the dimension is WITHHELD as an explicit measurement gap on our side — unscored and excluded from the lens, never a hedged middling score.
What it measures: Real static-analysis (SAST) findings — likely security bugs in the code, any language.
Method: Polyglot static analysis via semgrep across the repo using the pinned, image-baked p/security-audit + p/owasp-top-ten rulesets (no scan-time registry fetch); severity rules (ERROR/WARNING/INFO) map to a full-band severity-weighted score. Exhaustive, deterministic; degrades on parse failure.
Coverage: semgrep pattern rules over all files — exhaustive for the rule set, blind to classes of bug without a rule (clean = no rule matched).
3 finding(s): 0 critical, 2 high, 1 medium, 0 low. semgrep hit a parse error in 2 file(s) — `cuda-all-entrypoint.sh`, `cuda-entrypoint.sh` — so no absence of findings in them is evidence of anything, and nothing in them was analysed. Fix the syntax error (or exclude the file deliberately) and re-scan to cover them.
REDACTED
REDACTED
REDACTED
What to do
Resolve the 1 REDACTED finding(s) in Static Analysis (SAST) — start with REDACTED. — One of this dimension's main actionable groups (1 issue-level).
Resolve the 1 REDACTED finding(s) in Static Analysis (SAST) — start with REDACTED. — One of this dimension's main actionable groups (1 issue-level).
Resolve the 1 REDACTED finding(s) in Static Analysis (SAST) — start with REDACTED. — One of this dimension's main actionable groups (1 warning-level).
Detailed fixes: d29_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Whether any dependency has a known published vulnerability (CVE), direct or transitive, in ANY ecosystem the repository declares — Dart pub, Elixir and Erlang via Hex, Go modules, Java and Kotlin via Maven/Gradle, JavaScript/npm, .NET/NuGet, PHP/Composer, Python/PyPI, RubyGems, Rust/Cargo and Swift.
Method: Dependency-CVE scan across every ecosystem the repository declares, scored ONCE. Three sources are unioned and deduplicated by advisory identity (rule id + alias closure, CVE<->GHSA) scoped to package+version, keeping the worst severity: `osv-scanner --recursive` over osv.dev for Dart pub, Elixir/Hex (and Erlang, whose `rebar.lock` syft first converts to a CycloneDX SBOM the scanner reads, with rows attributed back to the lock), Go, Java and Kotlin via Maven/Gradle (and Scala, whose sbt build's pinned direct declarations are written into a CycloneDX SBOM the scanner reads, with rows attributed back to the build file), npm, PHP/Composer, Python/PyPI, RubyGems, Rust/Cargo and Swift; `trivy fs --scanners vuln` for npm lockfiles; and `dotnet list package --vulnerable --include-transitive` for NuGet (with per-advisory collapse of the project x target-framework fan-out), plus a DECLARED-dependency arm that resolves a published gem's gemspec against rubygems.org where no Gemfile.lock is committed. `SeverityScore(c,h,m,l, normalizer 8.0)`. NotApplicable only when NO ecosystem is readable; if any applicable ecosystem could not be scanned the findings are REPORTED and the score is withheld. Supersedes the npm and OSV arms, retired 2026-09-05. Advisory since rubric-2026.11.2 (CAI decision 0006): every row is a match against a vulnerability database that changes daily, so the reading is shown and never moves a scored dimension, lens or the headline.
+ 2 more group(s) — more in Appendix A; the complete list is findings.md.
What to do
Resolve the 29 High CVE finding(s) in Dependency Vulnerabilities — start with REDACTED (6), REDACTED (6), REDACTED (6). — One of this dimension's main actionable groups (29 issue-level).
Resolve the 8 Critical CVE finding(s) in Dependency Vulnerabilities — start with REDACTED (3), REDACTED (2), REDACTED. — One of this dimension's main actionable groups (8 issue-level).
Resolve the 5 Critical vulnerability finding(s) in Dependency Vulnerabilities — start with REDACTED, REDACTED, REDACTED. — One of this dimension's main actionable groups (5 issue-level).
Detailed fixes: d30_recommendation.md · top locations in Appendix A, every location in findings.md.
Resolve the 17 High IaC finding(s) in IaC & Container Security — start with REDACTED (4), REDACTED (3), REDACTED (3). — One of this dimension's main actionable groups (17 issue-level).
Resolve the 14 Medium IaC finding(s) in IaC & Container Security — start with REDACTED (4), REDACTED (2), REDACTED (2). — One of this dimension's main actionable groups (14 warning-level).
Resolve the 1 Low IaC finding(s) in IaC & Container Security — start with REDACTED. — One of this dimension's main actionable groups (1 recommendation-level).
Detailed fixes: d31_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Whether files that change together actually belong together — pairs that repeatedly co-change in git history despite having no explicit code dependency, surfacing the hidden/logical coupling (and boundaries in the wrong place) a static scan can't see.
Method: Pairwise co-occurrence over the per-commit file sets in git history (production source only — tests and generated dropped): Degree-of-Coupling = shared ÷ min individual revisions, reported above noise floors (each file ≥10 revisions, ≥5 shared commits, ≥50% strength); sweeping commits excluded. Deterministic over fixed history.
Coverage: Population: PRODUCTION source files only — test and generated files are dropped before pairing, so a class co-changing with its own test (trivially ~100%) can't drown the real production↔production coupling. Pairs ranked by Degree-of-Coupling. A non-source file is never a coupling PARTICIPANT either: documentation, schemas, config and data files are dropped with the rest, so a code↔docs pair — a command and the reference page that restates it — is not reported however strongly the two co-change; nor is coupling that runs THROUGH a build step or config file.
What it measures: Whether the build pipeline provides supply-chain integrity — generated provenance/attestation, signed artifacts (cosign/sigstore), an SBOM, and pinned build actions. Presence of the configuration, not a runtime guarantee.
Method: Supply-chain provenance/signing read deterministically from CI/build config (.github/workflows, .gitlab-ci.yml, azure-pipelines, Jenkinsfile, .circleci) + the release surface: four signals — generated provenance/attestation (SLSA/in-toto/actions-attest), artifact signing (cosign/sigstore/gitsign), an SBOM (syft/sbom-action/*.spdx.json/*.cdx.json), and SHA-pinned build actions — scored 10·present/denom. NotApplicable without a build pipeline. Detects configuration presence, not runtime enforcement.
What it measures: Whether any dependency the repository declares is published as MALICIOUS rather than merely vulnerable — a package that is an attacker's work, in any ecosystem osv-scanner reads. Reported apart from D30 because the answer is binary: there is no safe version to upgrade to, and the fix is to remove the package and rotate every credential it could have read.
Method: The same dependency scan D30 reads, partitioned on the scanner's own classification rather than rescanned: a row is MALICIOUS when its id is in the `MAL-` space (the ossf/malicious-packages feed) OR its `database_specific.cwe_ids` carries `CWE-506` ("Embedded Malicious Code"). Both channels are structural; the summary text is deliberately NOT read, because a malicious-package record whose summary says only "Critical severity vulnerability" is a real shape ([GHSA redacted]) and a text matcher misses it. Scored BINARY: any surviving row is 0, whatever its severity and however many CVEs sit beside it — a hostile dependency is not a quantity. Applicability and degradation are D30's: NotApplicable only when no ecosystem is readable, and an unscannable ecosystem degrades rather than reading clean. Advisory since rubric-2026.11.2 (CAI decision 0006): the malicious-package feed is a database that changes daily, so the reading is shown and never moves a scored dimension, lens or the headline.
What it measures: Whether anyone still ships security patches for the platform this repository RUNS ON — the runtime it pins and the framework majors its own constraints hold it to. Separate from D12 because the question differs: a current Django on an end-of-life Python is perfectly up to date and completely unsupported, and the fix is a migration rather than a version bump. What the repository says it merely SUPPORTS is never charged.
Method: End-of-life PLATFORM read from the repository's own declarations and graded against a FROZEN, dated table of vendor support dates — no network, no feed, no API, so this dimension answers identically inside a closed scan fence. Two subjects: a RUNTIME the project pins (a single or all-end-of-life TargetFramework, a .nvmrc or .python-version, a requires-python CAP) and a FRAMEWORK major a dependency constraint cannot move off (a caret, tilde or exact version; `vue@^2.7.16` pins Vue 2). A FLOOR is deliberately never charged — `requires-python = ">=3.8"` states what a package SUPPORTS, not what it runs on — and a multi-target project is charged only when EVERY target is out of support. Runtime 4.0/product capped 8.0, framework 1.5 capped 4.5. The table is safe to freeze because a statement about support that ended in the past cannot become false: it loses recall as it ages, never precision, and a test asserts every entry predates the freeze date. Disjoint from D31 (a container image's OS layer) and D29 (the toolchain a CI workflow installs). Abstains when the repository declares no platform this pass reads — never scores it clean.
1 end-of-life runtime(s) and 0 end-of-life framework(s), read from 1 platform declaration(s) and 81 dependency declaration(s). This dimension reads what the repository says about ITSELF — a pinned target framework, a version file, a capped requires-python, a Rust toolchain pin, a framework major a constraint cannot move off. A FLOOR is deliberately never charged: `requires-python = ">=3.8"` states what the package SUPPORTS, not what it runs on, and a well-maintained library declares exactly that while running its own CI on a current release. The end-of-life facts are FROZEN and dated, so this dimension needs no network and answers identically inside a closed scan fence; as the table ages it loses recall and never precision, because a statement about support that ended in the past cannot become false. The OS layer of a container image is D31's question and the toolchain a CI workflow installs is D29's; this row is neither.
Resolve the 1 End-of-life runtime finding(s) in Platform End-of-Life — start with rust-toolchain.toml. — One of this dimension's main actionable groups (1 warning-level).
Detailed fixes: d44_recommendation.md · top locations in Appendix A, every location in findings.md.
Other · Architecture — How the codebase splits by code ROLE — domain, application, infrastructure, test, generated. The significance map behind the knowledge/coupling weighting, and a DDD signal in its own right: a thin domain core under fat infrastructure is the anemic-domain smell, quantified. How each file's role is decided, because the split is only as good as that: a generated name or a build-output tree makes it Generated, a test project makes it Test, and otherwise the file's NAMESPACE and PATH words are matched against fixed vocabularies in a fixed ORDER — domain, then infrastructure, then application — so a file whose words hit two layers is counted under the earlier one. A production file matching none of them counts as application, so that share reads 'application or unclassified' rather than a measured application layer. Roles come from naming convention, never from what the code does. On this repository the split was taken from the source tree on disk rather than from a loaded .NET workspace, so a file's role is decided by its PATH segments alone — no declared namespace was available to add to the evidence — and generated output is excluded from the census entirely rather than counted as a generated share.
Method: Roslyn line-count by code ROLE: every source file classified Domain/Application/Infrastructure/Test/Generated by namespace + path convention (the shared CodeRoleClassifier), then significant lines summed per role. Deterministic; the advisory score is the business-logic (domain+application) share of production code.
Coverage: Population: ALL source files, each bucketed into ONE of five roles (Domain/Application/Infrastructure/Test/Generated) by namespace + path convention — a file whose layer isn't named in the convention falls to Application (the neutral default), and the split is line-count, not semantic depth or business value.
Other · Architecture — Whether the project-reference graph is acyclic (cycles block independent build/deploy and signal eroding boundaries).
Method: Project reference cycles via elementary-DFS over real .csproj references, using the engine shared with D5/D7; cyclic versus acyclic. Exhaustive, deterministic.
Other · Architecture — Whether dependencies point inward (Domain ← Application ← Infrastructure/Web) — the clean-architecture dependency rule, checked across the project graph.
Method: Layer violations by name-segment inference (Domain/Core to Application to Infrastructure/Web) over the project-reference graph. Exhaustive over all projects, deterministic.
Other · Architecture — Whether read (query) handlers stay side-effect-free — a query that writes persistent state or raises events breaks CQS and makes reads unsafe to retry, cache, or route to a read replica.
Method: Roslyn scan: CQRS handlers classified query-vs-command by interface (IQueryHandler/ICommandHandler/IRequestHandler<TQuery,TResult>) and name convention (*Query/Get*/Find* vs *Command); each query handler's body checked for persistent-state writes (SaveChanges/repository Add-Update) or event publishes by resolved invocation. Deterministic, type-level, exhaustive over the detected handlers.
Coverage: Population: CQRS handlers identified by IQueryHandler/ICommandHandler/IRequestHandler interface + *Query/Get*/Find*/*Command NAME convention; query purity then checked exhaustively within that set — a query handler using neither convention is invisible, and mutation is a resolved persistence/publish CALL, not full dataflow.
Maturity · Maturity — Whether the repository and its projects have a README, and whether it is substantive.
Method: Filesystem scan: README presence, word count, and headings for depth; git history for staleness. Exhaustive across root and project dirs, deterministic.
What to do
Explain in the README how to run the tests. Add a section called Testing to the README in the repository's top folder, saying how to run the test suite.
Maturity · Maturity — Whether key decisions (ADRs) and the high-level shape (C4/diagrams) are written down.
Method: Filesystem scan: ADR folder/naming conventions or content, plus Mermaid/PlantUML/C4/architecture.md discovery. Exhaustive, deterministic.
No Architecture Decision Records found — no conventional ADR directory, no numbered `NNNN-title` documents in any markup this check reads, and nothing ADR-shaped by content. Design rationale recorded elsewhere (a design-notes tree, a mailing list, pull-request discussion) is not visible to this check and is not re-findable per decision, so a future maintainer cannot ask why one choice was made and get an answer.
What to do
Record your significant design decisions. Write one short document per decision, giving its date, the situation, the decision and its consequences, and keep them together with your other design documents. A docs/adr folder with each file named by number and title is the easiest place to find them.
Maturity · Maturity — Whether the repository is organised deliberately: production code and tests are set apart in a way its ecosystem or the repository itself declares.
Method: Filesystem scan: src/test folder separation and namespace-prefix consistency (majority RootNamespace agreement). Exhaustive across projects, deterministic.
Readiness · Readiness — Whether an automated pipeline builds and tests every change.
Method: Filesystem scan: CI workflow files (.github/workflows, .gitlab-ci.yml, etc.) for build and test stages. Exhaustive, deterministic.
Do you agree with this assessment?
P10 · Library API & versioning10.0 / 10Exemplary○ Nothing flagged
Readiness · Readiness — For a library: a deliberate (small) public API surface and explicit semantic versioning so consumers can depend on it safely.
Method: Roslyn scan: public API surface area and semantic-versioning markers (SemVer attributes, changelog entries) for libraries; off .NET, a library is the ecosystem's publication act (an npm package that is not private and names an entry point, a PyPI distribution with a build system, a Rust library crate, a Maven/Gradle module that publishes, a Go module with no package main, a gemspec, a Composer library, a SwiftPM library product, a pub.dev or Hex package), its surface is the share of types the language's code model records as public (Rust, Swift, Java, Kotlin, Go, Dart; not measured where the model records no type visibility or, as in TypeScript, only module-level export), and its version is read from the manifest, a semver CHANGELOG, release tooling or semver git tags. Exhaustive, deterministic.
Readiness · Readiness — Whether SAST, secret/dependency scanning and performance benchmarking are wired in (presence, not runtime).
Method: Filesystem scan: SAST configuration, dependency-update automation, secret scanning, and a benchmark harness or benchmark step — in this repository's own ecosystem. Exhaustive, deterministic.
`.github/dependabot.yml` schedules updates, but no `package-ecosystem` entry covers `cargo`, `uv`, `gitsubmodule`, which this tree depends on — so those dependencies get no update pull requests and no advisory pressure, however complete the dep-scan credit looks. Add the missing entries to the `updates:` list. — .github/dependabot.yml:2
What to do
Dependabot is configured but does not watch `cargo`, `uv`, `gitsubmodule` — add those `package-ecosystem` entries to .github/dependabot.yml so those dependencies get the same automatic update and advisory pressure as the ones it already covers.
Readiness · Readiness — Whether releases are automated and safely reversible (probes, rolling updates, approval gates) — from manifests/pipeline files, not the live environment.
Method: Filesystem scan: deployment manifests/IaC (K8s YAML, Helm, Terraform) for rolling updates, probes, approval gates, migration hooks. Exhaustive, deterministic.
What to do
Nothing pauses a release for a human: publish as a draft release (or gate the release job on a protected tag/manual dispatch) so a bad build can be stopped before users can download it.
Readiness · Readiness — Whether releases are traceable — a maintained changelog and explicit version stamping.
Method: Filesystem scan: changelog file presence and version tags in csproj or git tags. Exhaustive, deterministic.
No CHANGELOG/HISTORY/RELEASES file — what shipped when isn't easy to reconstruct for support or audit. (Versioning/tagging makes releases traceable, but a changelog records the what.)
What to do
Keep a changelog (e.g. Keep-a-Changelog) recording what shipped in each release.
Readiness · Performance — Whether asynchronous code stays responsive — it avoids sync-over-async blocking (a .NET .Wait()/.GetAwaiter().GetResult(), a time.sleep or blocking HTTP call inside a Python coroutine, a *Sync call inside an async JavaScript function, block_on inside a Rust async fn, runBlocking inside a Kotlin suspend function, block() inside a Reactor publisher) that stalls a thread or event loop and risks deadlock, and, where the code is a reusable library on .NET, awaits with ConfigureAwait(false) so it never captures and stalls its caller's context.
Method: Production-source scan: sync-over-async blocking counted everywhere — .Wait()/.GetAwaiter().GetResult() in .NET; off .NET, read from each language's code model, a blocking call inside an async function (Python, TS/JS, Rust, Kotlin, Swift, Dart) or inside a Java method returning a Reactor Mono/Flux or a Scala method returning a Future or effect — and, for a .NET library with ≥5 awaits, the share of awaits using ConfigureAwait(false). Deterministic, syntax/text detection.
Other · Code Health — Whether the code avoids sync-over-async (deadlock-prone blocking on tasks) and async void.
Method: Roslyn syntax scan: async methods scanned for .Wait()/.GetAwaiter().GetResult() and async-void outside event handlers. Deterministic, hard fact per invocation.
Other · Code Health — Whether a `when` guard is free of side effects — a guard that increments a counter or assigns while deciding whether its arm matches applies that change during PATTERN MATCHING, on an arm that may not be selected, and skips it entirely when a short-circuit to its left answers first.
Method: Roslyn syntax: `when` guards on case labels and switch-expression arms, read for a mutation (`++`/`--`/assignment) sitting in a position the guard's own `&&`/`||`/`??`/`?:`/`?.` can skip. Deterministic, provable per finding. Advisory.
Do you agree with this assessment?
X30 · Support guard that admits what it rejects10.0 / 10Exemplary○ Nothing flagged
Other · Code Health — Whether a guard written as a NEGATED `||` says what its author meant — `!(a || b || x != k)` is `!a && !b && x == k` by De Morgan, so a bail-out that mixes capabilities the code needs with a fault it refuses turns inside out: it fires only where the capabilities are ABSENT, and lets every value the fault term names walk straight into the body that cannot handle it.
Method: Roslyn syntax only, no semantic model: every logical-not whose operand is a parenthesised `||` chain of two or more disjuncts, flattened (a left-nested `a || b || c` read once would see `(a || b)` as one disjunct). A site enters the population on that shape alone. A finding additionally needs the disjuncts to DISAGREE in polarity: at least one bare boolean read — an identifier or member access, never an invocation, which is a predicate rather than a capability flag — and at least one `x != <constant>`, the only form that negates into an exact-value pin (`== null` negates into a looser requirement and is outside the fault set). Consistently-polarised disjunctions, all-fault or all-capability, are counted and never reported; a negated `&&` is outside the population entirely. No same-receiver gate: it was measured to cost a real defect and remove no false positive. Deterministic, provable per finding. Advisory.
Other · Code Health — Whether log calls use message templates (queryable) rather than interpolated strings.
Method: Roslyn syntax scan: every log call-site counted; interpolated-string first-argument violations flagged. Population is all log calls, not estimated. On a repository with no .NET source it reads JavaScript/TypeScript off the engine’s own token stream (test, vendored and minified paths not) with the same rule: a `trace`/`debug`/`info`/`warn`/`error`/`fatal`/`verbose`/`silly`/`http`/`log` call on a receiver named for a logger, in a package whose own or an enclosing `package.json` declares a logger that carries values as fields (pino, pino-http, nestjs-pino, Fastify, winston, bunyan, tslog, LogTape, roarr), whose first deciding argument is a template literal with substitutions; an object-literal fields argument is stepped over, a plain string message clears the call. Python is read the same way: a `debug`/`info`/`warning`/`error`/`exception`/`critical`/`log` call on a logger in a file that imports the standard library’s `logging` or structlog (loguru, which formats at the call, is not judged), whose message is a whole f-string with a non-constant field; a `%`-formatted, `.format()`-ed or concatenated message is not convicted, as in C#. PHP is read the same way: a PSR-3 level call (`emergency`…`debug`, `log`) on a logger (`$logger`, `$this->logger`, Laravel’s `Log::`, `logger()`, `\Drupal::logger()`) in a package whose `composer.json` requires a PSR-3 logger or a framework that hands one out (psr/log, Monolog, Laravel, Symfony, Drupal, CakePHP), or a file that imports one, whose message is a double-quoted string interpolating a variable; a single-quoted message clears the call, and `sprintf` or `.` concatenation is not convicted, as in C#. Dart is read the same way: a `trace`/`debug`/`info`/`warn`/`error`/`fatal` call on Sentry’s structured logger (`Sentry.logger`) whose message is a whole string literal interpolating a value (`$name`, `${…}`); the `Sentry.logger.fmt` template API clears the call. Dart that logs only through loggers with one message and no fields (package:logging, package:logger, talker, dart:developer) is not applicable rather than a reach gap. Ruby is read the same way: a `trace`/`debug`/`info`/`warn`/`error`/`fatal` call on the console gem’s `Console` (declared by a Gemfile, gemspec or Gemfile.lock, or required) or, where SemanticLogger, Ougai or dry-logger is declared or required, on a receiver named for a logger (`logger`, `@logger`, `Rails.logger`, `log`), whose first deciding argument is, as the whole argument, a double-quoted string interpolating a non-constant value (`#{…}`, `#@ivar`); a plain literal clears the call, keyword arguments are the fields, and a block, `%`-format or `+` concatenation is not convicted. Ruby that logs only through the standard Logger (Rails.logger) is not applicable rather than a reach gap. Deterministic.
`info` at line 25 is handed an f-string as its message, so the logger receives one finished string — the values in it cannot be queried, filtered or aggregated as fields, every occurrence reads as a different message, and the string is built even when the record is filtered out. Pass the values beside a constant message instead: `logger.info("order %s moved to %s", order_id, status)` with the standard library's `logging`, which keeps the message and its arguments apart on the record, or `log.info("order moved", order_id=order_id, status=status)` with structlog. — backends/python/server/text_embeddings_server/utils/kernels.py:25
`warning` at line 34 is handed an f-string as its message, so the logger receives one finished string — the values in it cannot be queried, filtered or aggregated as fields, every occurrence reads as a different message, and the string is built even when the record is filtered out. Pass the values beside a constant message instead: `logger.warning("order %s moved to %s", order_id, status)` with the standard library's `logging`, which keeps the message and its arguments apart on the record, or `log.warning("order moved", order_id=order_id, status=status)` with structlog. — backends/python/server/text_embeddings_server/utils/kernels.py:34
`warn!` at line 229 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `warn!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/lib.rs:229
`info!` at line 236 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/lib.rs:236
`warn!` at line 319 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `warn!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/lib.rs:319
`warn!` at line 320 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `warn!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/lib.rs:320
`info!` at line 360 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/lib.rs:360
`info!` at line 235 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/main.rs:235
`error!` at line 208 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:208
`error!` at line 320 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:320
`error!` at line 342 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:342
`error!` at line 384 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:384
`error!` at line 506 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:506
`error!` at line 522 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:522
`error!` at line 639 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:639
`error!` at line 655 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:655
`error!` at line 826 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:826
`error!` at line 842 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:842
`error!` at line 1004 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1004
`error!` at line 1020 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1020
`warn!` at line 1159 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `warn!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1159
`error!` at line 1215 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1215
`error!` at line 1231 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1231
`error!` at line 1391 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1391
`error!` at line 1407 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1407
`error!` at line 1471 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1471
`error!` at line 1487 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1487
`info!` at line 1818 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1818
`info!` at line 1823 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1823
`info!` at line 1901 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1901
`info!` at line 1931 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1931
`info!` at line 1941 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/http/server.rs:1941
`error!` at line 888 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/grpc/server.rs:888
`error!` at line 903 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/grpc/server.rs:903
`error!` at line 911 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/grpc/server.rs:911
`error!` at line 956 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/grpc/server.rs:956
`error!` at line 1081 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/grpc/server.rs:1081
`error!` at line 1089 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/grpc/server.rs:1089
`error!` at line 1272 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/grpc/server.rs:1272
`info!` at line 1507 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`. — router/src/grpc/server.rs:1507
The score is the rank-weighted fold of these lenses (worst-heaviest), each including its meta-dimensions; a lens with a Critical contributor is capped at Fair (its band reads "gated by …") and is never the strongest area however high its average.
Unscored — 1 check(s) recorded observations but carry no score
These checks ran and found something, but they do not carry a score — either by design (an advisory check reports evidence rather than grading it) or because they could not be scored here. They are excluded from the score for that reason, not because there was nothing to see.
X10 Duplicated predicate — 6 observation(s) recorded · Advisory — this card reports evidence and never carries a score.
Not evidenced — 4 control(s) we could not find positive evidence for
These checks grade a working control, and the repository shows no evidence of one. That is deliberately not scored as a zero: a repository cannot show an ops runbook, a database TTL or an infrastructure-side audit log, so absence of evidence here is not evidence the control is missing. It is also not a statement that the check is irrelevant to this codebase — the thing it grades applies; we just could not see it. Excluded from the score either way.
C3 Audit Trail — Not assessed: these audit controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks audit controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
C4 Data Retention — Not assessed: these retention controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks retention controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
C5 Data-Subject Rights — Not assessed: these data-subject rights controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks data-subject rights controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
P5 DR & Backup — not evidenced — repo shows no backup/RTO/RPO controls; absence of evidence is not evidence of a working control
Not included — 71 check(s) not relevant to this codebase
These checks had nothing to measure here (no tests, no git history, the codebase is small, or the architecture style doesn't apply), so they're omitted above rather than scored low.
AC1 Text alternatives — No web markup found — accessibility is not applicable to this repository.
AC2 Forms & labels — No web markup found — accessibility is not applicable to this repository.
AC3 Page structure — No web markup found — accessibility is not applicable to this repository.
AC4 Keyboard semantics — No web markup found — accessibility is not applicable to this repository.
AC5 ARIA correctness — No web markup found — accessibility is not applicable to this repository.
AC6 Visual & motion safety — No web markup found — accessibility is not applicable to this repository.
AX1 Captive dependencies — Not applicable: this repository's JavaScript imports no dependency-injection container and defines no container of its own — nothing that both registers and resolves bindings, and nothing that names two lifetimes — so nothing holds one lifetime's instance while handing out another's; this repository's Python imports no dependency-injection container and defines no container of its own — nothing that both registers and resolves bindings, and nothing that names two lifetimes — so nothing holds one lifetime's instance while handing out another's; Rust with no dependency-injection crate has no container to hand one lifetime's instance to another — every value is owned by the code that builds it, and the borrow checker rejects a longer-lived value keeping a borrow of a shorter-lived one.
AX2 Stateful singletons — No container singleton was found, so there is no shared instance for concurrent requests to race on. No function in this Python code is served by a threaded web framework (a Flask, Bottle or FastAPI route, a Django or Pyramid view), so no module is shared between request threads. Rust's compiler refuses unsynchronised shared mutation — a value shared across threads must be Sync — so the race this check looks for cannot be written. TypeScript/JavaScript runs each process's requests on one event loop, so no two requests write a shared object at the same instant (interleaving across an await is a different defect).
AX5 Architecture & structure — not assessed — architecture style/structure is computed from a project graph (projects, types, module namespaces) that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
AX6 Interface segregation — not assessed — interface segregation is computed over a type surface that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
AX7 Slice cohesion — not applicable — not a vertical-slice architecture
AX8 Test isolation — no test/production split to check
AXB2 Runtime readiness — Advisory — this card reports evidence and never carries a score, so there is nothing missing here.
AXR1 Runtime accessibility — the service did not build — `cargo build --locked --bin 'text-embeddings-router'` exited 1: info: syncing channel updates for 1.92.0-x86_64-unknown-linux-gnu
error: could not download file from 'https://static.rust-lang.org/dist/channel-rust-1.92.0.toml.sha256' to '/usr/local/rustup/tmp/2qkq…; runtime evidence skipped (launch chosen because 'axum' dependency in router/Cargo.toml, binary 'text-embeddings-router') This is a statement about this run, not a statement about your application: nothing here says the surface is inaccessible, only that it was never rendered.
C1 Data Protection — Not assessed: these personal data controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks personal data controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
C2 Access Controls — Not assessed: these authorization controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks authorization controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
D10 Test Quality — ~3267 lines of test source are present (.rs) but no test cases reached the test census for this repository — no test root we could resolve declared them, or the files we read declare no cases in a test framework we recognise — so skipped/assertion-free tests couldn't be counted. Not scored — this is a gap in the analyzer, not a finding about this repository.
D11 Test Reliability — Test reliability not included — the .rs suite was found but not re-run
D18 Solution Shape — D18 scores the shape of a .NET solution; this repository has no .NET solution or project files, so the dimension does not apply.
D21 Naming Consistency — Naming consistency does not read html yet; not measured.
D22 Internal API Consistency — API consistency is measured on C# only; the exposed Rust surface was not measured.
D23 Boundary Type-Coupling — No bounded-context organisation was detected either — neither a context-shaped layout, nor 2+ sibling source directories each declaring an aggregate root, nor 2+ sibling projects each declaring an entity of their own or taking another project's integration events. Declaring this codebase's bounded contexts (≥2) would let cross-boundary type coupling be assessed. Declare them in `.codehealth/config.yaml` at the repository root (create it if absent), mapping each context name to the module-path or namespace prefixes that belong to it — e.g. `architecture:` → `contexts:` → `Billing: ["src/billing", "Acme.Billing"]`, `Catalog: ["src/catalog", "Acme.Catalog"]`.
D24 Comment Value — D24 reads C#, Visual Basic, TypeScript and JavaScript comments only; comments in this repository's .py, .rs source were not judged.
D25 ADR Conformance — no ADRs to check
D27 Navigability — D27 counts a call into the standard library or a dependency as unresolved for this repository's language (.rs), so too few call sites counted as resolved to assess navigability. Not scored: this is a gap in the analyzer, not a verdict about this repository.
D32 Data Compliance (PII/GDPR) — No personal data was found crossing a boundary the PII/GDPR ruleset checks — nothing written to a log or console sink, placed in a URL or query string, or persisted to browser storage. That is a clean result for the LEAK surface only: this ruleset detects personal data escaping, it does not inventory the personal data a repository holds, so it is not evidence that this repository has no personal-data surface. The personal-data map (Appendix C) and the C1-C5 compliance cards are what speak to that.
D34 Knowledge Freshness — knowledge concentrated in recent work — freshness signal contradicted by repo activity
D37 Vulnerability-disclosure Policy — No vulnerability-disclosure policy file found (SECURITY.md/.markdown/.rst/.txt at root or under .github/.forgejo/.gitea/docs, .well-known/security.txt). A coordinated-disclosure policy may live off-repo, so this is not evidenced rather than failed.
D40 Network Egress Confinement — No Kubernetes/orchestration workloads found in the repository manifests; network egress policy is a cluster-native control that may live at the platform/firewall layer, so there is nothing to assess here.
D41 Kernel & Syscall Confinement — No Kubernetes/orchestration workloads found in the repository manifests; seccomp/AppArmor/SELinux confinement is a workload-level control, so there is nothing to assess here.
D42 Runtime Threat Enforcement — No Kubernetes/orchestration workloads found in the repository manifests; runtime threat-detection and admission-control policy are cluster-level controls, so there is nothing to assess here.
D7 Architectural Integrity — no checkable ADRs, and no project-reference graph for the cycle pass to read — so this dimension makes no claim about dependency cycles in either direction (where this repository's language has an import-cycle lens, cycles are reported there). Architectural integrity not assessed
D8 Code Coverage — Coverage not measured — Rust suite
DM1 Domain Modelling — applicable but not scored (2 of 3 signals for this style — below the bar we score at): 7 aggregate root(s) (types guarding their own state behind command methods — this language has no AggregateRoot base to inherit); 5 value object(s); its 7 aggregate(s) are ORM models only — persistence, not behaviour: a CRUD model, measured, not claimed as DDD
ED2 Event/command shape — not scored — deciding whether a command has more than one competing handler requires resolving the call graph, and a call made through an inferred or generic receiver has no resolvable owner in the source. Reported as guidance rather than measured
ED5 Idempotency — no mutating command handlers or message consumers detected — idempotency check not applicable
ES1 Event Sourcing — not scored — this repository shows none of the 3 signals this lens looks for
GD1 Unfinished & placeholder code — no source files were read — this check reads C#, F#, JavaScript/TypeScript, Dart, Ruby and Scala, and none was read for this repository's product. That is a limit of the analyzer, not a finding about your code.
IC1 Incompleteness & stubs — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
M4 Documentation accuracy — Not applicable: the 17 current-state document(s) make no claim of a kind M4 checks. 32 claim(s) of kinds M4 does not check were not checked (path 1, command argument 10, environment variable 20, cd 1); M4 checks only links and versions and artefact counts.
P12 CI test-gate honesty — Reported, not scored — and nothing was matched here. The coverage check applies to any stack, and the automatic-re-run check to any GitHub-Actions workflow, but the checks for excluded tests, skipped tests and sleep-based synchronisation currently recognise only some ecosystems' test-runner idioms, so on a repository built with another stack the zeros below mean 'not checked', not 'clean'.
P7 Outbound HTTP resilience — no outbound HTTP usage detected — no HTTP client call or construction in the Python, Rust source (service entry point: router/src/http/server.rs:1783 (Router::new()))
P8 Schema migrations — no ORM, schema-migration tool or schema auto-create was found in this repository's dependency manifests or source, so there is no database schema for this check to judge
P9 Domain vs controller coverage — coverage not measured: this check compares the coverage of the analyzer's own test run across the domain and web layers, and that run measured none here — Test coverage (D8) says why
PF1 Benchmark discipline — Not applicable: no benchmark suite was found. This check searched for `#[bench]`, criterion's `bench_function`/`bench_with_input`, `#[divan::bench]` or `#[library_benchmark]` in any `.rs` file, or criterion, divan, iai or a `[[bench]]` target in a Cargo.toml, and for a `*benchmark*` script that this repository's CI runs. Benchmarks are credited as a bonus, so their absence is neither scored nor deducted.
PF2 Allocation hygiene — Not scored: this check counted allocation-aware idioms (Span, pools, stackalloc) per line, which describes what kind of code a repository is rather than whether it has an allocation problem, so it was withdrawn. Nothing here is a finding about your code.
S1 Web-Security Posture — Not assessed: these web-security controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks web-security controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
X12 Unreachable branch — X12 reached only 11% of this repository's production lines — rust (23,853 lines) was left unread. Not scored: a score over a minority of the product would be published as the whole of it.
X13 Undrained process stream — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X14 Bypassable address classification — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X15 Unvalidated length from an untrusted reader — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X16 Unfloored truncation loop — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
X17 Uncapped recursion over a caller-supplied document — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X18 Disposal-pattern correctness — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
X19 Unrestored process-global state — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X2 Cancellation propagation — no C# was loaded; this repository's Python and Rust were read, and no request in it can be cancelled by a caller — a Python request takes no token: asyncio cancels the task and trio or anyio the scope that made it, from outside the call, and requests, httpx, aiohttp and gRPC take a timeout, not a cancellation; a Rust request takes no token: a future is cancelled by dropping it, from outside the call (select!, a timeout, an aborted task), and reqwest, hyper and tonic take none
X20 Mistyped argument guard — This check asks whether an argument guard reports an EMPTY value as a NULL one, which needs a language that throws a null-specific argument exception — .NET's ArgumentNullException, the JVM's NullPointerException, Dart's ArgumentError.notNull. This repository contains none of those languages: the ones it is written in have a single exception for a bad argument, so there is no pair of exceptions to confuse and nothing here for it to assess. Not a gap in the analyzer and not a finding about your code.
X22 Contradicted release guard — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X23 Unguarded diagnostic materialisation — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X24 Document value interpolated into markup unescaped — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X25 Inert configuration knob — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X26 Unsynchronised callback handoff — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X27 Collection changed while being enumerated — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X28 Index access outside its own emptiness guard — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X29 Per-element action decided by a fixed element — not applicable — the code was read and holds no site of this kind, so there is nothing for this check to judge
X3 Exception handling — X3 reached only 11% of this repository's production lines — rust (23,853 lines) was left unread. Not scored: a score over a minority of the product would be published as the whole of it.
X32 Type resolved by simple name across every loaded assembly — This check is about how a .NET program searches the assemblies loaded into its process for a type, and this repository contains no .NET source, so there is nothing here for it to assess. Not a gap in the analyzer and not a finding about your code.
X5 Nullable reference types — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
X6 Hand-rolled structured-format parsing — Advisory — this card reports evidence and never carries a score, so there is nothing missing here.
X7 Silent fallback defaults — Advisory — this card reports evidence and never carries a score, so there is nothing missing here.
X9 Subsumed condition operand — Advisory — this card reports evidence and never carries a score, so there is nothing missing here.
Appendix A — Findings (grouped)
The findings behind the scores, grouped by severity, then by dimension and kind. The high-severity issues are enumerated in full below; items per group are capped at 25 with any overflow stated explicitly per group, never silently truncated. The complete machine-readable list of every finding (all severities) is the companion findings.md in this report's bundle.
MethodTooLong: CandleBackend.new backends/candle/src/lib.rs:163— MethodTooLong — new runs 331 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 231 over it, 3.31× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: OrtBackend.embed backends/ort/src/lib.rs:402— MethodTooLong — embed runs 199 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 99 over it, 1.99× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: Qwen3Model.forward backends/candle/src/models/qwen3.rs:518— MethodTooLong — forward runs 166 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 66 over it, 1.66× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: TextEmbeddingsService.rerank_stream router/src/grpc/server.rs:1068— MethodTooLong — rerank_stream runs 164 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 64 over it, 1.64× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: JinaBertModel.forward backends/candle/src/models/jina.rs:471— MethodTooLong — forward runs 163 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 63 over it, 1.63× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: JinaCodeBertModel.forward backends/candle/src/models/jina_code.rs:398— MethodTooLong — forward runs 163 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 63 over it, 1.63× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: NomicBertModel.forward backends/candle/src/models/nomic.rs:773— MethodTooLong — forward runs 153 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 53 over it, 1.53× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: Gemma3Model.forward backends/candle/src/models/gemma3.rs:721— MethodTooLong — forward runs 148 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 48 over it, 1.48× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: GTEModel.forward backends/candle/src/models/gte.rs:489— MethodTooLong — forward runs 144 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 44 over it, 1.44× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: BertModel.forward backends/candle/src/models/bert.rs:706— MethodTooLong — forward runs 143 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 43 over it, 1.43× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: ModernBertModel.forward backends/candle/src/models/modernbert.rs:633— MethodTooLong — forward runs 136 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 36 over it, 1.36× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: DistilBertModel.forward backends/candle/src/models/distilbert.rs:508— MethodTooLong — forward runs 134 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 34 over it, 1.34× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: TextEmbeddingsService.rerank router/src/grpc/server.rs:877— MethodTooLong — rerank runs 133 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 33 over it, 1.33× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: DebertaV2DisentangledSelfAttention.disentangled_attention_bias backends/candle/src/models/debertav2.rs:449— MethodTooLong — disentangled_attention_bias runs 127 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 27 over it, 1.27× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: Gemma3Attention.forward backends/candle/src/models/gemma3.rs:299— MethodTooLong — forward runs 127 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 27 over it, 1.27× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: DebertaV2Model.forward backends/candle/src/models/debertav2.rs:1121— MethodTooLong — forward runs 119 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 19 over it, 1.19× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: OrtBackend.prepare_inputs backends/ort/src/lib.rs:174— MethodTooLong — prepare_inputs runs 118 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 18 over it, 1.18× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: MPNetModel.forward backends/candle/src/models/mpnet.rs:506— MethodTooLong — forward runs 105 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 5 over it, 1.05× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
MethodTooLong: FlashNomicBertModel.forward backends/candle/src/models/flash_nomic.rs:272— MethodTooLong — forward runs 102 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 2 over it, 1.02× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
FileTooLong: http/server.rs router/src/http/server.rs— FileTooLong — 1466 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 966 over it, 2.93× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: models/debertav2.rs backends/candle/src/models/debertav2.rs— FileTooLong — 1059 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 559 over it, 2.12× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: grpc/server.rs router/src/grpc/server.rs— FileTooLong — 1059 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted), about 86% of them inside a single declaration: TextEmbeddingsService (7 blocks, 46-1364). The bar is 500 significant lines; this is 559 over it, 2.12× the bar. Moving the declarations that sit BESIDE it into sibling files will not shorten this file. Extract from INSIDE that declaration instead: lift each cohesive group of its body — the parts that share the same inputs and are named together — into its own unit in a sibling file, and have the original call them.
FileTooLong: models/gemma3.rs backends/candle/src/models/gemma3.rs— FileTooLong — 689 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 189 over it, 1.38× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: models/nomic.rs backends/candle/src/models/nomic.rs— FileTooLong — 679 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 179 over it, 1.36× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: src/lib.rs backends/src/lib.rs— FileTooLong — 670 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 170 over it, 1.34× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: models/bert.rs backends/candle/src/models/bert.rs— FileTooLong — 651 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 151 over it, 1.30× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: models/modernbert.rs backends/candle/src/models/modernbert.rs— FileTooLong — 614 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 114 over it, 1.23× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: models/qwen3.rs backends/candle/src/models/qwen3.rs— FileTooLong — 543 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 43 over it, 1.09× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: src/lib.rs router/src/lib.rs— FileTooLong — 536 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 36 over it, 1.07× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: src/lib.rs backends/candle/src/lib.rs— FileTooLong — 506 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted), about 76% of them inside a single declaration: CandleBackend (3 blocks, 156-726). The bar is 500 significant lines; this is 6 over it, 1.01× the bar. Moving the declarations that sit BESIDE it into sibling files will not shorten this file. Extract from INSIDE that declaration instead: lift each cohesive group of its body — the parts that share the same inputs and are named together — into its own unit in a sibling file, and have the original call them.
FileTooLong: models/jina.rs backends/candle/src/models/jina.rs— FileTooLong — 505 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted). The bar is 500 significant lines; this is 5 over it, 1.01× the bar. To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: src/lib.rs backends/ort/src/lib.rs— FileTooLong — 502 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted), about 90% of them inside a single declaration: OrtBackend (4 blocks, 61-722). The bar is 500 significant lines; this is 2 over it, 1.00× the bar. Moving the declarations that sit BESIDE it into sibling files will not shorten this file. Extract from INSIDE that declaration instead: lift each cohesive group of its body — the parts that share the same inputs and are named together — into its own unit in a sibling file, and have the original call them.
Duplicated block (10 lines × 2) backends/candle/src/lib.rs:369— backends/candle/src/lib.rs:369-378 | backends/candle/src/lib.rs:539-548 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (10 lines × 2) backends/candle/src/models/bert.rs:254— backends/candle/src/models/bert.rs:254-263 | backends/candle/src/models/jina.rs:227-236 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 37 separate duplicated blocks between them, totalling at least 547 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:254` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 2) backends/candle/src/models/bert.rs:409— backends/candle/src/models/bert.rs:409-418 | backends/candle/src/models/jina.rs:383-392 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 37 separate duplicated blocks between them, totalling at least 547 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:409` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 2) backends/candle/src/models/flash_gte.rs:140— backends/candle/src/models/flash_gte.rs:140-149 | backends/candle/src/models/gte.rs:270-279 — before extracting anything, compare `backends/candle/src/models/flash_gte.rs` and `backends/candle/src/models/gte.rs` as WHOLE FILES: this scan already matched 7 separate duplicated blocks between them, totalling at least 107 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_gte.rs:140` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 2) backends/candle/src/models/jina.rs:412— backends/candle/src/models/jina.rs:412-421 | backends/candle/src/models/jina_code.rs:343-352 — before extracting anything, compare `backends/candle/src/models/jina.rs` and `backends/candle/src/models/jina_code.rs` as WHOLE FILES: this scan already matched 29 separate duplicated blocks between them, totalling at least 536 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (10 lines × 2) backends/candle/src/models/modernbert.rs:722— backends/candle/src/models/modernbert.rs:722-731 | backends/candle/src/models/modernbert.rs:733-742 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
Duplicated block (10 lines × 2) backends/python/src/lib.rs:88— backends/python/src/lib.rs:88-97 | backends/python/src/lib.rs:116-125 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/python/src/lib.rs:88` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (10 lines × 2) router/src/http/server.rs:600— router/src/http/server.rs:600-609 | router/src/http/server.rs:1171-1180 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:600` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 2) backends/candle/src/models/flash_nomic.rs:244— backends/candle/src/models/flash_nomic.rs:244-253 | backends/candle/src/models/nomic.rs:742-751 — before extracting anything, compare `backends/candle/src/models/flash_nomic.rs` and `backends/candle/src/models/nomic.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 106 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_nomic.rs:244` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 2) backends/candle/src/models/debertav2.rs:665— backends/candle/src/models/debertav2.rs:665-674 | backends/candle/src/models/debertav2.rs:726-735 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/debertav2.rs:665` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (10 lines × 2) backends/python/server/text_embeddings_server/models/default_model.py:68— backends/python/server/text_embeddings_server/models/default_model.py:68-77 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:461-470 — the copies sit in sibling files of one directory, so a shared home is within easy reach: extract the block into a single shared function the call sites can all reach — a file they already depend on, or a new one alongside them — and call it from both call sites, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `backends/python/server/text_embeddings_server/models/flash_qwen3.py:461` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
TodoComment backends/src/lib.rs:527— // TODO(alvarobartt): eventually detach the Sentence Transformers module handling — source code is not a task system: move the work to your tracker and leave a reference instead (e.g. `// REF: #123`), so the task is planned where tasks live and the ticket links back to the code.
TodoComment backends/src/lib.rs:729— // TODO: Most likely all the download functions could benefit from the information defined in — source code is not a task system: move the work to your tracker and leave a reference instead (e.g. `// REF: #123`), so the task is planned where tasks live and the ticket links back to the code.
TodoComment backends/candle/src/lib.rs:323— // TODO(alvarobartt): Enable Flash Attention with BF16 once supported on Metal — source code is not a task system: move the work to your tracker and leave a reference instead (e.g. `// REF: #123`), so the task is planned where tasks live and the ticket links back to the code.
TodoComment backends/candle/src/lib.rs:368— // TODO(alvarobartt): Enable Flash Attention with BF16 once supported on Metal — source code is not a task system: move the work to your tracker and leave a reference instead (e.g. `// REF: #123`), so the task is planned where tasks live and the ticket links back to the code.
TodoComment backends/candle/src/lib.rs:456— // TODO(alvarobartt): Enable Flash Attention with BF16 once supported on CUDA — source code is not a task system: move the work to your tracker and leave a reference instead (e.g. `// REF: #123`), so the task is planned where tasks live and the ticket links back to the code.
TodoComment backends/candle/src/lib.rs:538— // TODO(alvarobartt): Enable Flash Attention with BF16 once supported on CUDA — source code is not a task system: move the work to your tracker and leave a reference instead (e.g. `// REF: #123`), so the task is planned where tasks live and the ticket links back to the code.
TodoComment backends/candle/src/models/debertav2.rs:794— // TODO: In order to fully test ConvLayer a model needs to be found has a configuration where — source code is not a task system: move the work to your tracker and leave a reference instead (e.g. `// REF: #123`), so the task is planned where tasks live and the ticket links back to the code.
TodoComment backends/candle/src/models/qwen3.rs:31— // TODO(alvarobartt): Migrate to `is_causal` instead — source code is not a task system: move the work to your tracker and leave a reference instead (e.g. `// REF: #123`), so the task is planned where tasks live and the ticket links back to the code.
TodoComment backends/candle/src/models/qwen3.rs:392— // TODO(alvarobartt): Eventually extend Qwen3 for Voyage instead of adding `projection` here — source code is not a task system: move the work to your tracker and leave a reference instead (e.g. `// REF: #123`), so the task is planned where tasks live and the ticket links back to the code.
FunctionTooLong: text_embeddings_router::run router/src/lib.rs:47— FunctionTooLong — text_embeddings_router::run runs 257 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 157 over it, 2.57× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
FunctionTooLong: text_embeddings_router::http::server::run router/src/http/server.rs:1626— FunctionTooLong — text_embeddings_router::http::server::run runs 236 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 136 over it, 2.36× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
FunctionTooLong: text_embeddings_router::http::server::openai_embed router/src/http/server.rs:1127— FunctionTooLong — text_embeddings_router::http::server::openai_embed runs 139 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 39 over it, 1.39× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
FunctionTooLong: text_embeddings_router::http::server::predict router/src/http/server.rs:104— FunctionTooLong — text_embeddings_router::http::server::predict runs 123 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 23 over it, 1.23× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
FunctionTooLong: text_embeddings_router::http::server::rerank router/src/http/server.rs:305— FunctionTooLong — text_embeddings_router::http::server::rerank runs 119 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 19 over it, 1.19× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
FunctionTooLong: text_embeddings_router::http::server::embed_sparse router/src/http/server.rs:764— FunctionTooLong — text_embeddings_router::http::server::embed_sparse runs 115 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 15 over it, 1.15× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
FunctionTooLong: text_embeddings_backend::init_backend backends/src/lib.rs:447— FunctionTooLong — text_embeddings_backend::init_backend runs 112 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 12 over it, 1.12× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
FunctionTooLong: text_embeddings_router::http::server::embed router/src/http/server.rs:584— FunctionTooLong — text_embeddings_router::http::server::embed runs 112 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 12 over it, 1.12× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
FunctionTooLong: text_embeddings_router::http::server::embed_all router/src/http/server.rs:951— FunctionTooLong — text_embeddings_router::http::server::embed_all runs 108 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted) in one body. The bar is 100 significant lines; this is 8 over it, 1.08× the bar. This is length, not branching: a long straight-line body scores low on complexity and is still read whole to change any part of it, so the complexity numbers beside this row neither confirm nor excuse it. To reduce it, extract each cohesive step of the body — the runs of statements that work on the same values and would earn the same name — into its own named unit, and have this one call them in order.
D30 · Dependency Vulnerabilities· Medium advisory (unmaintained) · ×8
Duplicated block (17 lines × 2) backends/candle/src/layers/rms_norm.rs:55— backends/candle/src/layers/rms_norm.rs:55-71 | backends/candle/src/models/gemma3.rs:90-106 — before extracting anything, compare `backends/candle/src/layers/rms_norm.rs` and `backends/candle/src/models/gemma3.rs` as WHOLE FILES: this scan already matched 4 separate duplicated blocks between them, totalling at least 63 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. The two sit in different directories, so one cannot simply be deleted in favour of the other while both are reached separately: hoist the shared part into a location both already depend on and have each file call it, and retire whichever file turns out to have no caller of its own left. Extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/layers/rms_norm.rs:55` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (17 lines × 2) backends/candle/src/models/flash_modernbert.rs:143— backends/candle/src/models/flash_modernbert.rs:143-159 | backends/candle/src/models/modernbert.rs:324-340 — before extracting anything, compare `backends/candle/src/models/flash_modernbert.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 113 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_modernbert.rs:143` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (17 lines × 2) backends/candle/src/models/flash_nomic.rs:114— backends/candle/src/models/flash_nomic.rs:114-130 | backends/candle/src/models/nomic.rs:614-630 — before extracting anything, compare `backends/candle/src/models/flash_nomic.rs` and `backends/candle/src/models/nomic.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 106 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (17 lines × 2) core/src/infer.rs:370— core/src/infer.rs:370-386 | core/src/infer.rs:436-452 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `core/src/infer.rs:370` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (17 lines × 2) router/src/grpc/server.rs:96— router/src/grpc/server.rs:96-112 | router/src/grpc/server.rs:217-233 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:96` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `router/src/grpc/server.rs:94` calls `unwrap_or` and `router/src/grpc/server.rs:216` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (17 lines × 2) router/src/grpc/server.rs:147— router/src/grpc/server.rs:147-163 | router/src/grpc/server.rs:213-229 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:147` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it. Note first that the copies are not typed on the same thing: the declarations holding them bind `request` to `EmbedSparseRequest` in one and `EmbedAllRequest` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (17 lines × 2) router/src/grpc/server.rs:898— router/src/grpc/server.rs:898-914 | router/src/grpc/server.rs:1076-1092 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
Duplicated block (17 lines × 2) router/src/http/server.rs:686— router/src/http/server.rs:686-702 | router/src/http/server.rs:1261-1277 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:686` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (9 lines × 2) backends/candle/src/lib.rs:324— backends/candle/src/lib.rs:324-332 | backends/candle/src/lib.rs:457-465 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (9 lines × 2) backends/candle/src/models/bert.rs:509— backends/candle/src/models/bert.rs:509-517 | backends/candle/src/models/bert.rs:541-549 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:509` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9 lines × 2) backends/candle/src/models/flash_jina_code.rs:233— backends/candle/src/models/flash_jina_code.rs:233-241 | backends/candle/src/models/jina_code.rs:287-295 — before extracting anything, compare `backends/candle/src/models/flash_jina_code.rs` and `backends/candle/src/models/jina_code.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 121 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_jina_code.rs:233` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9 lines × 2) backends/candle/src/models/flash_mistral.rs:47— backends/candle/src/models/flash_mistral.rs:47-55 | backends/candle/src/models/flash_qwen2.rs:59-67 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_mistral.rs:47` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/flash_mistral.rs:57` calls `unwrap_or` and `backends/candle/src/models/flash_qwen2.rs:68` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (9 lines × 2) backends/src/lib.rs:535— backends/src/lib.rs:535-543 | backends/src/lib.rs:892-900 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/src/lib.rs:535` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9 lines × 2) core/src/infer.rs:548— core/src/infer.rs:548-556 | core/src/infer.rs:579-587 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `core/src/infer.rs:548` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9 lines × 2) backends/candle/src/layers/rms_norm.rs:12— backends/candle/src/layers/rms_norm.rs:12-20 | backends/candle/src/models/gemma3.rs:45-53 — before extracting anything, compare `backends/candle/src/layers/rms_norm.rs` and `backends/candle/src/models/gemma3.rs` as WHOLE FILES: this scan already matched 4 separate duplicated blocks between them, totalling at least 63 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. The two sit in different directories, so one cannot simply be deleted in favour of the other while both are reached separately: hoist the shared part into a location both already depend on and have each file call it, and retire whichever file turns out to have no caller of its own left. Extracting one helper per block leaves the fork in place.
Duplicated block (9 lines × 2) backends/candle/src/models/distilbert.rs:97— backends/candle/src/models/distilbert.rs:97-105 | backends/candle/src/models/gte.rs:63-71 — before extracting anything, compare `backends/candle/src/models/distilbert.rs` and `backends/candle/src/models/gte.rs` as WHOLE FILES: this scan already matched 24 separate duplicated blocks between them, totalling at least 362 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/distilbert.rs:97` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (15 lines × 2) backends/candle/src/layers/layer_norm.rs:51— backends/candle/src/layers/layer_norm.rs:51-65 | backends/candle/src/layers/layer_norm.rs:133-147 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/layers/layer_norm.rs:51` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (15 lines × 2) backends/candle/src/models/distilbert.rs:259— backends/candle/src/models/distilbert.rs:259-273 | backends/candle/src/models/flash_distilbert.rs:110-124 — before extracting anything, compare `backends/candle/src/models/distilbert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 6 separate duplicated blocks between them, totalling at least 97 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (15 lines × 2) backends/candle/src/models/flash_gte.rs:243— backends/candle/src/models/flash_gte.rs:243-257 | backends/candle/src/models/gte.rs:458-472 — before extracting anything, compare `backends/candle/src/models/flash_gte.rs` and `backends/candle/src/models/gte.rs` as WHOLE FILES: this scan already matched 7 separate duplicated blocks between them, totalling at least 107 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (15 lines × 2) router/src/grpc/server.rs:756— router/src/grpc/server.rs:756-770 | router/src/grpc/server.rs:832-846 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
Duplicated block (15 lines × 2) router/src/grpc/server.rs:931— router/src/grpc/server.rs:931-945 | router/src/http/server.rs:356-370 — `router/src/grpc/server.rs` and `router/src/http/server.rs` are one unit implemented once per sibling directory, so they are most likely parallel implementations of one contract rather than a copy of each other — this scan matched 7 separate duplicated blocks between them, totalling at least 91 lines. If both are selected at run time, neither can be retired in favour of the other, and the lines that DIFFER between them are the reason both exist. The move that pays here is to hoist the identical part into a shared location the whole family can reach and give what differs a parameter or a seam, so a change lands once instead of once per sibling; extracting one helper per block leaves every sibling to drift on its own. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:931` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `router/src/http/server.rs:372` calls `unwrap_or` and `router/src/grpc/server.rs:947` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (15 lines × 2) backends/candle/src/models/flash_gte.rs:25— backends/candle/src/models/flash_gte.rs:25-39 | backends/candle/src/models/gte.rs:51-65 — before extracting anything, compare `backends/candle/src/models/flash_gte.rs` and `backends/candle/src/models/gte.rs` as WHOLE FILES: this scan already matched 7 separate duplicated blocks between them, totalling at least 107 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (15 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:297— backends/python/server/text_embeddings_server/models/flash_mistral.py:297-311 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:286-300 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. The `return` at the foot of the matched lines is the enclosing body's own terminal exit, not an early one: it moves with them unchanged, and each site calls the extracted unit from the position that `return` occupied — no decision has to be handed back and re-acted on.
Duplicated block (7 lines × 2) backends/candle/src/models/distilbert.rs:532— backends/candle/src/models/distilbert.rs:532-539 | backends/candle/src/models/modernbert.rs:654-660 — before extracting anything, compare `backends/candle/src/models/distilbert.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 192 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/distilbert.rs:532` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/distilbert.rs:540` calls `push` and `backends/candle/src/models/modernbert.rs:661` does not — after which the two agree again for 4 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (7 lines × 2) backends/candle/src/models/gemma3.rs:461— backends/candle/src/models/gemma3.rs:461-467 | backends/candle/src/models/qwen3.rs:264-271 — before extracting anything, compare `backends/candle/src/models/gemma3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 15 separate duplicated blocks between them, totalling at least 209 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/gemma3.rs:461` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (7 lines × 2) backends/candle/src/models/gemma3.rs:761— backends/candle/src/models/gemma3.rs:761-767 | backends/candle/src/models/qwen3.rs:561-567 — before extracting anything, compare `backends/candle/src/models/gemma3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 15 separate duplicated blocks between them, totalling at least 209 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/gemma3.rs:761` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/gemma3.rs:770` calls `broadcast_as`, `contiguous` and `backends/candle/src/models/qwen3.rs:568` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (7 lines × 2) backends/candle/src/models/flash_mistral.rs:160— backends/candle/src/models/flash_mistral.rs:160-166 | backends/candle/src/models/flash_qwen2.rs:171-177 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_mistral.rs:160` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (7 lines × 2) backends/candle/src/models/bert.rs:342— backends/candle/src/models/bert.rs:342-348 | backends/candle/src/models/flash_bert.rs:195-201 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/flash_bert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (7 lines × 2) backends/candle/src/models/distilbert.rs:302— backends/candle/src/models/distilbert.rs:302-308 | backends/candle/src/models/flash_distilbert.rs:154-160 — before extracting anything, compare `backends/candle/src/models/distilbert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 6 separate duplicated blocks between them, totalling at least 97 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (7 lines × 2) backends/candle/src/models/flash_nomic.rs:165— backends/candle/src/models/flash_nomic.rs:165-171 | backends/candle/src/models/nomic.rs:664-670 — before extracting anything, compare `backends/candle/src/models/flash_nomic.rs` and `backends/candle/src/models/nomic.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 106 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (20 lines × 2) backends/candle/src/layers/cublaslt.rs:68— backends/candle/src/layers/cublaslt.rs:68-87 | backends/candle/src/layers/cublaslt.rs:109-128 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/layers/cublaslt.rs:68` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (20 lines × 2) backends/candle/src/models/flash_bert.rs:92— backends/candle/src/models/flash_bert.rs:92-111 | backends/candle/src/models/flash_distilbert.rs:73-92 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/flash_bert.rs:112` calls `forward`, `Some` and `backends/candle/src/models/flash_distilbert.rs:94` does not — after which the two agree again for 4 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (20 lines × 2) backends/candle/src/models/flash_bert.rs:231— backends/candle/src/models/flash_bert.rs:231-250 | backends/candle/src/models/flash_bert.rs:297-316 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:231` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (20 lines × 2) backends/candle/src/models/flash_bert.rs:464— backends/candle/src/models/flash_bert.rs:464-483 | backends/candle/src/models/flash_distilbert.rs:333-352 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:464` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (20 lines × 2) backends/candle/src/models/flash_modernbert.rs:30— backends/candle/src/models/flash_modernbert.rs:30-49 | backends/candle/src/models/modernbert.rs:175-194 — before extracting anything, compare `backends/candle/src/models/flash_modernbert.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 113 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (20 lines × 2) backends/candle/src/models/gemma3.rs:348— backends/candle/src/models/gemma3.rs:348-367 | backends/candle/src/models/qwen3.rs:181-200 — before extracting anything, compare `backends/candle/src/models/gemma3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 15 separate duplicated blocks between them, totalling at least 209 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/gemma3.rs:348` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11 lines × 2) backends/candle/src/models/bert.rs:642— backends/candle/src/models/bert.rs:642-652 | backends/candle/src/models/flash_bert.rs:311-321 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/flash_bert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:642` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11 lines × 2) backends/candle/src/models/distilbert.rs:51— backends/candle/src/models/distilbert.rs:51-61 | backends/candle/src/models/mpnet.rs:58-68 — before extracting anything, compare `backends/candle/src/models/distilbert.rs` and `backends/candle/src/models/mpnet.rs` as WHOLE FILES: this scan already matched 14 separate duplicated blocks between them, totalling at least 238 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (11 lines × 2) backends/candle/src/models/flash_mistral.rs:34— backends/candle/src/models/flash_mistral.rs:34-44 | backends/candle/src/models/gemma3.rs:173-183 — the copies sit in sibling files of one directory, so a shared home is within easy reach: extract the block into a single shared function the call sites can all reach — a file they already depend on, or a new one alongside them — and call it from both call sites, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/gemma3.rs:173` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11 lines × 2) backends/candle/src/models/flash_qwen3.rs:431— backends/candle/src/models/flash_qwen3.rs:431-441 | backends/candle/src/models/qwen3.rs:639-649 — before extracting anything, compare `backends/candle/src/models/flash_qwen3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 225 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_qwen3.rs:431` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11 lines × 2) router/src/grpc/server.rs:156— router/src/grpc/server.rs:156-166 | router/src/grpc/server.rs:291-301 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:156` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11 lines × 2) backends/candle/src/models/gte.rs:360— backends/candle/src/models/gte.rs:360-371 | backends/candle/src/models/nomic.rs:678-688 — before extracting anything, compare `backends/candle/src/models/gte.rs` and `backends/candle/src/models/nomic.rs` as WHOLE FILES: this scan already matched 24 separate duplicated blocks between them, totalling at least 399 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/gte.rs:360` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (6 lines × 2) backends/candle/src/models/distilbert.rs:78— backends/candle/src/models/distilbert.rs:78-83 | backends/candle/src/models/flash_distilbert.rs:25-30 — before extracting anything, compare `backends/candle/src/models/distilbert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 6 separate duplicated blocks between them, totalling at least 97 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (6 lines × 2) backends/candle/src/models/flash_mistral.rs:27— backends/candle/src/models/flash_mistral.rs:27-32 | backends/candle/src/models/flash_qwen2.rs:31-36 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&MistralConfig` in one and `&Qwen2Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/flash_qwen2.rs:37` calls `pp`, `get` and `backends/candle/src/models/flash_mistral.rs:34` does not — after which the two agree again for 4 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (6 lines × 2) backends/candle/src/models/flash_modernbert.rs:77— backends/candle/src/models/flash_modernbert.rs:77-82 | backends/candle/src/models/modernbert.rs:217-222 — before extracting anything, compare `backends/candle/src/models/flash_modernbert.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 113 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/modernbert.rs:215` calls `device` and `backends/candle/src/models/flash_modernbert.rs:75` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (6 lines × 2) backends/src/lib.rs:372— backends/src/lib.rs:372-378 | backends/src/lib.rs:406-411 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/src/lib.rs:372` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (6 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:38— backends/python/server/text_embeddings_server/models/flash_mistral.py:38-43 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:25-30 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (6 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:255— backends/python/server/text_embeddings_server/models/flash_mistral.py:255-260 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:244-249 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (14 lines × 2) backends/candle/src/models/flash_jina.rs:235— backends/candle/src/models/flash_jina.rs:235-248 | backends/candle/src/models/flash_jina_code.rs:286-299 — before extracting anything, compare `backends/candle/src/models/flash_jina.rs` and `backends/candle/src/models/flash_jina_code.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 288 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_jina.rs:235` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (14 lines × 2) backends/candle/src/models/flash_mistral.rs:265— backends/candle/src/models/flash_mistral.rs:265-278 | backends/candle/src/models/flash_qwen2.rs:286-299 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_mistral.rs:265` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&MistralConfig` in one and `&Qwen2Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (14 lines × 2) backends/candle/src/models/gte.rs:97— backends/candle/src/models/gte.rs:97-110 | backends/candle/src/models/nomic.rs:519-532 — before extracting anything, compare `backends/candle/src/models/gte.rs` and `backends/candle/src/models/nomic.rs` as WHOLE FILES: this scan already matched 24 separate duplicated blocks between them, totalling at least 399 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/gte.rs:97` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (14 lines × 2) core/src/infer.rs:130— core/src/infer.rs:130-143 | core/src/infer.rs:181-194 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `core/src/infer.rs:130` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (14 lines × 2) backends/python/server/text_embeddings_server/models/flash_bert.py:381— backends/python/server/text_embeddings_server/models/flash_bert.py:381-394 | backends/python/server/text_embeddings_server/models/flash_mistral.py:438-451 — the copies sit in sibling files of one directory, so a shared home is within easy reach: extract the block into a single shared function the call sites can all reach — a file they already depend on, or a new one alongside them — and call it from both call sites, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `backends/python/server/text_embeddings_server/models/flash_bert.py:381` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
D30 · Dependency Vulnerabilities· Medium vulnerability · ×4
REDACTED
REDACTED
REDACTED
REDACTED
D4 · Code Duplication· Members sharing a duplicated core (5 members, 50+ identical tokens) · ×4
Members sharing a duplicated core (5 members, 50+ identical tokens) backends/candle/src/models/bert.rs:375— backends/candle/src/models/bert.rs:375-402 | backends/candle/src/models/bert.rs:428-451 | backends/candle/src/models/distilbert.rs:335-356 | backends/candle/src/models/gte.rs:300-320 | backends/candle/src/models/jina.rs:349-376 — These 5 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 5 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 5 times.
Members sharing a duplicated core (5 members, 50+ identical tokens) backends/candle/src/models/bert.rs:577— backends/candle/src/models/bert.rs:577-630 | backends/candle/src/models/bert.rs:636-704 | backends/candle/src/models/flash_bert.rs:230-290 | backends/candle/src/models/flash_bert.rs:296-367 | backends/candle/src/models/jina.rs:411-469 — These 5 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 5 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 5 times.
Members sharing a duplicated core (5 members, 50+ identical tokens) backends/candle/src/models/flash_gte.rs:167— backends/candle/src/models/flash_gte.rs:167-237 | backends/candle/src/models/flash_mistral.rs:240-301 | backends/candle/src/models/flash_qwen2.rs:251-322 | backends/candle/src/models/flash_qwen3.rs:303-392 | backends/candle/src/models/gte.rs:392-452 — These 5 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 5 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 5 times.
Members sharing a duplicated core (5 members, 50+ identical tokens) backends/candle/src/models/flash_mistral.rs:128— backends/candle/src/models/flash_mistral.rs:128-154 | backends/candle/src/models/flash_qwen2.rs:139-165 | backends/candle/src/models/flash_qwen3.rs:187-213 | backends/candle/src/models/gemma3.rs:481-505 | backends/candle/src/models/qwen3.rs:285-311 — These 5 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 5 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 5 times.
Duplicated block (33 lines × 2) backends/candle/src/models/flash_gte.rs:177— backends/candle/src/models/flash_gte.rs:177-209 | backends/candle/src/models/gte.rs:393-425 — before extracting anything, compare `backends/candle/src/models/flash_gte.rs` and `backends/candle/src/models/gte.rs` as WHOLE FILES: this scan already matched 7 separate duplicated blocks between them, totalling at least 107 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (33 lines × 2) backends/candle/src/models/flash_mistral.rs:74— backends/candle/src/models/flash_mistral.rs:74-106 | backends/candle/src/models/flash_qwen2.rs:85-117 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_mistral.rs:74` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (33 lines × 2) router/src/grpc/server.rs:1023— router/src/grpc/server.rs:1023-1055 | router/src/grpc/server.rs:1284-1316 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:1023` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (33 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:158— backends/python/server/text_embeddings_server/models/flash_mistral.py:158-190 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:127-159 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (16 lines × 2) backends/candle/src/models/flash_bert.rs:100— backends/candle/src/models/flash_bert.rs:100-115 | backends/candle/src/models/flash_jina.rs:102-117 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_jina.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 293 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:100` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (16 lines × 2) backends/candle/src/models/flash_gte.rs:111— backends/candle/src/models/flash_gte.rs:111-126 | backends/candle/src/models/gte.rs:241-256 — before extracting anything, compare `backends/candle/src/models/flash_gte.rs` and `backends/candle/src/models/gte.rs` as WHOLE FILES: this scan already matched 7 separate duplicated blocks between them, totalling at least 107 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (16 lines × 2) core/src/infer.rs:206— core/src/infer.rs:206-221 | core/src/infer.rs:306-321 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `core/src/infer.rs:206` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (16 lines × 2) router/src/grpc/server.rs:731— router/src/grpc/server.rs:731-746 | router/src/grpc/server.rs:781-796 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:731` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note first that the copies are not typed on the same thing: the declarations holding them bind `request` to `Request<PredictRequest>` in one and `Request<PredictPairRequest>` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (12 lines × 2) backends/candle/src/models/bert.rs:70— backends/candle/src/models/bert.rs:70-81 | backends/candle/src/models/mpnet.rs:44-55 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/mpnet.rs` as WHOLE FILES: this scan already matched 21 separate duplicated blocks between them, totalling at least 346 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:70` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&BertConfig` in one and `&MPNetConfig` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (12 lines × 2) router/src/grpc/server.rs:1001— router/src/grpc/server.rs:1001-1014 | router/src/http/server.rs:426-437 — `router/src/grpc/server.rs` and `router/src/http/server.rs` are one unit implemented once per sibling directory, so they are most likely parallel implementations of one contract rather than a copy of each other — this scan matched 7 separate duplicated blocks between them, totalling at least 91 lines. If both are selected at run time, neither can be retired in favour of the other, and the lines that DIFFER between them are the reason both exist. The move that pays here is to hoist the identical part into a shared location the whole family can reach and give what differs a parameter or a seam, so a change lands once instead of once per sibling; extracting one helper per block leaves every sibling to drift on its own. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:1001` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (12 lines × 2) router/src/http/server.rs:590— router/src/http/server.rs:590-601 | router/src/http/server.rs:957-968 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:590` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (12 lines × 2) backends/candle/src/models/flash_modernbert.rs:201— backends/candle/src/models/flash_modernbert.rs:201-212 | backends/candle/src/models/modernbert.rs:384-395 — before extracting anything, compare `backends/candle/src/models/flash_modernbert.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 113 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Hotspot: router/src/lib.rs router/src/lib.rs:47— router/src/lib.rs changed 4 times in last 90 days, and the most complex body those changes touched has cyclomatic complexity 28 in text_embeddings_router::run at line 47. Frequent change and high complexity in one file compound: schedule the next change to it to include carving out the part being edited, with the area under test before it moves. Counted over 2026-08-08..2026-11-06, the 90 days ending at the analysed commit. Reproduce with `git log --since='2026-08-08 13:10:20 +01:00' --until='2026-11-06 13:10:20 +01:00' --full-history --no-merges -- router/src/lib.rs`: merges are excluded because a merge re-states changes already counted at their own commits, and history is NOT path-simplified because a change that reached the file through a merged branch is still a change to it. That command counts raw commits and can read HIGHER than this row, which counts a cherry-picked re-land, and a revert together with the commit it undoes, once each — a difference of several commits on a file whose history was re-landed or reverted inside the window.
Hotspot: router/src/http/server.rs router/src/http/server.rs:1626— router/src/http/server.rs changed 3 times in last 90 days, and the most complex body those changes touched has cyclomatic complexity 18 in text_embeddings_router::http::server::run at line 1626. 1 of those changes was a fix/bug commit, and the other 2 changed it for other reasons — this file is under both repair and feature pressure. Before the next change lands here, make sure the area it touches is under test, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time. Counted over 2026-08-08..2026-11-06, the 90 days ending at the analysed commit. Reproduce with `git log --since='2026-08-08 13:10:20 +01:00' --until='2026-11-06 13:10:20 +01:00' --full-history --no-merges -- router/src/http/server.rs`: merges are excluded because a merge re-states changes already counted at their own commits, and history is NOT path-simplified because a change that reached the file through a merged branch is still a change to it. That command counts raw commits and can read HIGHER than this row, which counts a cherry-picked re-land, and a revert together with the commit it undoes, once each — a difference of several commits on a file whose history was re-landed or reverted inside the window.
Hotspot: backends/candle/src/models/qwen3.rs backends/candle/src/models/qwen3.rs:518— backends/candle/src/models/qwen3.rs changed 2 times in last 90 days, and the most complex body those changes touched has cyclomatic complexity 25 in Qwen3Model::forward at line 518. 1 of those changes was a fix/bug commit, and the other 1 changed it for other reasons — this file is under both repair and feature pressure. Before the next change lands here, make sure the area it touches is under test, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time. Counted over 2026-08-08..2026-11-06, the 90 days ending at the analysed commit. Reproduce with `git log --since='2026-08-08 13:10:20 +01:00' --until='2026-11-06 13:10:20 +01:00' --full-history --no-merges -- backends/candle/src/models/qwen3.rs`: merges are excluded because a merge re-states changes already counted at their own commits, and history is NOT path-simplified because a change that reached the file through a merged branch is still a change to it. That command counts raw commits and can read HIGHER than this row, which counts a cherry-picked re-land, and a revert together with the commit it undoes, once each — a difference of several commits on a file whose history was re-landed or reverted inside the window.
text_embeddings_backend::init_backend (cognitive 56) backends/src/lib.rs:447— text_embeddings_backend::init_backend has cognitive complexity 56 (threshold 15). Drivers by points: if/else 17 (39 pts), match/switch 7 (16 pts), boolean chains 1 (nesting depth added 31). To reduce it, flatten the nesting: this score is depth rather than breadth — most of its points come from checks stacked inside one another, so the work sits several levels in. Invert each enclosing check into an early exit (a return, or the language's equivalent) so the happy path stays at one level, and where a level cannot be exited early, lift the block it encloses into its own named function.
text_embeddings_backend::download_dense_modules (cognitive 24) backends/src/lib.rs:875— text_embeddings_backend::download_dense_modules has cognitive complexity 24 (threshold 15). Drivers by points: if/else 4 (14 pts), match/switch 3 (6 pts), loops 1 (4 pts) (nesting depth added 16). The drivers above price the dispatch low by construction — a dispatch is charged once however many cases it lists, while each branch inside an arm is charged in full — so most of this count is what the case bodies hold, and the arms are where it can be reduced. To reduce it, keep the dispatch but shrink the arms: move each non-trivial case body into its own named function (or onto the value being matched) so the dispatch reads one line per case, and group related cases into a sub-dispatch. Keep every case explicit, and make the behaviour for cases you do not list a deliberate choice rather than an accident.
text_embeddings_backend::download_onnx (cognitive 23) backends/src/lib.rs:728— text_embeddings_backend::download_onnx has cognitive complexity 23 (threshold 15). Drivers by points: match/switch 8 (17 pts), if/else 2 (3 pts), loops 1 (2 pts), boolean chains 1 (nesting depth added 11). To reduce it, keep the dispatch but shrink the arms: move each non-trivial case body into its own named function (or onto the value being matched) so the dispatch reads one line per case, and group related cases into a sub-dispatch. Keep every case explicit, and make the behaviour for cases you do not list a deliberate choice rather than an accident.
text_embeddings_router::run (cognitive 40) router/src/lib.rs:47— text_embeddings_router::run has cognitive complexity 40 (threshold 15). Drivers by points: if/else 21 (25 pts), match/switch 7 (10 pts), boolean chains 4, loops 1 (nesting depth added 7). To reduce it, split the body: most of this score is breadth rather than depth — checks laid out side by side rather than stacked — so group the statements between the checks into named steps and move each step into its own function. Some of it IS depth: where a check sits inside another whose only job is to reach it, merge the two into one condition, and where an else follows a branch that already returns, drop the trailing else and let the rest of the body continue at one level.
text_embeddings_router::http::server::run (cognitive 24) router/src/http/server.rs:1626— text_embeddings_router::http::server::run has cognitive complexity 24 (threshold 15). Drivers by points: if/else 16 (18 pts), match/switch 4 (6 pts) (nesting depth added 4). To reduce it, split the body: most of this score is breadth rather than depth — checks laid out side by side rather than stacked — so group the statements between the checks into named steps and move each step into its own function. Some of it IS depth: where a check sits inside another whose only job is to reach it, merge the two into one condition, and where an else follows a branch that already returns, drop the trailing else and let the rest of the body continue at one level.
text_embeddings_router::get_backend_model_type (cognitive 18) router/src/lib.rs:416— text_embeddings_router::get_backend_model_type has cognitive complexity 18 (threshold 15). Drivers by points: if/else 6 (12 pts), match/switch 2 (3 pts), boolean chains 2, loops 1 (nesting depth added 7). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
D4 · Code Duplication· Members sharing a duplicated core (4 members, 50+ identical tokens) · ×3
Members sharing a duplicated core (4 members, 50+ identical tokens) backends/candle/src/layers/layer_norm.rs:22— backends/candle/src/layers/layer_norm.rs:22-79 | backends/candle/src/layers/layer_norm.rs:104-161 | backends/candle/src/layers/rms_norm.rs:26-95 | backends/candle/src/models/gemma3.rs:59-131 — These 4 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 4 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 4 times.
Members sharing a duplicated core (4 members, 50+ identical tokens) backends/candle/src/models/flash_jina.rs:234— backends/candle/src/models/flash_jina.rs:234-297 | backends/candle/src/models/flash_jina_code.rs:285-343 | backends/candle/src/models/jina.rs:411-469 | backends/candle/src/models/jina_code.rs:342-396 — These 4 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 4 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 4 times.
Members sharing a duplicated core (4 members, 50+ identical tokens) backends/candle/src/models/flash_mistral.rs:179— backends/candle/src/models/flash_mistral.rs:179-201 | backends/candle/src/models/flash_qwen2.rs:190-212 | backends/candle/src/models/flash_qwen3.rs:238-260 | backends/candle/src/models/qwen3.rs:336-358 — These 4 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 4 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 4 times.
D4 · Code Duplication· Edited copy of a member (11 corresponding lines) · ×3
Edited copy of a member (11 corresponding lines) backends/candle/src/models/flash_mistral.rs:156— backends/candle/src/models/flash_mistral.rs:156-166 | backends/candle/src/models/flash_qwen2.rs:167-177 — These two members are one piece of code written twice and then edited apart: 11 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (11 corresponding lines) backends/candle/src/models/debertav2.rs:755— backends/candle/src/models/debertav2.rs:755-765 | backends/candle/src/models/nomic.rs:322-335 — These two members are one piece of code written twice and then edited apart: 11 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (11 corresponding lines) backends/candle/src/models/flash_distilbert.rs:162— backends/candle/src/models/flash_distilbert.rs:162-173 | backends/candle/src/models/flash_nomic.rs:180-190 — These two members are one piece of code written twice and then edited apart: 11 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (14 corresponding lines) · ×3
Edited copy of a member (14 corresponding lines) backends/candle/src/models/flash_modernbert.rs:220— backends/candle/src/models/flash_modernbert.rs:220-233 | backends/candle/src/models/modernbert.rs:404-422 — These two members are one piece of code written twice and then edited apart: 14 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (14 corresponding lines) backends/candle/src/models/bert.rs:88— backends/candle/src/models/bert.rs:88-101 | backends/candle/src/models/jina.rs:59-74 — These two members are one piece of code written twice and then edited apart: 14 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (14 corresponding lines) backends/python/server/text_embeddings_server/utils/flash_attn.py:86— backends/python/server/text_embeddings_server/utils/flash_attn.py:86-100 | backends/python/server/text_embeddings_server/utils/flash_attn.py:116-129 — These two members are one piece of code written twice and then edited apart: 14 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (8 corresponding lines) · ×3
Edited copy of a member (8 corresponding lines) backends/candle/src/models/flash_nomic.rs:164— backends/candle/src/models/flash_nomic.rs:164-171 | backends/candle/src/models/nomic.rs:663-670 — These two members are one piece of code written twice and then edited apart: 8 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (8 corresponding lines) backends/candle/src/models/bert.rs:341— backends/candle/src/models/bert.rs:341-348 | backends/candle/src/models/flash_bert.rs:194-201 — These two members are one piece of code written twice and then edited apart: 8 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (8 corresponding lines) backends/candle/src/models/distilbert.rs:301— backends/candle/src/models/distilbert.rs:301-308 | backends/candle/src/models/flash_distilbert.rs:153-160 — These two members are one piece of code written twice and then edited apart: 8 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Duplicated block (38 lines × 2) backends/candle/src/models/bert.rs:861— backends/candle/src/models/bert.rs:861-898 | backends/candle/src/models/distilbert.rs:654-691 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:861` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/bert.rs:860` calls `index_select` and `backends/candle/src/models/distilbert.rs:652` does not — after which the two agree again for 4 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (38 lines × 2) backends/candle/src/models/flash_jina_code.rs:175— backends/candle/src/models/flash_jina_code.rs:175-212 | backends/candle/src/models/jina_code.rs:227-264 — before extracting anything, compare `backends/candle/src/models/flash_jina_code.rs` and `backends/candle/src/models/jina_code.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 121 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_jina_code.rs:175` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (38 lines × 2) backends/candle/src/models/bert.rs:583— backends/candle/src/models/bert.rs:583-620 | backends/candle/src/models/flash_bert.rs:245-282 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/flash_bert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:583` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (22 lines × 2) backends/candle/src/models/flash_modernbert.rs:275— backends/candle/src/models/flash_modernbert.rs:275-296 | backends/candle/src/models/modernbert.rs:527-548 — before extracting anything, compare `backends/candle/src/models/flash_modernbert.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 113 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_modernbert.rs:275` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (22 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:231— backends/python/server/text_embeddings_server/models/flash_mistral.py:231-252 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:220-241 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (22 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:124— backends/python/server/text_embeddings_server/models/flash_mistral.py:124-145 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:312-333 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. The `return` at the foot of the matched lines is the enclosing body's own terminal exit, not an early one: it moves with them unchanged, and each site calls the extracted unit from the position that `return` occupied — no decision has to be handed back and re-acted on.
Duplicated block (18 lines × 2) backends/candle/src/models/flash_bert.rs:273— backends/candle/src/models/flash_bert.rs:273-290 | backends/candle/src/models/flash_bert.rs:350-367 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:273` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (18 lines × 2) core/src/infer.rs:351— core/src/infer.rs:351-368 | core/src/infer.rs:417-434 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `core/src/infer.rs:351` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (18 lines × 2) router/src/grpc/server.rs:981— router/src/grpc/server.rs:981-998 | router/src/http/server.rs:406-423 — `router/src/grpc/server.rs` and `router/src/http/server.rs` are one unit implemented once per sibling directory, so they are most likely parallel implementations of one contract rather than a copy of each other — this scan matched 7 separate duplicated blocks between them, totalling at least 91 lines. If both are selected at run time, neither can be retired in favour of the other, and the lines that DIFFER between them are the reason both exist. The move that pays here is to hoist the identical part into a shared location the whole family can reach and give what differs a parameter or a seam, so a change lands once instead of once per sibling; extracting one helper per block leaves every sibling to drift on its own. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:981` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11 lines × 3) backends/candle/src/models/bert.rs:184— backends/candle/src/models/bert.rs:184-194 | backends/candle/src/models/distilbert.rs:125-135 | backends/candle/src/models/jina.rs:157-167 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:184` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11 lines × 3) backends/candle/src/models/flash_bert.rs:88— backends/candle/src/models/flash_bert.rs:88-98 | backends/candle/src/models/flash_distilbert.rs:69-79 | backends/candle/src/models/flash_jina.rs:90-100 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:88` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/flash_bert.rs:83` calls `clone` and `backends/candle/src/models/flash_distilbert.rs:64` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (11 lines × 3) backends/candle/src/models/flash_qwen3.rs:36— backends/candle/src/models/flash_qwen3.rs:36-46 | backends/candle/src/models/gemma3.rs:163-173 | backends/candle/src/models/qwen3.rs:62-72 — before extracting anything, compare `backends/candle/src/models/flash_qwen3.rs` and `backends/candle/src/models/gemma3.rs` as WHOLE FILES: this scan already matched 3 separate duplicated blocks between them, totalling at least 37 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&Qwen3Config` in one and `&Gemma3Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (9 lines × 3) backends/candle/src/models/distilbert.rs:114— backends/candle/src/models/distilbert.rs:114-122 | backends/candle/src/models/gte.rs:86-94 | backends/candle/src/models/nomic.rs:508-516 — before extracting anything, compare `backends/candle/src/models/distilbert.rs` and `backends/candle/src/models/gte.rs` as WHOLE FILES: this scan already matched 24 separate duplicated blocks between them, totalling at least 362 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (9 lines × 3) backends/candle/src/models/flash_bert.rs:81— backends/candle/src/models/flash_bert.rs:81-89 | backends/candle/src/models/flash_jina.rs:83-91 | backends/candle/src/models/flash_jina_code.rs:99-107 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_jina.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 293 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (9 lines × 3) router/src/http/server.rs:696— router/src/http/server.rs:696-704 | router/src/http/server.rs:882-890 | router/src/http/server.rs:1059-1067 — all 3 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:696` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 4) backends/candle/src/layers/layer_norm.rs:31— backends/candle/src/layers/layer_norm.rs:31-38 | backends/candle/src/layers/layer_norm.rs:113-120 | backends/candle/src/layers/rms_norm.rs:40-47 | backends/candle/src/models/gemma3.rs:73-81 — before extracting anything, compare `backends/candle/src/layers/rms_norm.rs` and `backends/candle/src/models/gemma3.rs` as WHOLE FILES: this scan already matched 4 separate duplicated blocks between them, totalling at least 63 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. The two sit in different directories, so one cannot simply be deleted in favour of the other while both are reached separately: hoist the shared part into a location both already depend on and have each file call it, and retire whichever file turns out to have no caller of its own left. Extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/layers/layer_norm.rs:39` calls `sum_keepdim`, `broadcast_sub` and `backends/candle/src/layers/rms_norm.rs:48` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (8 lines × 4) backends/candle/src/models/flash_bert.rs:389— backends/candle/src/models/flash_bert.rs:389-396 | backends/candle/src/models/flash_distilbert.rs:260-267 | backends/candle/src/models/flash_jina.rs:319-326 | backends/candle/src/models/flash_jina_code.rs:365-372 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:389` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 4) backends/candle/src/models/gemma3.rs:356— backends/candle/src/models/gemma3.rs:356-363 | backends/candle/src/models/gemma3.rs:365-372 | backends/candle/src/models/qwen3.rs:189-196 | backends/candle/src/models/qwen3.rs:198-205 — before extracting anything, compare `backends/candle/src/models/gemma3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 15 separate duplicated blocks between them, totalling at least 209 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (8 lines × 3) backends/candle/src/models/bert.rs:325— backends/candle/src/models/bert.rs:325-332 | backends/candle/src/models/flash_bert.rs:178-185 | backends/candle/src/models/mpnet.rs:290-297 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/flash_bert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:325` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 3) backends/candle/src/models/bert.rs:810— backends/candle/src/models/bert.rs:810-817 | backends/candle/src/models/jina.rs:618-625 | backends/candle/src/models/jina_code.rs:545-552 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 37 separate duplicated blocks between them, totalling at least 547 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:810` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (8 lines × 3) core/src/infer.rs:129— core/src/infer.rs:129-136 | core/src/infer.rs:180-187 | core/src/infer.rs:335-342 — all 3 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `core/src/infer.rs:129` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 2) router/src/grpc/server.rs:310— router/src/grpc/server.rs:310-317 | router/src/http/server.rs:155-162 — `router/src/grpc/server.rs` and `router/src/http/server.rs` are one unit implemented once per sibling directory, so they are most likely parallel implementations of one contract rather than a copy of each other — this scan matched 7 separate duplicated blocks between them, totalling at least 91 lines. If both are selected at run time, neither can be retired in favour of the other, and the lines that DIFFER between them are the reason both exist. The move that pays here is to hoist the identical part into a shared location the whole family can reach and give what differs a parameter or a seam, so a change lands once instead of once per sibling; extracting one helper per block leaves every sibling to drift on its own. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:310` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 2) backends/candle/src/models/flash_distilbert.rs:42— backends/candle/src/models/flash_distilbert.rs:42-49 | backends/candle/src/models/flash_gte.rs:37-44 — before extracting anything, compare `backends/candle/src/models/flash_distilbert.rs` and `backends/candle/src/models/flash_gte.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 107 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_distilbert.rs:42` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:362— backends/python/server/text_embeddings_server/models/flash_mistral.py:362-369 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:382-389 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (7 lines × 3) backends/candle/src/models/bert.rs:407— backends/candle/src/models/bert.rs:407-413 | backends/candle/src/models/gte.rs:323-330 | backends/candle/src/models/jina.rs:381-387 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/gte.rs` as WHOLE FILES: this scan already matched 25 separate duplicated blocks between them, totalling at least 400 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (7 lines × 3) router/src/grpc/server.rs:1022— router/src/grpc/server.rs:1022-1028 | router/src/grpc/server.rs:1283-1289 | router/src/http/server.rs:443-449 — `router/src/grpc/server.rs` and `router/src/http/server.rs` are one unit implemented once per sibling directory, so they are most likely parallel implementations of one contract rather than a copy of each other — this scan matched 7 separate duplicated blocks between them, totalling at least 91 lines. If both are selected at run time, neither can be retired in favour of the other, and the lines that DIFFER between them are the reason both exist. The move that pays here is to hoist the identical part into a shared location the whole family can reach and give what differs a parameter or a seam, so a change lands once instead of once per sibling; extracting one helper per block leaves every sibling to drift on its own.
Duplicated block (7 lines × 3) backends/candle/src/models/flash_mistral.rs:157— backends/candle/src/models/flash_mistral.rs:157-163 | backends/candle/src/models/flash_qwen2.rs:168-174 | backends/candle/src/models/flash_qwen3.rs:216-222 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (5 lines × 2) backends/candle/src/models/flash_gte.rs:74— backends/candle/src/models/flash_gte.rs:74-78 | backends/candle/src/models/flash_modernbert.rs:86-90 — before extracting anything, compare `backends/candle/src/models/flash_gte.rs` and `backends/candle/src/models/flash_modernbert.rs` as WHOLE FILES: this scan already matched 7 separate duplicated blocks between them, totalling at least 133 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (5 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:30— backends/python/server/text_embeddings_server/models/flash_mistral.py:30-34 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:41-45 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. The `return` at the foot of the matched lines is the enclosing body's own terminal exit, not an early one: it moves with them unchanged, and each site calls the extracted unit from the position that `return` occupied — no decision has to be handed back and re-acted on.
Duplicated block (5 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:117— backends/python/server/text_embeddings_server/models/flash_mistral.py:117-121 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:305-309 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `MistralConfig` in one and `Qwen3Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
text_embeddings_router::run (cyclomatic 28) router/src/lib.rs:47— text_embeddings_router::run has cyclomatic complexity 28 (threshold 15). To reduce it, split the body: these branches sit side by side rather than nested inside one another, so extracting each one on its own would leave a function per branch. Group the statements between the checks into named steps and move each step into its own function, so the body reads as a short sequence of named stages.
text_embeddings_router::http::server::run (cyclomatic 18) router/src/http/server.rs:1626— text_embeddings_router::http::server::run has cyclomatic complexity 18 (threshold 15). To reduce it, split the body: these branches sit side by side rather than nested inside one another, so extracting each one on its own would leave a function per branch. Group the statements between the checks into named steps and move each step into its own function, so the body reads as a short sequence of named stages.
OrtBackend::embed (cognitive 112) backends/ort/src/lib.rs:402— OrtBackend::embed has cognitive complexity 112 (threshold 15). Drivers by points: loops 11 (50 pts), if/else 17 (39 pts), match/switch 5 (20 pts), boolean chains 3 (nesting depth added 76). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
OrtBackend::prepare_inputs (cognitive 44) backends/ort/src/lib.rs:174— OrtBackend::prepare_inputs has cognitive complexity 44 (threshold 15). Drivers by points: loops 7 (26 pts), if/else 10 (16 pts), match/switch 1 (2 pts) (nesting depth added 26). To reduce it, break up the iteration: give each loop body a named function, and split a multi-phase loop into one function per phase so no single body carries the whole pipeline.
text_embeddings_core::queue::queue_blocking_task (cognitive 31) core/src/queue.rs:99— text_embeddings_core::queue::queue_blocking_task has cognitive complexity 31 (threshold 15). Drivers by points: if/else 7 (21 pts), match/switch 2 (6 pts), loops 2 (4 pts) (nesting depth added 20). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
text_embeddings_core::tokenization::tokenizer_worker (cognitive 17) core/src/tokenization.rs:171— text_embeddings_core::tokenization::tokenizer_worker has cognitive complexity 17 (threshold 15). Drivers by points: if/else 3 (9 pts), match/switch 3 (7 pts), loops 1 (nesting depth added 10). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
ClassTooLong: TextEmbeddingsService router/src/grpc/server.rs:46— ClassTooLong — 908 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted), 9 methods, 7 blocks, lines 46-1364. The bar is 400 significant lines; this is 508 over it, 2.27× the bar. To reduce it, group the members that share the same data into a smaller type of their own and delegate to it, so no single type carries every responsibility.
ClassTooLong: OrtBackend backends/ort/src/lib.rs:61— ClassTooLong — 450 significant lines (blank, comment-only and punctuation-only lines excluded, and inline test code — #[cfg(test)] modules and bare #[test] functions — not counted), 3 methods, 4 blocks, lines 61-722. The bar is 400 significant lines; this is 50 over it, 1.13× the bar. To reduce it, group the members that share the same data into a smaller type of their own and delegate to it, so no single type carries every responsibility.
D30 · Dependency Vulnerabilities· Medium advisory (unsound) · ×2
REDACTED
REDACTED
D4 · Code Duplication· Edited copy of a member (30 corresponding lines) · ×2
Edited copy of a member (30 corresponding lines) backends/candle/src/models/flash_modernbert.rs:130— backends/candle/src/models/flash_modernbert.rs:130-159 | backends/candle/src/models/modernbert.rs:311-340 — These two members are one piece of code written twice and then edited apart: 30 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (30 corresponding lines) backends/candle/src/layers/cublaslt.rs:58— backends/candle/src/layers/cublaslt.rs:58-87 | backends/candle/src/layers/cublaslt.rs:99-128 — These two members are one piece of code written twice and then edited apart: 30 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Members sharing a duplicated core (10 members, 50+ identical tokens) · ×2
Members sharing a duplicated core (10 members, 50+ identical tokens) backends/candle/src/models/bert.rs:118— backends/candle/src/models/bert.rs:118-168 | backends/candle/src/models/distilbert.rs:77-111 | backends/candle/src/models/flash_bert.rs:25-73 | backends/candle/src/models/flash_distilbert.rs:24-56 | backends/candle/src/models/flash_jina.rs:26-75 | backends/candle/src/models/flash_jina_code.rs:29-91 | backends/candle/src/models/flash_qwen2.rs:26-75 | backends/candle/src/models/jina.rs:91-141 | backends/candle/src/models/jina_code.rs:26-89 | backends/candle/src/models/mpnet.rs:84-131 — These 10 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 10 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 10 times.
Members sharing a duplicated core (10 members, 50+ identical tokens) router/src/grpc/server.rs:880— router/src/grpc/server.rs:880-1055 | router/src/http/server.rs:109-278 | router/src/http/server.rs:310-472 | router/src/http/server.rs:503-558 | router/src/http/server.rs:589-738 | router/src/http/server.rs:769-924 | router/src/http/server.rs:956-1101 | router/src/http/server.rs:1133-1328 | router/src/http/server.rs:1351-1434 | router/src/http/server.rs:1457-1509 — These 10 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 10 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 10 times.
D4 · Code Duplication· Edited copy of a member (31 corresponding lines) · ×2
Edited copy of a member (31 corresponding lines) backends/candle/src/models/bert.rs:51— backends/candle/src/models/bert.rs:51-81 | backends/candle/src/models/jina.rs:19-52 — These two members are one piece of code written twice and then edited apart: 31 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (31 corresponding lines) backends/candle/src/models/flash_gte.rs:242— backends/candle/src/models/flash_gte.rs:242-274 | backends/candle/src/models/gte.rs:457-487 — These two members are one piece of code written twice and then edited apart: 31 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (13 corresponding lines) · ×2
Edited copy of a member (13 corresponding lines) backends/python/server/text_embeddings_server/models/jinaBert_model.py:238— backends/python/server/text_embeddings_server/models/jinaBert_model.py:238-250 | backends/python/server/text_embeddings_server/models/jinaBert_model.py:302-320 — These two members are one piece of code written twice and then edited apart: 13 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (13 corresponding lines) backends/candle/src/models/bert.rs:406— backends/candle/src/models/bert.rs:406-418 | backends/candle/src/models/jina.rs:380-392 — These two members are one piece of code written twice and then edited apart: 13 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (12 corresponding lines) · ×2
Edited copy of a member (12 corresponding lines) backends/candle/src/models/flash_bert.rs:174— backends/candle/src/models/flash_bert.rs:174-185 | backends/candle/src/models/mpnet.rs:283-297 — These two members are one piece of code written twice and then edited apart: 12 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Edited copy of a member (12 corresponding lines) backends/candle/src/models/bert.rs:321— backends/candle/src/models/bert.rs:321-332 | backends/candle/src/models/debertav2.rs:774-791 — These two members are one piece of code written twice and then edited apart: 12 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Duplicated block (26 lines × 2) backends/candle/src/models/flash_qwen2.rs:293— backends/candle/src/models/flash_qwen2.rs:293-318 | backends/candle/src/models/flash_qwen3.rs:362-387 — before extracting anything, compare `backends/candle/src/models/flash_qwen2.rs` and `backends/candle/src/models/flash_qwen3.rs` as WHOLE FILES: this scan already matched 15 separate duplicated blocks between them, totalling at least 269 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_qwen2.rs:293` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&Qwen2Config` in one and `&Qwen3Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (26 lines × 2) backends/candle/src/models/gemma3.rs:887— backends/candle/src/models/gemma3.rs:887-912 | backends/candle/src/models/qwen3.rs:723-748 — before extracting anything, compare `backends/candle/src/models/gemma3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 15 separate duplicated blocks between them, totalling at least 209 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/gemma3.rs:887` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (24 lines × 2) backends/candle/src/models/bert.rs:671— backends/candle/src/models/bert.rs:671-694 | backends/candle/src/models/flash_bert.rs:336-359 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/flash_bert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:671` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (24 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:89— backends/python/server/text_embeddings_server/models/flash_mistral.py:89-112 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:91-114 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (23 lines × 2) backends/candle/src/models/flash_nomic.rs:24— backends/candle/src/models/flash_nomic.rs:24-46 | backends/candle/src/models/nomic.rs:465-487 — before extracting anything, compare `backends/candle/src/models/flash_nomic.rs` and `backends/candle/src/models/nomic.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 106 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (23 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:50— backends/python/server/text_embeddings_server/models/flash_mistral.py:50-72 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:52-74 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. The `return` at the foot of the matched lines is the enclosing body's own terminal exit, not an early one: it moves with them unchanged, and each site calls the extracted unit from the position that `return` occupied — no decision has to be handed back and re-acted on. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `MistralConfig` in one and `Qwen3Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (21 lines × 4) backends/candle/src/models/flash_mistral.rs:129— backends/candle/src/models/flash_mistral.rs:129-149 | backends/candle/src/models/flash_qwen2.rs:140-160 | backends/candle/src/models/flash_qwen3.rs:188-208 | backends/candle/src/models/qwen3.rs:286-306 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_mistral.rs:129` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&MistralConfig` in one and `&Qwen2Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (21 lines × 4) router/src/http/server.rs:258— router/src/http/server.rs:258-278 | router/src/http/server.rs:718-738 | router/src/http/server.rs:904-924 | router/src/http/server.rs:1081-1101 — all 4 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:258` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (21 lines × 2) backends/candle/src/models/modernbert.rs:669— backends/candle/src/models/modernbert.rs:669-689 | backends/candle/src/models/mpnet.rs:547-567 — before extracting anything, compare `backends/candle/src/models/modernbert.rs` and `backends/candle/src/models/mpnet.rs` as WHOLE FILES: this scan already matched 14 separate duplicated blocks between them, totalling at least 245 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/modernbert.rs:669` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (21 lines × 2) router/src/http/server.rs:696— router/src/http/server.rs:696-716 | router/src/http/server.rs:1059-1079 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:696` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (19 lines × 2) backends/candle/src/models/bert.rs:376— backends/candle/src/models/bert.rs:376-394 | backends/candle/src/models/jina.rs:350-368 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 37 separate duplicated blocks between them, totalling at least 547 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (19 lines × 2) backends/candle/src/models/flash_mistral.rs:283— backends/candle/src/models/flash_mistral.rs:283-301 | backends/candle/src/models/flash_qwen2.rs:304-322 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_mistral.rs:283` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&MistralConfig` in one and `&Qwen2Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (15–16 lines × 10) backends/candle/src/models/bert.rs:231— backends/candle/src/models/bert.rs:231-246 | backends/candle/src/models/distilbert.rs:172-187 | backends/candle/src/models/gemma3.rs:440-454 | backends/candle/src/models/gte.rs:147-162 | backends/candle/src/models/jina.rs:204-219 | backends/candle/src/models/jina_code.rs:173-188 | backends/candle/src/models/modernbert.rs:273-287 | backends/candle/src/models/mpnet.rs:192-207 | backends/candle/src/models/nomic.rs:569-584 | backends/candle/src/models/qwen3.rs:243-257 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:231` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (15–16 lines × 10) backends/candle/src/models/flash_bert.rs:441— backends/candle/src/models/flash_bert.rs:441-456 | backends/candle/src/models/flash_distilbert.rs:310-325 | backends/candle/src/models/flash_gte.rs:367-382 | backends/candle/src/models/flash_jina.rs:371-386 | backends/candle/src/models/flash_jina_code.rs:417-432 | backends/candle/src/models/flash_mistral.rs:387-402 | backends/candle/src/models/flash_modernbert.rs:398-412 | backends/candle/src/models/flash_nomic.rs:366-381 | backends/candle/src/models/flash_qwen2.rs:408-423 | backends/candle/src/models/flash_qwen3.rs:485-500 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:441` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (14 lines × 3) backends/candle/src/models/bert.rs:149— backends/candle/src/models/bert.rs:149-162 | backends/candle/src/models/jina.rs:122-135 | backends/candle/src/models/jina_code.rs:68-81 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 37 separate duplicated blocks between them, totalling at least 547 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:149` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (14 lines × 3) backends/candle/src/models/flash_bert.rs:54— backends/candle/src/models/flash_bert.rs:54-67 | backends/candle/src/models/flash_jina.rs:55-68 | backends/candle/src/models/flash_jina_code.rs:69-82 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_jina.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 293 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:54` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (13–14 lines × 2) backends/candle/src/models/flash_jina.rs:101— backends/candle/src/models/flash_jina.rs:101-113 | backends/candle/src/models/flash_jina_code.rs:137-150 — before extracting anything, compare `backends/candle/src/models/flash_jina.rs` and `backends/candle/src/models/flash_jina_code.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 288 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_jina.rs:101` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (13–14 lines × 2) backends/candle/src/models/flash_nomic.rs:229— backends/candle/src/models/flash_nomic.rs:229-241 | backends/candle/src/models/nomic.rs:726-739 — before extracting anything, compare `backends/candle/src/models/flash_nomic.rs` and `backends/candle/src/models/nomic.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 106 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_nomic.rs:229` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (12 lines × 4) backends/candle/src/models/flash_gte.rs:203— backends/candle/src/models/flash_gte.rs:203-214 | backends/candle/src/models/flash_mistral.rs:270-281 | backends/candle/src/models/flash_qwen2.rs:291-302 | backends/candle/src/models/flash_qwen3.rs:360-371 — before extracting anything, compare `backends/candle/src/models/flash_gte.rs` and `backends/candle/src/models/flash_mistral.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 163 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_gte.rs:203` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `>EConfig` in one and `&MistralConfig` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (12 lines × 4) backends/candle/src/models/bert.rs:350— backends/candle/src/models/bert.rs:350-361 | backends/candle/src/models/distilbert.rs:310-321 | backends/candle/src/models/jina.rs:324-335 | backends/candle/src/models/jina_code.rs:313-324 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (10 lines × 3) router/src/grpc/server.rs:604— router/src/grpc/server.rs:604-613 | router/src/grpc/server.rs:644-653 | router/src/grpc/server.rs:684-693 — all 3 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:604` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 3) router/src/http/server.rs:597— router/src/http/server.rs:597-606 | router/src/http/server.rs:786-795 | router/src/http/server.rs:964-973 — all 3 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:597` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (7 lines × 5) backends/candle/src/models/flash_bert.rs:370— backends/candle/src/models/flash_bert.rs:370-376 | backends/candle/src/models/flash_distilbert.rs:244-250 | backends/candle/src/models/flash_jina.rs:300-306 | backends/candle/src/models/flash_jina_code.rs:346-352 | backends/candle/src/models/flash_nomic.rs:273-279 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/flash_bert.rs:378` calls `from_vec` and `backends/candle/src/models/flash_distilbert.rs:252` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (7 lines × 5) backends/candle/src/models/flash_distilbert.rs:250— backends/candle/src/models/flash_distilbert.rs:250-256 | backends/candle/src/models/flash_mistral.rs:310-316 | backends/candle/src/models/flash_modernbert.rs:334-340 | backends/candle/src/models/flash_qwen2.rs:331-337 | backends/candle/src/models/flash_qwen3.rs:401-407 — before extracting anything, compare `backends/candle/src/models/flash_distilbert.rs` and `backends/candle/src/models/flash_mistral.rs` as WHOLE FILES: this scan already matched 5 separate duplicated blocks between them, totalling at least 82 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (7 lines × 4) backends/candle/src/models/bert.rs:253— backends/candle/src/models/bert.rs:253-259 | backends/candle/src/models/distilbert.rs:194-200 | backends/candle/src/models/jina.rs:226-232 | backends/candle/src/models/jina_code.rs:195-201 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:253` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/bert.rs:260` calls `forward`, `Some` and `backends/candle/src/models/distilbert.rs:202` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (7 lines × 4) backends/candle/src/models/flash_gte.rs:277— backends/candle/src/models/flash_gte.rs:277-283 | backends/candle/src/models/flash_mistral.rs:304-310 | backends/candle/src/models/flash_qwen2.rs:325-331 | backends/candle/src/models/flash_qwen3.rs:395-401 — before extracting anything, compare `backends/candle/src/models/flash_gte.rs` and `backends/candle/src/models/flash_mistral.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 163 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/flash_gte.rs:285` calls `from_vec` and `backends/candle/src/models/flash_mistral.rs:312` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
CandleBackend::new (cyclomatic 79) backends/candle/src/lib.rs:163— CandleBackend::new has cyclomatic complexity 79 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
OrtBackend::embed (cyclomatic 33) backends/ort/src/lib.rs:402— OrtBackend::embed has cyclomatic complexity 33 (threshold 15). To reduce it, break up the iteration: give each loop body a named function, and split a multi-phase loop into one function per phase so no single body carries the whole pipeline.
models.get_model (cyclomatic 28) backends/python/server/text_embeddings_server/models/__init__.py:63— models.get_model has cyclomatic complexity 28 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
Qwen3Model::forward (cyclomatic 25) backends/candle/src/models/qwen3.rs:518— Qwen3Model::forward has cyclomatic complexity 25 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
JinaBertModel::forward (cyclomatic 24) backends/candle/src/models/jina.rs:471— JinaBertModel::forward has cyclomatic complexity 24 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
JinaCodeBertModel::forward (cyclomatic 24) backends/candle/src/models/jina_code.rs:398— JinaCodeBertModel::forward has cyclomatic complexity 24 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
BertModel::forward (cyclomatic 23) backends/candle/src/models/bert.rs:706— BertModel::forward has cyclomatic complexity 23 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
DistilBertModel::forward (cyclomatic 23) backends/candle/src/models/distilbert.rs:508— DistilBertModel::forward has cyclomatic complexity 23 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
NomicBertModel::forward (cyclomatic 23) backends/candle/src/models/nomic.rs:773— NomicBertModel::forward has cyclomatic complexity 23 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
text_embeddings_backend::init_backend (cyclomatic 23) backends/src/lib.rs:447— text_embeddings_backend::init_backend has cyclomatic complexity 23 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
DebertaV2Model::forward (cyclomatic 22) backends/candle/src/models/debertav2.rs:1121— DebertaV2Model::forward has cyclomatic complexity 22 (threshold 15). To reduce it, separate the branches: extract each independent case into its own named function so the top-level body reads as a short sequence of named decisions.
GTEModel::forward (cyclomatic 21) backends/candle/src/models/gte.rs:489— GTEModel::forward has cyclomatic complexity 21 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
MPNetModel::forward (cyclomatic 21) backends/candle/src/models/mpnet.rs:506— MPNetModel::forward has cyclomatic complexity 21 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
Gemma3Model::forward (cyclomatic 20) backends/candle/src/models/gemma3.rs:721— Gemma3Model::forward has cyclomatic complexity 20 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
ModernBertModel::forward (cyclomatic 20) backends/candle/src/models/modernbert.rs:633— ModernBertModel::forward has cyclomatic complexity 20 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashNomicBertModel::forward (cyclomatic 18) backends/candle/src/models/flash_nomic.rs:272— FlashNomicBertModel::forward has cyclomatic complexity 18 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashQwen3Model::forward (cyclomatic 18) backends/candle/src/models/flash_qwen3.rs:394— FlashQwen3Model::forward has cyclomatic complexity 18 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashBertModel::forward (cyclomatic 17) backends/candle/src/models/flash_bert.rs:369— FlashBertModel::forward has cyclomatic complexity 17 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashDistilBertModel::forward (cyclomatic 17) backends/candle/src/models/flash_distilbert.rs:243— FlashDistilBertModel::forward has cyclomatic complexity 17 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashGTEModel::forward (cyclomatic 17) backends/candle/src/models/flash_gte.rs:276— FlashGTEModel::forward has cyclomatic complexity 17 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashMistralModel::forward (cyclomatic 17) backends/candle/src/models/flash_mistral.rs:303— FlashMistralModel::forward has cyclomatic complexity 17 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashModernBertModel::forward (cyclomatic 17) backends/candle/src/models/flash_modernbert.rs:327— FlashModernBertModel::forward has cyclomatic complexity 17 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashQwen2Model::forward (cyclomatic 17) backends/candle/src/models/flash_qwen2.rs:324— FlashQwen2Model::forward has cyclomatic complexity 17 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashJinaBertModel::forward (cyclomatic 16) backends/candle/src/models/flash_jina.rs:299— FlashJinaBertModel::forward has cyclomatic complexity 16 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
FlashJinaCodeBertModel::forward (cyclomatic 16) backends/candle/src/models/flash_jina_code.rs:345— FlashJinaCodeBertModel::forward has cyclomatic complexity 16 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
HackComment router/src/main.rs:237— // Hack to trim pages regularly — a workaround marked in source: record what it is compensating for and what would allow its removal (the upstream fix, the API it is waiting on, the invariant it restores), so the next reader can judge whether it is still needed rather than rediscovering why it is there.
CandleBackend::new (cognitive 95) backends/candle/src/lib.rs:163— CandleBackend::new has cognitive complexity 95 (threshold 15). Drivers by points: if/else 44 (68 pts), boolean chains 11, match/switch 7 (11 pts), loops 2 (5 pts) (nesting depth added 31). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
Qwen3Model::forward (cognitive 62) backends/candle/src/models/qwen3.rs:518— Qwen3Model::forward has cognitive complexity 62 (threshold 15). Drivers by points: if/else 25 (40 pts), loops 6 (17 pts), match/switch 2 (4 pts), boolean chains 1 (nesting depth added 28). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
JinaBertModel::forward (cognitive 58) backends/candle/src/models/jina.rs:471— JinaBertModel::forward has cognitive complexity 58 (threshold 15). Drivers by points: if/else 20 (36 pts), loops 5 (16 pts), match/switch 2 (4 pts), boolean chains 2 (nesting depth added 29). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
JinaCodeBertModel::forward (cognitive 58) backends/candle/src/models/jina_code.rs:398— JinaCodeBertModel::forward has cognitive complexity 58 (threshold 15). Drivers by points: if/else 20 (36 pts), loops 5 (16 pts), match/switch 2 (4 pts), boolean chains 2 (nesting depth added 29). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
BertModel::forward (cognitive 55) backends/candle/src/models/bert.rs:706— BertModel::forward has cognitive complexity 55 (threshold 15). Drivers by points: if/else 17 (33 pts), loops 5 (16 pts), match/switch 2 (4 pts), boolean chains 2 (nesting depth added 29). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
DistilBertModel::forward (cognitive 55) backends/candle/src/models/distilbert.rs:508— DistilBertModel::forward has cognitive complexity 55 (threshold 15). Drivers by points: if/else 17 (33 pts), loops 5 (16 pts), match/switch 2 (4 pts), boolean chains 2 (nesting depth added 29). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
NomicBertModel::forward (cognitive 51) backends/candle/src/models/nomic.rs:773— NomicBertModel::forward has cognitive complexity 51 (threshold 15). Drivers by points: if/else 17 (28 pts), loops 5 (16 pts), match/switch 2 (4 pts), boolean chains 3 (nesting depth added 24). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
Gemma3Model::forward (cognitive 50) backends/candle/src/models/gemma3.rs:721— Gemma3Model::forward has cognitive complexity 50 (threshold 15). Drivers by points: if/else 19 (30 pts), loops 6 (17 pts), match/switch 1 (2 pts), boolean chains 1 (nesting depth added 23). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
GTEModel::forward (cognitive 48) backends/candle/src/models/gte.rs:489— GTEModel::forward has cognitive complexity 48 (threshold 15). Drivers by points: if/else 15 (26 pts), loops 5 (16 pts), match/switch 2 (4 pts), boolean chains 2 (nesting depth added 24). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
MPNetModel::forward (cognitive 46) backends/candle/src/models/mpnet.rs:506— MPNetModel::forward has cognitive complexity 46 (threshold 15). Drivers by points: if/else 13 (22 pts), loops 6 (18 pts), match/switch 2 (4 pts), boolean chains 2 (nesting depth added 23). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
DebertaV2Model::forward (cognitive 45) backends/candle/src/models/debertav2.rs:1121— DebertaV2Model::forward has cognitive complexity 45 (threshold 15). Drivers by points: if/else 13 (22 pts), loops 5 (16 pts), match/switch 3 (5 pts), boolean chains 2 (nesting depth added 22). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
ModernBertModel::forward (cognitive 45) backends/candle/src/models/modernbert.rs:633— ModernBertModel::forward has cognitive complexity 45 (threshold 15). Drivers by points: if/else 13 (22 pts), loops 5 (16 pts), match/switch 3 (5 pts), boolean chains 2 (nesting depth added 22). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashBertModel::forward (cognitive 38) backends/candle/src/models/flash_bert.rs:369— FlashBertModel::forward has cognitive complexity 38 (threshold 15). Drivers by points: if/else 13 (23 pts), loops 2 (7 pts), match/switch 3 (7 pts), boolean chains 1 (nesting depth added 19). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashDistilBertModel::forward (cognitive 38) backends/candle/src/models/flash_distilbert.rs:243— FlashDistilBertModel::forward has cognitive complexity 38 (threshold 15). Drivers by points: if/else 13 (23 pts), loops 2 (7 pts), match/switch 3 (7 pts), boolean chains 1 (nesting depth added 19). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashNomicBertModel::forward (cognitive 37) backends/candle/src/models/flash_nomic.rs:272— FlashNomicBertModel::forward has cognitive complexity 37 (threshold 15). Drivers by points: if/else 13 (21 pts), loops 2 (7 pts), match/switch 3 (7 pts), boolean chains 2 (nesting depth added 17). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashQwen3Model::forward (cognitive 37) backends/candle/src/models/flash_qwen3.rs:394— FlashQwen3Model::forward has cognitive complexity 37 (threshold 15). Drivers by points: if/else 13 (21 pts), loops 3 (8 pts), match/switch 3 (7 pts), boolean chains 1 (nesting depth added 17). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
models.get_model (cognitive 37) backends/python/server/text_embeddings_server/models/__init__.py:63— models.get_model has cognitive complexity 37 (threshold 15). Drivers by points: if/else 13 (20 pts), boolean chains 10, error handling 3 (7 pts) (nesting depth added 11). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashGTEModel::forward (cognitive 35) backends/candle/src/models/flash_gte.rs:276— FlashGTEModel::forward has cognitive complexity 35 (threshold 15). Drivers by points: if/else 11 (19 pts), loops 3 (8 pts), match/switch 3 (7 pts), boolean chains 1 (nesting depth added 17). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashMistralModel::forward (cognitive 35) backends/candle/src/models/flash_mistral.rs:303— FlashMistralModel::forward has cognitive complexity 35 (threshold 15). Drivers by points: if/else 11 (19 pts), loops 3 (8 pts), match/switch 3 (7 pts), boolean chains 1 (nesting depth added 17). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashModernBertModel::forward (cognitive 35) backends/candle/src/models/flash_modernbert.rs:327— FlashModernBertModel::forward has cognitive complexity 35 (threshold 15). Drivers by points: if/else 11 (19 pts), loops 3 (8 pts), match/switch 3 (7 pts), boolean chains 1 (nesting depth added 17). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashQwen2Model::forward (cognitive 35) backends/candle/src/models/flash_qwen2.rs:324— FlashQwen2Model::forward has cognitive complexity 35 (threshold 15). Drivers by points: if/else 11 (19 pts), loops 3 (8 pts), match/switch 3 (7 pts), boolean chains 1 (nesting depth added 17). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashJinaBertModel::forward (cognitive 34) backends/candle/src/models/flash_jina.rs:299— FlashJinaBertModel::forward has cognitive complexity 34 (threshold 15). Drivers by points: if/else 11 (19 pts), loops 2 (7 pts), match/switch 3 (7 pts), boolean chains 1 (nesting depth added 17). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
FlashJinaCodeBertModel::forward (cognitive 34) backends/candle/src/models/flash_jina_code.rs:345— FlashJinaCodeBertModel::forward has cognitive complexity 34 (threshold 15). Drivers by points: if/else 11 (19 pts), loops 2 (7 pts), match/switch 3 (7 pts), boolean chains 1 (nesting depth added 17). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
BackendProcess::new (cognitive 21) backends/python/src/management.rs:19— BackendProcess::new has cognitive complexity 21 (threshold 15). Drivers by points: if/else 7 (12 pts), loops 3 (7 pts), match/switch 2 (nesting depth added 9). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
TooManyFields: ModernBertConfig backends/candle/src/models/modernbert.rs:13— TooManyFields — 35 stored fields. The bar is 30 stored fields; this is 5 over it, 1.17× the bar. This is width in DATA, not behaviour: every reader that takes the whole type couples to all of its fields, so a change to any one of them is a change every reader has to be checked against. To reduce it, group the fields that are read together by the same callers into a smaller type of their own, and have this one hold that type as a single member — each reader then names only the group it uses.
D4 · Code Duplication· Near-duplicate member family (10 members, 56 shared lines) · ×1
Near-duplicate member family (10 members, 56 shared lines) backends/candle/src/models/flash_bert.rs:369— backends/candle/src/models/flash_bert.rs:369-526 | backends/candle/src/models/flash_distilbert.rs:243-396 | backends/candle/src/models/flash_gte.rs:276-428 | backends/candle/src/models/flash_jina.rs:299-432 | backends/candle/src/models/flash_jina_code.rs:345-478 | backends/candle/src/models/flash_mistral.rs:303-448 | backends/candle/src/models/flash_modernbert.rs:327-453 | backends/candle/src/models/flash_nomic.rs:272-427 | backends/candle/src/models/flash_qwen2.rs:324-469 | backends/candle/src/models/flash_qwen3.rs:394-546 — These 10 members are variants of one another: a block of 56 lines reported below appears in every one of them, and the pairwise near-duplicate rows they would otherwise produce are collapsed into this row. Read them as one construct written 10 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 10 times.
D4 · Code Duplication· Near-duplicate member family (9 members, 36 shared lines) · ×1
Near-duplicate member family (9 members, 36 shared lines) backends/candle/src/models/bert.rs:170— backends/candle/src/models/bert.rs:170-263 | backends/candle/src/models/distilbert.rs:113-203 | backends/candle/src/models/gte.rs:85-178 | backends/candle/src/models/jina.rs:143-236 | backends/candle/src/models/jina_code.rs:91-207 | backends/candle/src/models/modernbert.rs:213-298 | backends/candle/src/models/mpnet.rs:138-224 | backends/candle/src/models/nomic.rs:507-600 | backends/candle/src/models/qwen3.rs:143-271 — These 9 members are variants of one another: a block of 36 lines reported below appears in every one of them, and the pairwise near-duplicate rows they would otherwise produce are collapsed into this row. Read them as one construct written 9 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 9 times.
D4 · Code Duplication· Near-duplicate member family (3 members, 21 shared lines) · ×1
Near-duplicate member family (3 members, 21 shared lines) backends/candle/src/models/flash_qwen3.rs:116— backends/candle/src/models/flash_qwen3.rs:116-173 | backends/candle/src/models/gemma3.rs:305-467 | backends/candle/src/models/qwen3.rs:143-271 — These 3 members are variants of one another: a block of 21 lines reported below appears in every one of them, and the pairwise near-duplicate rows they would otherwise produce are collapsed into this row. Read them as one construct written 3 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 3 times.
D4 · Code Duplication· Members sharing a duplicated core (12 members, 50+ identical tokens) · ×1
Members sharing a duplicated core (12 members, 50+ identical tokens) backends/candle/src/models/bert.rs:706— backends/candle/src/models/bert.rs:706-932 | backends/candle/src/models/debertav2.rs:1121-1311 | backends/candle/src/models/distilbert.rs:508-725 | backends/candle/src/models/gemma3.rs:721-917 | backends/candle/src/models/gte.rs:489-709 | backends/candle/src/models/jina.rs:471-721 | backends/candle/src/models/jina_code.rs:398-648 | backends/candle/src/models/modernbert.rs:633-831 | backends/candle/src/models/mpnet.rs:506-681 | backends/candle/src/models/nomic.rs:773-1003 | backends/candle/src/models/qwen3.rs:518-753 | backends/ort/src/lib.rs:178-341 — These 12 members share a duplicated core: a run of at least 50 identical tokens appears in every one of them. That run is NOT broken out as duplicated-block rows below — it is what admitted this row, and the blocks below cover only the part of it that clears the block floor, so they understate the correspondence. Read the members as one construct written 12 times. The repair is at the members' grain — factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Extracting the individual blocks below is not the same fix: it leaves every body in place and the next edit still has to be made 12 times.
D4 · Code Duplication· Edited copy of a member (77 corresponding lines) · ×1
Edited copy of a member (77 corresponding lines) backends/candle/src/models/flash_qwen3.rs:31— backends/candle/src/models/flash_qwen3.rs:31-107 | backends/candle/src/models/qwen3.rs:57-135 — These two members are one piece of code written twice and then edited apart: 77 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (72 corresponding lines) · ×1
Edited copy of a member (72 corresponding lines) backends/candle/src/models/flash_modernbert.rs:251— backends/candle/src/models/flash_modernbert.rs:251-325 | backends/candle/src/models/modernbert.rs:508-579 — These two members are one piece of code written twice and then edited apart: 72 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (75 corresponding lines) · ×1
Edited copy of a member (75 corresponding lines) router/src/grpc/server.rs:391— router/src/grpc/server.rs:391-471 | router/src/grpc/server.rs:484-558 — These two members are one piece of code written twice and then edited apart: 75 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (63 corresponding lines) · ×1
Edited copy of a member (63 corresponding lines) backends/candle/src/models/flash_nomic.rs:207— backends/candle/src/models/flash_nomic.rs:207-270 | backends/candle/src/models/nomic.rs:709-771 — These two members are one piece of code written twice and then edited apart: 63 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (54 corresponding lines) · ×1
Edited copy of a member (54 corresponding lines) backends/candle/src/models/distilbert.rs:448— backends/candle/src/models/distilbert.rs:448-506 | backends/candle/src/models/flash_distilbert.rs:188-241 — These two members are one piece of code written twice and then edited apart: 54 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (39 corresponding lines) · ×1
Edited copy of a member (39 corresponding lines) backends/python/server/text_embeddings_server/models/flash_bert.py:355— backends/python/server/text_embeddings_server/models/flash_bert.py:355-394 | backends/python/server/text_embeddings_server/models/flash_mistral.py:413-451 — These two members are one piece of code written twice and then edited apart: 39 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (42 corresponding lines) · ×1
Edited copy of a member (42 corresponding lines) core/src/infer.rs:125— core/src/infer.rs:125-166 | core/src/infer.rs:176-221 — These two members are one piece of code written twice and then edited apart: 42 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copies of one member (4 members, 38+ corresponding lines) · ×1
Edited copies of one member (4 members, 38+ corresponding lines) backends/candle/src/models/flash_gte.rs:60— backends/candle/src/models/flash_gte.rs:60-97 | backends/candle/src/models/flash_mistral.rs:73-114 | backends/candle/src/models/flash_modernbert.rs:74-117 | backends/candle/src/models/flash_qwen2.rs:84-125 — These 4 members are one piece of code written 4 times and then edited apart: every two of them correspond over at least 38 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 4 times and the copies will drift further apart.
D4 · Code Duplication· Edited copy of a member (36 corresponding lines) · ×1
Edited copy of a member (36 corresponding lines) backends/candle/src/models/flash_bert.rs:80— backends/candle/src/models/flash_bert.rs:80-115 | backends/candle/src/models/flash_jina.rs:82-117 — These two members are one piece of code written twice and then edited apart: 36 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (28 corresponding lines) · ×1
Edited copy of a member (28 corresponding lines) backends/python/src/lib.rs:77— backends/python/src/lib.rs:77-104 | backends/python/src/lib.rs:106-133 — These two members are one piece of code written twice and then edited apart: 28 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copies of one member (6 members, 28+ corresponding lines) · ×1
Edited copies of one member (6 members, 28+ corresponding lines) backends/candle/src/models/flash_gte.rs:24— backends/candle/src/models/flash_gte.rs:24-51 | backends/candle/src/models/flash_modernbert.rs:29-65 | backends/candle/src/models/flash_nomic.rs:23-58 | backends/candle/src/models/gte.rs:50-77 | backends/candle/src/models/modernbert.rs:174-206 | backends/candle/src/models/nomic.rs:464-499 — These 6 members are one piece of code written 6 times and then edited apart: every two of them correspond over at least 28 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 6 times and the copies will drift further apart.
D4 · Code Duplication· Edited copies of one member (3 members, 42+ corresponding lines) · ×1
Edited copies of one member (3 members, 42+ corresponding lines) router/src/grpc/server.rs:80— router/src/grpc/server.rs:80-121 | router/src/grpc/server.rs:138-187 | router/src/grpc/server.rs:204-248 — These 3 members are one piece of code written 3 times and then edited apart: every two of them correspond over at least 42 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 3 times and the copies will drift further apart.
D4 · Code Duplication· Edited copy of a member (24 corresponding lines) · ×1
Edited copy of a member (24 corresponding lines) backends/python/server/text_embeddings_server/models/flash_mistral.py:195— backends/python/server/text_embeddings_server/models/flash_mistral.py:195-218 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:180-207 — These two members are one piece of code written twice and then edited apart: 24 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copies of one member (3 members, 31+ corresponding lines) · ×1
Edited copies of one member (3 members, 31+ corresponding lines) backends/candle/src/models/bert.rs:476— backends/candle/src/models/bert.rs:476-517 | backends/candle/src/models/bert.rs:519-549 | backends/candle/src/models/distilbert.rs:383-420 — These 3 members are one piece of code written 3 times and then edited apart: every two of them correspond over at least 31 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 3 times and the copies will drift further apart.
D4 · Code Duplication· Edited copy of a member (33 corresponding lines) · ×1
Edited copy of a member (33 corresponding lines) backends/candle/src/models/flash_distilbert.rs:63— backends/candle/src/models/flash_distilbert.rs:63-95 | backends/candle/src/models/flash_nomic.rs:67-100 — These two members are one piece of code written twice and then edited apart: 33 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copies of one member (3 members, 29+ corresponding lines) · ×1
Edited copies of one member (3 members, 29+ corresponding lines) backends/python/server/text_embeddings_server/models/classification_model.py:24— backends/python/server/text_embeddings_server/models/classification_model.py:24-52 | backends/python/server/text_embeddings_server/models/default_model.py:25-53 | backends/python/server/text_embeddings_server/models/masked_model.py:25-54 — These 3 members are one piece of code written 3 times and then edited apart: every two of them correspond over at least 29 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 3 times and the copies will drift further apart.
D4 · Code Duplication· Edited copy of a member (23 corresponding lines) · ×1
Edited copy of a member (23 corresponding lines) backends/candle/src/models/distilbert.rs:215— backends/candle/src/models/distilbert.rs:215-237 | backends/candle/src/models/nomic.rs:164-196 — These two members are one piece of code written twice and then edited apart: 23 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copies of one member (4 members, 18+ corresponding lines) · ×1
Edited copies of one member (4 members, 18+ corresponding lines) backends/candle/src/models/flash_jina.rs:171— backends/candle/src/models/flash_jina.rs:171-188 | backends/candle/src/models/flash_jina_code.rs:219-241 | backends/candle/src/models/jina.rs:289-306 | backends/candle/src/models/jina_code.rs:270-295 — These 4 members are one piece of code written 4 times and then edited apart: every two of them correspond over at least 18 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 4 times and the copies will drift further apart.
D4 · Code Duplication· Edited copy of a member (21 corresponding lines) · ×1
Edited copy of a member (21 corresponding lines) backends/candle/src/models/flash_modernbert.rs:168— backends/candle/src/models/flash_modernbert.rs:168-188 | backends/candle/src/models/modernbert.rs:347-371 — These two members are one piece of code written twice and then edited apart: 21 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copies of one member (4 members, 20+ corresponding lines) · ×1
Edited copies of one member (4 members, 20+ corresponding lines) router/src/grpc/server.rs:594— router/src/grpc/server.rs:594-613 | router/src/grpc/server.rs:634-653 | router/src/grpc/server.rs:674-693 | router/src/grpc/server.rs:718-746 — These 4 members are one piece of code written 4 times and then edited apart: every two of them correspond over at least 20 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 4 times and the copies will drift further apart.
D4 · Code Duplication· Edited copy of a member (18 corresponding lines) · ×1
Edited copy of a member (18 corresponding lines) backends/candle/src/models/debertav2.rs:660— backends/candle/src/models/debertav2.rs:660-677 | backends/candle/src/models/debertav2.rs:721-738 — These two members are one piece of code written twice and then edited apart: 18 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copies of one member (3 members, 16+ corresponding lines) · ×1
Edited copies of one member (3 members, 16+ corresponding lines) backends/candle/src/models/distilbert.rs:33— backends/candle/src/models/distilbert.rs:33-48 | backends/candle/src/models/mpnet.rs:34-55 | backends/candle/src/models/nomic.rs:67-82 — These 3 members are one piece of code written 3 times and then edited apart: every two of them correspond over at least 16 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 3 times and the copies will drift further apart.
D4 · Code Duplication· Edited copies of one member (4 members, 15+ corresponding lines) · ×1
Edited copies of one member (4 members, 15+ corresponding lines) backends/python/server/text_embeddings_server/models/flash_bert.py:329— backends/python/server/text_embeddings_server/models/flash_bert.py:329-343 | backends/python/server/text_embeddings_server/models/flash_mistral.py:385-401 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:402-419 | backends/python/server/text_embeddings_server/models/jinaBert_model.py:524-538 — These 4 members are one piece of code written 4 times and then edited apart: every two of them correspond over at least 15 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 4 times and the copies will drift further apart.
D4 · Code Duplication· Edited copy of a member (27 corresponding lines) · ×1
Edited copy of a member (27 corresponding lines) backends/python/server/text_embeddings_server/models/flash_mistral.py:325— backends/python/server/text_embeddings_server/models/flash_mistral.py:325-351 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:345-371 — These two members are one piece of code written twice and then edited apart: 27 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (20 corresponding lines) · ×1
Edited copy of a member (20 corresponding lines) backends/python/server/text_embeddings_server/models/flash_mistral.py:273— backends/python/server/text_embeddings_server/models/flash_mistral.py:273-292 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:262-281 — These two members are one piece of code written twice and then edited apart: 20 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (10 corresponding lines) · ×1
Edited copy of a member (10 corresponding lines) backends/candle/src/models/flash_jina.rs:197— backends/candle/src/models/flash_jina.rs:197-206 | backends/candle/src/models/flash_jina_code.rs:250-259 — These two members are one piece of code written twice and then edited apart: 10 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copies of one member (3 members, 15+ corresponding lines) · ×1
Edited copies of one member (3 members, 15+ corresponding lines) backends/candle/src/models/flash_gte.rs:135— backends/candle/src/models/flash_gte.rs:135-149 | backends/candle/src/models/gte.rs:264-279 | backends/candle/src/models/qwen3.rs:366-385 — These 3 members are one piece of code written 3 times and then edited apart: every two of them correspond over at least 15 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 3 times and the copies will drift further apart.
D4 · Code Duplication· Edited copies of one member (3 members, 26+ corresponding lines) · ×1
Edited copies of one member (3 members, 26+ corresponding lines) core/src/tokenization.rs:74— core/src/tokenization.rs:74-101 | core/src/tokenization.rs:109-135 | core/src/tokenization.rs:142-167 — These 3 members are one piece of code written 3 times and then edited apart: every two of them correspond over at least 26 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 3 times and the copies will drift further apart.
D4 · Code Duplication· Edited copies of one member (4 members, 13+ corresponding lines) · ×1
Edited copies of one member (4 members, 13+ corresponding lines) backends/candle/src/models/distilbert.rs:279— backends/candle/src/models/distilbert.rs:279-291 | backends/candle/src/models/flash_distilbert.rs:131-143 | backends/candle/src/models/flash_nomic.rs:139-154 | backends/candle/src/models/nomic.rs:638-653 — These 4 members are one piece of code written 4 times and then edited apart: every two of them correspond over at least 13 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 4 times and the copies will drift further apart.
D4 · Code Duplication· Edited copies of one member (4 members, 11+ corresponding lines) · ×1
Edited copies of one member (4 members, 11+ corresponding lines) backends/candle/src/models/flash_qwen3.rs:215— backends/candle/src/models/flash_qwen3.rs:215-225 | backends/candle/src/models/gemma3.rs:507-519 | backends/candle/src/models/gte.rs:216-227 | backends/candle/src/models/qwen3.rs:313-323 — These 4 members are one piece of code written 4 times and then edited apart: every two of them correspond over at least 11 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 4 times and the copies will drift further apart.
D4 · Code Duplication· Edited copies of one member (4 members, 10+ corresponding lines) · ×1
Edited copies of one member (4 members, 10+ corresponding lines) backends/candle/src/models/bert.rs:455— backends/candle/src/models/bert.rs:455-464 | backends/candle/src/models/distilbert.rs:360-371 | backends/candle/src/models/gte.rs:322-336 | backends/candle/src/models/modernbert.rs:472-484 — These 4 members are one piece of code written 4 times and then edited apart: every two of them correspond over at least 10 lines, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below: the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain. Factor the shared implementation out once and have all of them call it with their differences as parameters or as an injected step, or, where the difference is systematic, generate them from one template. Left alone, the next edit has to be made 4 times and the copies will drift further apart.
D4 · Code Duplication· Edited copy of a member (9 corresponding lines) · ×1
Edited copy of a member (9 corresponding lines) backends/candle/src/models/debertav2.rs:621— backends/candle/src/models/debertav2.rs:621-629 | backends/candle/src/models/nomic.rs:402-411 — These two members are one piece of code written twice and then edited apart: 9 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (4 corresponding lines) · ×1
Edited copy of a member (4 corresponding lines) backends/python/server/text_embeddings_server/models/flash_mistral.py:23— backends/python/server/text_embeddings_server/models/flash_mistral.py:23-26 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:34-37 — These two members are one piece of code written twice and then edited apart: 4 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
D4 · Code Duplication· Edited copy of a member (6 corresponding lines) · ×1
Edited copy of a member (6 corresponding lines) core/src/tokenization.rs:439— core/src/tokenization.rs:439-445 | router/src/http/types.rs:17-22 — These two members are one piece of code written twice and then edited apart: 6 consecutive lines correspond almost exactly, broken only by small local edits. Most of that correspondence is NOT reported as duplicated blocks below — the edits cut it into fragments and only the largest of them clear the block floor, so the rows below understate it. The repair is at the members' grain — factor the shared implementation into one the two call with their differences as parameters or as an injected step, or, where the difference is systematic (an extra return value, one transport against another), generate one from the other. Left alone, the next edit has to be made twice and the two will drift further apart.
Duplicated block (103 lines × 2) backends/candle/src/models/jina.rs:532— backends/candle/src/models/jina.rs:532-634 | backends/candle/src/models/jina_code.rs:459-561 — before extracting anything, compare `backends/candle/src/models/jina.rs` and `backends/candle/src/models/jina_code.rs` as WHOLE FILES: this scan already matched 29 separate duplicated blocks between them, totalling at least 536 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/jina.rs:532` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (48 lines × 2) router/src/grpc/server.rs:424— router/src/grpc/server.rs:424-471 | router/src/grpc/server.rs:511-558 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:424` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (45 lines × 2) backends/candle/src/models/flash_qwen3.rs:313— backends/candle/src/models/flash_qwen3.rs:313-357 | backends/candle/src/models/qwen3.rs:410-454 — before extracting anything, compare `backends/candle/src/models/flash_qwen3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 225 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/qwen3.rs:458` calls `unwrap_or` and `backends/candle/src/models/flash_qwen3.rs:360` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (52–56 lines × 2) backends/src/lib.rs:141— backends/src/lib.rs:141-192 | backends/src/lib.rs:201-256 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (42–43 lines × 6) backends/candle/src/models/bert.rs:738— backends/candle/src/models/bert.rs:738-780 | backends/candle/src/models/distilbert.rs:538-579 | backends/candle/src/models/gte.rs:521-563 | backends/candle/src/models/jina.rs:503-545 | backends/candle/src/models/jina_code.rs:430-472 | backends/candle/src/models/nomic.rs:805-847 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:738` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/bert.rs:736` calls `push` and `backends/candle/src/models/distilbert.rs:536` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (35–42 lines × 9) backends/candle/src/models/flash_bert.rs:408— backends/candle/src/models/flash_bert.rs:408-449 | backends/candle/src/models/flash_gte.rs:334-375 | backends/candle/src/models/flash_jina.rs:338-379 | backends/candle/src/models/flash_jina_code.rs:384-425 | backends/candle/src/models/flash_mistral.rs:354-395 | backends/candle/src/models/flash_modernbert.rs:371-405 | backends/candle/src/models/flash_nomic.rs:333-374 | backends/candle/src/models/flash_qwen2.rs:375-416 | backends/candle/src/models/flash_qwen3.rs:452-493 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_gte.rs` as WHOLE FILES: this scan already matched 9 separate duplicated blocks between them, totalling at least 154 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:408` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (40–41 lines × 2) backends/candle/src/models/flash_qwen3.rs:49— backends/candle/src/models/flash_qwen3.rs:49-88 | backends/candle/src/models/qwen3.rs:77-117 — before extracting anything, compare `backends/candle/src/models/flash_qwen3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 225 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_qwen3.rs:49` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (38 lines × 3) backends/candle/src/models/bert.rs:278— backends/candle/src/models/bert.rs:278-315 | backends/candle/src/models/flash_bert.rs:130-167 | backends/candle/src/models/mpnet.rs:239-276 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/flash_bert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:278` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&BertConfig` in one and `&MPNetConfig` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (35 lines × 3) backends/candle/src/models/bert.rs:782— backends/candle/src/models/bert.rs:782-816 | backends/candle/src/models/gte.rs:565-599 | backends/candle/src/models/nomic.rs:849-883 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/gte.rs` as WHOLE FILES: this scan already matched 25 separate duplicated blocks between them, totalling at least 400 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:782` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (33–34 lines × 9) backends/candle/src/models/bert.rs:893— backends/candle/src/models/bert.rs:893-926 | backends/candle/src/models/debertav2.rs:1272-1305 | backends/candle/src/models/distilbert.rs:686-719 | backends/candle/src/models/gte.rs:670-703 | backends/candle/src/models/jina.rs:682-715 | backends/candle/src/models/jina_code.rs:609-642 | backends/candle/src/models/modernbert.rs:793-825 | backends/candle/src/models/mpnet.rs:642-675 | backends/candle/src/models/nomic.rs:964-997 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:893` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (32–33 lines × 2) backends/candle/src/models/flash_qwen3.rs:119— backends/candle/src/models/flash_qwen3.rs:119-151 | backends/candle/src/models/qwen3.rs:148-179 — before extracting anything, compare `backends/candle/src/models/flash_qwen3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 225 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/qwen3.rs:146` calls `device` and `backends/candle/src/models/flash_qwen3.rs:117` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (27–32 lines × 7) backends/candle/src/models/bert.rs:199— backends/candle/src/models/bert.rs:199-230 | backends/candle/src/models/distilbert.rs:140-171 | backends/candle/src/models/gte.rs:115-146 | backends/candle/src/models/jina.rs:172-203 | backends/candle/src/models/jina_code.rs:141-172 | backends/candle/src/models/nomic.rs:537-568 | backends/candle/src/models/qwen3.rs:216-242 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:199` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (32 lines × 3) backends/candle/src/models/flash_mistral.rs:306— backends/candle/src/models/flash_mistral.rs:306-337 | backends/candle/src/models/flash_qwen2.rs:327-358 | backends/candle/src/models/flash_qwen3.rs:397-428 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (27–29 lines × 9) backends/candle/src/models/bert.rs:217— backends/candle/src/models/bert.rs:217-245 | backends/candle/src/models/distilbert.rs:158-186 | backends/candle/src/models/gte.rs:133-161 | backends/candle/src/models/jina.rs:190-218 | backends/candle/src/models/jina_code.rs:159-187 | backends/candle/src/models/modernbert.rs:260-286 | backends/candle/src/models/mpnet.rs:178-206 | backends/candle/src/models/nomic.rs:555-583 | backends/candle/src/models/qwen3.rs:230-256 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:217` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (27–28 lines × 2) backends/candle/src/layers/rms_norm.rs:27— backends/candle/src/layers/rms_norm.rs:27-53 | backends/candle/src/models/gemma3.rs:60-87 — before extracting anything, compare `backends/candle/src/layers/rms_norm.rs` and `backends/candle/src/models/gemma3.rs` as WHOLE FILES: this scan already matched 4 separate duplicated blocks between them, totalling at least 63 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. The two sit in different directories, so one cannot simply be deleted in favour of the other while both are reached separately: hoist the shared part into a location both already depend on and have each file call it, and retire whichever file turns out to have no caller of its own left. Extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/layers/rms_norm.rs:27` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (21–27 lines × 10) backends/candle/src/models/flash_bert.rs:397— backends/candle/src/models/flash_bert.rs:397-423 | backends/candle/src/models/flash_distilbert.rs:270-294 | backends/candle/src/models/flash_gte.rs:323-349 | backends/candle/src/models/flash_jina.rs:327-353 | backends/candle/src/models/flash_jina_code.rs:373-399 | backends/candle/src/models/flash_mistral.rs:343-369 | backends/candle/src/models/flash_modernbert.rs:362-382 | backends/candle/src/models/flash_nomic.rs:322-348 | backends/candle/src/models/flash_qwen2.rs:364-390 | backends/candle/src/models/flash_qwen3.rs:441-467 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:397` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (26–27 lines × 2) backends/candle/src/models/flash_gte.rs:205— backends/candle/src/models/flash_gte.rs:205-231 | backends/candle/src/models/flash_mistral.rs:272-297 — before extracting anything, compare `backends/candle/src/models/flash_gte.rs` and `backends/candle/src/models/flash_mistral.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 163 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_gte.rs:205` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `>EConfig` in one and `&MistralConfig` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (34 lines × 2) backends/candle/src/models/distilbert.rs:465— backends/candle/src/models/distilbert.rs:465-498 | backends/candle/src/models/flash_distilbert.rs:203-236 — before extracting anything, compare `backends/candle/src/models/distilbert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 6 separate duplicated blocks between them, totalling at least 97 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/distilbert.rs:465` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (25 lines × 2) backends/candle/src/layers/layer_norm.rs:23— backends/candle/src/layers/layer_norm.rs:23-47 | backends/candle/src/layers/layer_norm.rs:105-129 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/layers/layer_norm.rs:23` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (20–23 lines × 10) backends/candle/src/models/flash_bert.rs:492— backends/candle/src/models/flash_bert.rs:492-514 | backends/candle/src/models/flash_distilbert.rs:362-384 | backends/candle/src/models/flash_gte.rs:394-416 | backends/candle/src/models/flash_jina.rs:398-420 | backends/candle/src/models/flash_jina_code.rs:444-466 | backends/candle/src/models/flash_mistral.rs:414-436 | backends/candle/src/models/flash_modernbert.rs:423-442 | backends/candle/src/models/flash_nomic.rs:393-415 | backends/candle/src/models/flash_qwen2.rs:435-457 | backends/candle/src/models/flash_qwen3.rs:512-534 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:492` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/flash_distilbert.rs:361` calls `Some` and `backends/candle/src/models/flash_bert.rs:491` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (22–23 lines × 2) backends/candle/src/models/nomic.rs:121— backends/candle/src/models/nomic.rs:121-143 | backends/candle/src/models/nomic.rs:172-193 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/nomic.rs:121` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (22 lines × 4) router/src/http/server.rs:623— router/src/http/server.rs:623-644 | router/src/http/server.rs:810-831 | router/src/http/server.rs:988-1009 | router/src/http/server.rs:1199-1220 — all 4 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:623` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (21 lines × 6) router/src/http/server.rs:639— router/src/http/server.rs:639-659 | router/src/http/server.rs:826-846 | router/src/http/server.rs:1004-1024 | router/src/http/server.rs:1215-1235 | router/src/http/server.rs:1391-1411 | router/src/http/server.rs:1471-1491 — all 6 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:639` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (21 lines × 3) backends/candle/src/models/flash_qwen3.rs:127— backends/candle/src/models/flash_qwen3.rs:127-147 | backends/candle/src/models/gemma3.rs:323-343 | backends/candle/src/models/qwen3.rs:156-176 — before extracting anything, compare `backends/candle/src/models/flash_qwen3.rs` and `backends/candle/src/models/gemma3.rs` as WHOLE FILES: this scan already matched 3 separate duplicated blocks between them, totalling at least 37 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (18–20 lines × 6) backends/candle/src/models/bert.rs:119— backends/candle/src/models/bert.rs:119-138 | backends/candle/src/models/flash_bert.rs:26-43 | backends/candle/src/models/flash_jina.rs:27-44 | backends/candle/src/models/flash_jina_code.rs:30-47 | backends/candle/src/models/jina.rs:92-111 | backends/candle/src/models/jina_code.rs:27-46 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/flash_bert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (20 lines × 3) backends/candle/src/models/bert.rs:611— backends/candle/src/models/bert.rs:611-630 | backends/candle/src/models/bert.rs:685-704 | backends/candle/src/models/jina.rs:450-469 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 37 separate duplicated blocks between them, totalling at least 547 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:611` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (2–28 lines × 10) backends/candle/src/models/bert.rs:139— backends/candle/src/models/bert.rs:139-155 | backends/candle/src/models/distilbert.rs:92-94 | backends/candle/src/models/flash_bert.rs:44-60 | backends/candle/src/models/flash_distilbert.rs:37-39 | backends/candle/src/models/flash_jina.rs:45-61 | backends/candle/src/models/flash_jina_code.rs:48-75 | backends/candle/src/models/flash_qwen2.rs:56-57 | backends/candle/src/models/jina.rs:112-128 | backends/candle/src/models/jina_code.rs:47-74 | backends/candle/src/models/mpnet.rs:105-107 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (19 lines × 7) backends/candle/src/models/bert.rs:842— backends/candle/src/models/bert.rs:842-860 | backends/candle/src/models/distilbert.rs:636-654 | backends/candle/src/models/gte.rs:638-656 | backends/candle/src/models/jina.rs:650-668 | backends/candle/src/models/jina_code.rs:577-595 | backends/candle/src/models/mpnet.rs:609-627 | backends/candle/src/models/nomic.rs:932-950 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:842` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/bert.rs:861` calls `index_select` and `backends/candle/src/models/distilbert.rs:655` does not — after which the two agree again for 4 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (19 lines × 4) backends/candle/src/models/flash_mistral.rs:183— backends/candle/src/models/flash_mistral.rs:183-201 | backends/candle/src/models/flash_qwen2.rs:194-212 | backends/candle/src/models/flash_qwen3.rs:242-260 | backends/candle/src/models/qwen3.rs:340-358 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_mistral.rs:183` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&MistralConfig` in one and `&Qwen2Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (17–19 lines × 4) backends/candle/src/models/gemma3.rs:722— backends/candle/src/models/gemma3.rs:722-738 | backends/candle/src/models/modernbert.rs:634-651 | backends/candle/src/models/mpnet.rs:507-525 | backends/candle/src/models/qwen3.rs:519-536 — before extracting anything, compare `backends/candle/src/models/gemma3.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 4 separate duplicated blocks between them, totalling at least 61 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/gemma3.rs:722` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (18–19 lines × 2) router/src/grpc/server.rs:342— router/src/grpc/server.rs:342-360 | router/src/http/server.rs:1358-1375 — `router/src/grpc/server.rs` and `router/src/http/server.rs` are one unit implemented once per sibling directory, so they are most likely parallel implementations of one contract rather than a copy of each other — this scan matched 7 separate duplicated blocks between them, totalling at least 91 lines. If both are selected at run time, neither can be retired in favour of the other, and the lines that DIFFER between them are the reason both exist. The move that pays here is to hoist the identical part into a shared location the whole family can reach and give what differs a parameter or a seam, so a change lands once instead of once per sibling; extracting one helper per block leaves every sibling to drift on its own. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:342` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (18 lines × 3) router/src/http/server.rs:661— router/src/http/server.rs:661-678 | router/src/http/server.rs:848-865 | router/src/http/server.rs:1026-1043 — all 3 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:661` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (29 lines × 2) backends/candle/src/models/flash_jina.rs:136— backends/candle/src/models/flash_jina.rs:136-164 | backends/candle/src/models/jina.rs:255-283 — before extracting anything, compare `backends/candle/src/models/flash_jina.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 120 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_jina.rs:136` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (16–18 lines × 2) backends/candle/src/models/flash_qwen3.rs:32— backends/candle/src/models/flash_qwen3.rs:32-47 | backends/candle/src/models/qwen3.rs:58-75 — before extracting anything, compare `backends/candle/src/models/flash_qwen3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 225 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_qwen3.rs:32` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (17–18 lines × 2) backends/candle/src/models/gemma3.rs:775— backends/candle/src/models/gemma3.rs:775-792 | backends/candle/src/models/qwen3.rs:571-587 — before extracting anything, compare `backends/candle/src/models/gemma3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 15 separate duplicated blocks between them, totalling at least 209 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/gemma3.rs:775` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (14–17 lines × 8) backends/candle/src/models/bert.rs:834— backends/candle/src/models/bert.rs:834-850 | backends/candle/src/models/distilbert.rs:628-644 | backends/candle/src/models/gte.rs:630-646 | backends/candle/src/models/jina.rs:642-658 | backends/candle/src/models/jina_code.rs:569-585 | backends/candle/src/models/modernbert.rs:760-773 | backends/candle/src/models/mpnet.rs:601-617 | backends/candle/src/models/nomic.rs:924-940 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:834` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (17 lines × 3) backends/candle/src/models/debertav2.rs:1164— backends/candle/src/models/debertav2.rs:1164-1180 | backends/candle/src/models/modernbert.rs:668-684 | backends/candle/src/models/mpnet.rs:546-562 — before extracting anything, compare `backends/candle/src/models/debertav2.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 7 separate duplicated blocks between them, totalling at least 122 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/debertav2.rs:1164` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/debertav2.rs:1162` calls `push` and `backends/candle/src/models/mpnet.rs:544` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (16–17 lines × 2) backends/candle/src/models/gemma3.rs:679— backends/candle/src/models/gemma3.rs:679-694 | backends/candle/src/models/qwen3.rs:456-472 — before extracting anything, compare `backends/candle/src/models/gemma3.rs` and `backends/candle/src/models/qwen3.rs` as WHOLE FILES: this scan already matched 15 separate duplicated blocks between them, totalling at least 209 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&Gemma3Config` in one and `&Qwen3Config` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (11–16 lines × 11) backends/candle/src/models/bert.rs:715— backends/candle/src/models/bert.rs:715-730 | backends/candle/src/models/debertav2.rs:1130-1144 | backends/candle/src/models/distilbert.rs:517-531 | backends/candle/src/models/gemma3.rs:730-740 | backends/candle/src/models/gte.rs:498-513 | backends/candle/src/models/jina.rs:480-495 | backends/candle/src/models/jina_code.rs:407-422 | backends/candle/src/models/modernbert.rs:642-653 | backends/candle/src/models/mpnet.rs:516-527 | backends/candle/src/models/nomic.rs:782-797 | backends/candle/src/models/qwen3.rs:528-538 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:715` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (14–16 lines × 9) backends/candle/src/models/bert.rs:829— backends/candle/src/models/bert.rs:829-844 | backends/candle/src/models/debertav2.rs:1225-1240 | backends/candle/src/models/distilbert.rs:623-638 | backends/candle/src/models/gte.rs:625-640 | backends/candle/src/models/jina.rs:637-652 | backends/candle/src/models/jina_code.rs:564-579 | backends/candle/src/models/modernbert.rs:755-768 | backends/candle/src/models/mpnet.rs:596-611 | backends/candle/src/models/nomic.rs:919-934 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:829` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/gte.rs:623` calls `forward`, `as_ref` and `backends/candle/src/models/bert.rs:827` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (15–16 lines × 5) router/src/http/server.rs:209— router/src/http/server.rs:209-223 | router/src/http/server.rs:656-671 | router/src/http/server.rs:843-858 | router/src/http/server.rs:1021-1036 | router/src/http/server.rs:1232-1247 — all 5 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:209` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `router/src/http/server.rs:672` calls `clone` and `router/src/http/server.rs:1249` does not — after which the two agree again for 4 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (16 lines × 3) backends/candle/src/models/bert.rs:153— backends/candle/src/models/bert.rs:153-168 | backends/candle/src/models/jina.rs:126-141 | backends/candle/src/models/mpnet.rs:116-131 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 37 separate duplicated blocks between them, totalling at least 547 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:153` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note first that the copies are not typed on the same thing: the declarations holding them bind `config` to `&BertConfig` in one and `&MPNetConfig` in another, and the duplicated lines use it. The extracted unit therefore needs a parameter type that fits BOTH — their common supertype where they have one, or a new abstraction over them where they do not — and settling that is the step that comes BEFORE the extraction above. Where the two types are deliberately unrelated, the duplication is the price of that separation and the honest resolution is to record the decision rather than to extract.
Duplicated block (13–16 lines × 3) backends/candle/src/models/flash_mistral.rs:212— backends/candle/src/models/flash_mistral.rs:212-224 | backends/candle/src/models/flash_qwen2.rs:223-235 | backends/candle/src/models/flash_qwen3.rs:271-286 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit.
Duplicated block (14–16 lines × 2) backends/candle/src/lib.rs:288— backends/candle/src/lib.rs:288-301 | backends/candle/src/lib.rs:400-415 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
Duplicated block (15 lines × 9) backends/candle/src/models/bert.rs:918— backends/candle/src/models/bert.rs:918-932 | backends/candle/src/models/distilbert.rs:711-725 | backends/candle/src/models/flash_nomic.rs:413-427 | backends/candle/src/models/gte.rs:695-709 | backends/candle/src/models/jina.rs:707-721 | backends/candle/src/models/jina_code.rs:634-648 | backends/candle/src/models/modernbert.rs:817-831 | backends/candle/src/models/mpnet.rs:667-681 | backends/candle/src/models/nomic.rs:989-1003 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:918` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (13–15 lines × 4) backends/candle/src/models/bert.rs:855— backends/candle/src/models/bert.rs:855-869 | backends/candle/src/models/debertav2.rs:1251-1265 | backends/candle/src/models/modernbert.rs:776-788 | backends/candle/src/models/mpnet.rs:622-636 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Each matched range is the entire body of the declaration above it, so the region is already a complete unit: move that whole declaration to the shared location and have each site call it, rather than lifting the lines out of their bodies. Any `return` inside it is the body's own exit and keeps its meaning in the moved unit. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/debertav2.rs:1266` calls `Some` and `backends/candle/src/models/modernbert.rs:789` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (14 lines × 9) router/src/http/server.rs:203— router/src/http/server.rs:203-216 | router/src/http/server.rs:379-392 | router/src/http/server.rs:517-530 | router/src/http/server.rs:650-663 | router/src/http/server.rs:837-850 | router/src/http/server.rs:1015-1028 | router/src/http/server.rs:1226-1239 | router/src/http/server.rs:1402-1415 | router/src/http/server.rs:1482-1495 — all 9 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `router/src/http/server.rs:646` calls `Err` and `router/src/http/server.rs:200` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (14 lines × 5) router/src/http/server.rs:258— router/src/http/server.rs:258-271 | router/src/http/server.rs:718-731 | router/src/http/server.rs:904-917 | router/src/http/server.rs:1081-1094 | router/src/http/server.rs:1298-1311 — all 5 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:258` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it.
Duplicated block (13 lines × 10) backends/candle/src/models/bert.rs:914— backends/candle/src/models/bert.rs:914-926 | backends/candle/src/models/debertav2.rs:1293-1305 | backends/candle/src/models/distilbert.rs:707-719 | backends/candle/src/models/flash_nomic.rs:409-421 | backends/candle/src/models/gte.rs:691-703 | backends/candle/src/models/jina.rs:703-715 | backends/candle/src/models/jina_code.rs:630-642 | backends/candle/src/models/modernbert.rs:813-825 | backends/candle/src/models/mpnet.rs:663-675 | backends/candle/src/models/nomic.rs:985-997 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:914` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (13 lines × 6) backends/candle/src/models/bert.rs:740— backends/candle/src/models/bert.rs:740-752 | backends/candle/src/models/debertav2.rs:1153-1165 | backends/candle/src/models/gte.rs:523-535 | backends/candle/src/models/jina.rs:505-517 | backends/candle/src/models/jina_code.rs:432-444 | backends/candle/src/models/nomic.rs:807-819 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:740` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (13 lines × 2) backends/candle/src/models/flash_modernbert.rs:261— backends/candle/src/models/flash_modernbert.rs:261-273 | backends/candle/src/models/modernbert.rs:509-521 — before extracting anything, compare `backends/candle/src/models/flash_modernbert.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 113 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_modernbert.rs:261` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11–13 lines × 2) backends/ort/src/lib.rs:407— backends/ort/src/lib.rs:407-419 | backends/ort/src/lib.rs:699-709 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/ort/src/lib.rs:712` calls `e`, `to_owned` and `backends/ort/src/lib.rs:421` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (12 lines × 7) backends/candle/src/models/bert.rs:740— backends/candle/src/models/bert.rs:740-751 | backends/candle/src/models/debertav2.rs:1153-1164 | backends/candle/src/models/distilbert.rs:540-551 | backends/candle/src/models/gte.rs:523-534 | backends/candle/src/models/jina.rs:505-516 | backends/candle/src/models/jina_code.rs:432-443 | backends/candle/src/models/nomic.rs:807-818 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:740` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/bert.rs:752` calls `push` and `backends/candle/src/models/distilbert.rs:552` does not — after which the two agree again for 4 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (14 lines × 4) backends/candle/src/models/bert.rs:595— backends/candle/src/models/bert.rs:595-608 | backends/candle/src/models/bert.rs:659-672 | backends/candle/src/models/flash_bert.rs:257-270 | backends/candle/src/models/flash_bert.rs:324-337 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/flash_bert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:595` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (3–12 lines × 3) backends/candle/src/models/bert.rs:811— backends/candle/src/models/bert.rs:811-822 | backends/candle/src/models/debertav2.rs:1198-1209 | backends/candle/src/models/mpnet.rs:582-584 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:811` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/mpnet.rs:582` calls `from_vec` and `backends/candle/src/models/bert.rs:811` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (11–12 lines × 2) backends/candle/src/models/flash_jina.rs:256— backends/candle/src/models/flash_jina.rs:256-267 | backends/candle/src/models/jina.rs:423-433 — before extracting anything, compare `backends/candle/src/models/flash_jina.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 120 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_jina.rs:256` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9–12 lines × 2) backends/candle/src/models/flash_modernbert.rs:169— backends/candle/src/models/flash_modernbert.rs:169-177 | backends/candle/src/models/modernbert.rs:348-359 — before extracting anything, compare `backends/candle/src/models/flash_modernbert.rs` and `backends/candle/src/models/modernbert.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 113 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (11 lines × 5) backends/candle/src/models/flash_bert.rs:373— backends/candle/src/models/flash_bert.rs:373-383 | backends/candle/src/models/flash_gte.rs:280-290 | backends/candle/src/models/flash_jina.rs:303-313 | backends/candle/src/models/flash_jina_code.rs:349-359 | backends/candle/src/models/flash_nomic.rs:276-286 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_gte.rs` as WHOLE FILES: this scan already matched 9 separate duplicated blocks between them, totalling at least 154 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/flash_gte.rs:292` calls `forward` and `backends/candle/src/models/flash_bert.rs:385` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (10–11 lines × 4) backends/candle/src/models/bert.rs:807— backends/candle/src/models/bert.rs:807-816 | backends/candle/src/models/distilbert.rs:604-614 | backends/candle/src/models/gte.rs:590-599 | backends/candle/src/models/nomic.rs:874-883 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:807` it runs out through the closing brace of the declaration holding it and carries on into the declaration that follows — the window is the tail of one member plus the head of the next, so no call can be substituted for those exact lines, and the smallest declaration that contains all of them is the type they sit in. The repeated unit is the member each site sits in: where those members' bodies are the same, move one whole member to the shared location and have the others delegate to it; where the copies are a run of near-identical overloads or wrappers that differ only in their signatures, the repetition IS the run — a one-line delegation has no helper inside it to lift — so generate the run from the set it enumerates, or accept it and keep each member's own documentation with it. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/gte.rs:599` calls `from_vec` and `backends/candle/src/models/distilbert.rs:614` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (8–10 lines × 10) backends/candle/src/models/bert.rs:189— backends/candle/src/models/bert.rs:189-198 | backends/candle/src/models/distilbert.rs:130-139 | backends/candle/src/models/gemma3.rs:404-412 | backends/candle/src/models/gte.rs:105-114 | backends/candle/src/models/jina.rs:162-171 | backends/candle/src/models/jina_code.rs:131-140 | backends/candle/src/models/modernbert.rs:243-250 | backends/candle/src/models/mpnet.rs:158-166 | backends/candle/src/models/nomic.rs:527-536 | backends/candle/src/models/qwen3.rs:207-215 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:189` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/gte.rs:102` calls `apply_rotary` and `backends/candle/src/models/bert.rs:187` does not — after which the two agree again for 4 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (10 lines × 10) router/src/grpc/server.rs:955— router/src/grpc/server.rs:955-964 | router/src/http/server.rs:207-216 | router/src/http/server.rs:383-392 | router/src/http/server.rs:521-530 | router/src/http/server.rs:654-663 | router/src/http/server.rs:841-850 | router/src/http/server.rs:1019-1028 | router/src/http/server.rs:1230-1239 | router/src/http/server.rs:1406-1415 | router/src/http/server.rs:1486-1495 — `router/src/grpc/server.rs` and `router/src/http/server.rs` are one unit implemented once per sibling directory, so they are most likely parallel implementations of one contract rather than a copy of each other — this scan matched 7 separate duplicated blocks between them, totalling at least 91 lines. If both are selected at run time, neither can be retired in favour of the other, and the lines that DIFFER between them are the reason both exist. The move that pays here is to hoist the identical part into a shared location the whole family can reach and give what differs a parameter or a seam, so a change lands once instead of once per sibling; extracting one helper per block leaves every sibling to drift on its own. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/grpc/server.rs:955` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9–10 lines × 4) backends/candle/src/models/bert.rs:171— backends/candle/src/models/bert.rs:171-180 | backends/candle/src/models/jina.rs:144-153 | backends/candle/src/models/jina_code.rs:92-100 | backends/candle/src/models/mpnet.rs:139-148 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/jina.rs` as WHOLE FILES: this scan already matched 37 separate duplicated blocks between them, totalling at least 547 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (7–10 lines × 2) backends/candle/src/models/flash_bert.rs:175— backends/candle/src/models/flash_bert.rs:175-181 | backends/candle/src/models/mpnet.rs:284-293 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/mpnet.rs` as WHOLE FILES: this scan already matched 6 separate duplicated blocks between them, totalling at least 99 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (8–9 lines × 12) backends/candle/src/models/flash_bert.rs:442— backends/candle/src/models/flash_bert.rs:442-450 | backends/candle/src/models/flash_bert.rs:469-477 | backends/candle/src/models/flash_distilbert.rs:311-319 | backends/candle/src/models/flash_distilbert.rs:338-346 | backends/candle/src/models/flash_gte.rs:368-376 | backends/candle/src/models/flash_jina.rs:372-380 | backends/candle/src/models/flash_jina_code.rs:418-426 | backends/candle/src/models/flash_mistral.rs:388-396 | backends/candle/src/models/flash_modernbert.rs:399-406 | backends/candle/src/models/flash_nomic.rs:367-375 | backends/candle/src/models/flash_qwen2.rs:409-417 | backends/candle/src/models/flash_qwen3.rs:486-494 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_bert.rs:442` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/flash_bert.rs:466` calls `as_ref`, `unwrap`, `forward` and `backends/candle/src/models/flash_bert.rs:441` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (9 lines × 7) router/src/http/server.rs:506— router/src/http/server.rs:506-514 | router/src/http/server.rs:639-647 | router/src/http/server.rs:826-834 | router/src/http/server.rs:1004-1012 | router/src/http/server.rs:1215-1223 | router/src/http/server.rs:1391-1399 | router/src/http/server.rs:1471-1479 — all 7 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:506` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9 lines × 4) backends/candle/src/models/bert.rs:394— backends/candle/src/models/bert.rs:394-402 | backends/candle/src/models/distilbert.rs:348-356 | backends/candle/src/models/gte.rs:312-320 | backends/candle/src/models/jina.rs:368-376 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:394` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8–9 lines × 3) backends/candle/src/models/flash_distilbert.rs:64— backends/candle/src/models/flash_distilbert.rs:64-71 | backends/candle/src/models/flash_gte.rs:61-69 | backends/candle/src/models/flash_nomic.rs:68-76 — before extracting anything, compare `backends/candle/src/models/flash_distilbert.rs` and `backends/candle/src/models/flash_gte.rs` as WHOLE FILES: this scan already matched 8 separate duplicated blocks between them, totalling at least 107 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/flash_nomic.rs:81` calls `apply_rotary_inplace` and `backends/candle/src/models/flash_distilbert.rs:76` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (8 lines × 7) backends/candle/src/models/bert.rs:174— backends/candle/src/models/bert.rs:174-181 | backends/candle/src/models/flash_bert.rs:83-90 | backends/candle/src/models/flash_jina.rs:85-92 | backends/candle/src/models/flash_jina_code.rs:101-108 | backends/candle/src/models/jina.rs:147-154 | backends/candle/src/models/jina_code.rs:94-101 | backends/candle/src/models/mpnet.rs:142-149 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/flash_bert.rs` as WHOLE FILES: this scan already matched 12 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/mpnet.rs:140` calls `device` and `backends/candle/src/models/flash_bert.rs:81` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (6–8 lines × 5) backends/candle/src/models/flash_bert.rs:378— backends/candle/src/models/flash_bert.rs:378-385 | backends/candle/src/models/flash_distilbert.rs:251-256 | backends/candle/src/models/flash_gte.rs:285-290 | backends/candle/src/models/flash_jina.rs:308-315 | backends/candle/src/models/flash_jina_code.rs:354-361 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/flash_bert.rs:376` calls `from_vec` and `backends/candle/src/models/flash_distilbert.rs:247` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (6–7 lines × 7) backends/candle/src/models/bert.rs:180— backends/candle/src/models/bert.rs:180-185 | backends/candle/src/models/distilbert.rs:121-126 | backends/candle/src/models/gte.rs:93-98 | backends/candle/src/models/jina.rs:153-158 | backends/candle/src/models/modernbert.rs:221-226 | backends/candle/src/models/mpnet.rs:148-154 | backends/candle/src/models/nomic.rs:515-520 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (6 lines × 13) backends/candle/src/models/bert.rs:176— backends/candle/src/models/bert.rs:176-181 | backends/candle/src/models/distilbert.rs:117-122 | backends/candle/src/models/flash_bert.rs:85-90 | backends/candle/src/models/flash_distilbert.rs:66-71 | backends/candle/src/models/flash_gte.rs:63-69 | backends/candle/src/models/flash_jina.rs:87-92 | backends/candle/src/models/flash_jina_code.rs:103-108 | backends/candle/src/models/flash_nomic.rs:70-76 | backends/candle/src/models/gte.rs:89-94 | backends/candle/src/models/jina.rs:149-154 | backends/candle/src/models/jina_code.rs:96-101 | backends/candle/src/models/mpnet.rs:144-149 | backends/candle/src/models/nomic.rs:511-516 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/bert.rs:174` calls `clone` and `backends/candle/src/models/distilbert.rs:115` does not — after which the two agree again for 4 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (6 lines × 8) backends/candle/src/models/bert.rs:721— backends/candle/src/models/bert.rs:721-726 | backends/candle/src/models/debertav2.rs:1135-1140 | backends/candle/src/models/distilbert.rs:522-527 | backends/candle/src/models/gte.rs:504-509 | backends/candle/src/models/jina.rs:486-491 | backends/candle/src/models/jina_code.rs:413-418 | backends/candle/src/models/nomic.rs:788-793 | backends/ort/src/lib.rs:191-197 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/bert.rs:719` calls `with_capacity` and `backends/candle/src/models/debertav2.rs:1132` does not — after which the two agree again for 2 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (5 lines × 8) backends/candle/src/models/bert.rs:199— backends/candle/src/models/bert.rs:199-205 | backends/candle/src/models/distilbert.rs:140-146 | backends/candle/src/models/gemma3.rs:413-417 | backends/candle/src/models/gte.rs:115-121 | backends/candle/src/models/jina.rs:172-178 | backends/candle/src/models/jina_code.rs:141-147 | backends/candle/src/models/nomic.rs:537-543 | backends/candle/src/models/qwen3.rs:216-220 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:199` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (5 lines × 7) backends/candle/src/models/flash_bert.rs:87— backends/candle/src/models/flash_bert.rs:87-92 | backends/candle/src/models/flash_distilbert.rs:68-73 | backends/candle/src/models/flash_jina.rs:89-94 | backends/candle/src/models/flash_jina_code.rs:105-110 | backends/candle/src/models/flash_modernbert.rs:79-83 | backends/candle/src/models/flash_nomic.rs:73-78 | backends/candle/src/models/jina_code.rs:98-103 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_distilbert.rs` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 190 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just before the matched lines, `backends/candle/src/models/flash_bert.rs:83` calls `clone` and `backends/candle/src/models/flash_distilbert.rs:64` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (5 lines × 6) backends/candle/src/models/bert.rs:253— backends/candle/src/models/bert.rs:253-257 | backends/candle/src/models/distilbert.rs:194-198 | backends/candle/src/models/gte.rs:169-173 | backends/candle/src/models/jina.rs:226-230 | backends/candle/src/models/jina_code.rs:195-199 | backends/candle/src/models/nomic.rs:591-595 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:253` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. ★ These copies have DRIFTED, and that is worth reading before extracting anything: just after the matched lines, `backends/candle/src/models/bert.rs:260` calls `forward`, `Some` and `backends/candle/src/models/distilbert.rs:202` does not — after which the two agree again for 3 more lines. One of those two behaviours is the intended one and the other is what a copy-paste left behind, so decide which BEFORE unifying them: extracting the shared part will silently settle it, and if the copy that skips the call is the wrong one, that bug is already live.
Duplicated block (5 lines × 5) backends/candle/src/models/flash_mistral.rs:139— backends/candle/src/models/flash_mistral.rs:139-143 | backends/candle/src/models/flash_qwen2.rs:150-154 | backends/candle/src/models/flash_qwen3.rs:198-202 | backends/candle/src/models/gemma3.rs:490-494 | backends/candle/src/models/qwen3.rs:296-300 — before extracting anything, compare `backends/candle/src/models/flash_mistral.rs` and `backends/candle/src/models/flash_qwen2.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 331 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (11 lines × 10) backends/candle/src/models/bert.rs:944— backends/candle/src/models/bert.rs:944-954 | backends/candle/src/models/debertav2.rs:1495-1505 | backends/candle/src/models/distilbert.rs:737-747 | backends/candle/src/models/flash_bert.rs:538-548 | backends/candle/src/models/flash_gte.rs:440-450 | backends/candle/src/models/flash_jina.rs:444-454 | backends/candle/src/models/flash_modernbert.rs:465-475 | backends/candle/src/models/gte.rs:721-731 | backends/candle/src/models/jina.rs:733-743 | backends/candle/src/models/modernbert.rs:843-853 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/debertav2.rs` as WHOLE FILES: this scan already matched 10 separate duplicated blocks between them, totalling at least 149 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (12 lines × 3) backends/candle/src/models/flash_bert.rs:203— backends/candle/src/models/flash_bert.rs:203-214 | backends/candle/src/models/flash_jina.rs:208-219 | backends/candle/src/models/flash_jina_code.rs:261-272 — before extracting anything, compare `backends/candle/src/models/flash_bert.rs` and `backends/candle/src/models/flash_jina.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 293 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place.
Duplicated block (9 lines × 8) router/src/http/server.rs:320— router/src/http/server.rs:320-328 | router/src/http/server.rs:506-514 | router/src/http/server.rs:639-647 | router/src/http/server.rs:826-834 | router/src/http/server.rs:1004-1012 | router/src/http/server.rs:1215-1223 | router/src/http/server.rs:1391-1399 | router/src/http/server.rs:1471-1479 — all 8 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `router/src/http/server.rs:320` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 5) backends/candle/src/models/bert.rs:395— backends/candle/src/models/bert.rs:395-402 | backends/candle/src/models/bert.rs:444-451 | backends/candle/src/models/distilbert.rs:349-356 | backends/candle/src/models/gte.rs:313-320 | backends/candle/src/models/jina.rs:369-376 — before extracting anything, compare `backends/candle/src/models/bert.rs` and `backends/candle/src/models/distilbert.rs` as WHOLE FILES: this scan already matched 27 separate duplicated blocks between them, totalling at least 440 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/bert.rs:395` it runs out through the closing brace of the declaration holding it — the window is that declaration's tail, not a fragment that begins part-way through something, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (5 lines × 4) backends/candle/src/models/flash_jina.rs:275— backends/candle/src/models/flash_jina.rs:275-279 | backends/candle/src/models/flash_jina_code.rs:322-326 | backends/candle/src/models/jina.rs:444-448 | backends/candle/src/models/jina_code.rs:372-376 — before extracting anything, compare `backends/candle/src/models/flash_jina.rs` and `backends/candle/src/models/flash_jina_code.rs` as WHOLE FILES: this scan already matched 20 separate duplicated blocks between them, totalling at least 288 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/candle/src/models/flash_jina.rs:275` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (25 lines × 3) backends/python/server/text_embeddings_server/models/flash_bert.py:355— backends/python/server/text_embeddings_server/models/flash_bert.py:355-379 | backends/python/server/text_embeddings_server/models/flash_mistral.py:413-437 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:431-455 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/python/server/text_embeddings_server/models/flash_bert.py:355` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (19–20 lines × 2) backends/python/server/text_embeddings_server/models/flash_mistral.py:200— backends/python/server/text_embeddings_server/models/flash_mistral.py:200-218 | backends/python/server/text_embeddings_server/models/flash_qwen3.py:188-207 — before extracting anything, compare `backends/python/server/text_embeddings_server/models/flash_mistral.py` and `backends/python/server/text_embeddings_server/models/flash_qwen3.py` as WHOLE FILES: this scan already matched 13 separate duplicated blocks between them, totalling at least 214 lines, which is the signature of one file having been copied from the other rather than of a helper waiting to be extracted. If that is what happened, the fix is to keep one copy and have the other call it (or delete it), which resolves this row and its siblings together — extracting one helper per block leaves the fork in place. Read the line range as the matched WINDOW rather than a finished unit: at `backends/python/server/text_embeddings_server/models/flash_qwen3.py:188` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The `return` at the foot of the matched lines is the enclosing body's own terminal exit, not an early one: it moves with them unchanged, and each site calls the extracted unit from the position that `return` occupied — no decision has to be handed back and re-acted on.
Duplicated block (18–19 lines × 3) backends/python/server/text_embeddings_server/models/classification_model.py:30— backends/python/server/text_embeddings_server/models/classification_model.py:30-48 | backends/python/server/text_embeddings_server/models/default_model.py:33-51 | backends/python/server/text_embeddings_server/models/masked_model.py:33-50 — the copies sit in sibling files of one directory, so a shared home is within easy reach: extract the block into a single shared function the call sites can all reach — a file they already depend on, or a new one alongside them — and call it from all 3 call sites, so a change lands once.
Duplicated block (8–10 lines × 2) backends/python/server/text_embeddings_server/models/default_model.py:61— backends/python/server/text_embeddings_server/models/default_model.py:61-70 | backends/python/server/text_embeddings_server/models/masked_model.py:62-69 — the copies sit in sibling files of one directory, so a shared home is within easy reach: extract the block into a single shared function the call sites can all reach — a file they already depend on, or a new one alongside them — and call it from both call sites, so a change lands once.
Duplicated block (8–9 lines × 2) backends/python/server/text_embeddings_server/models/__init__.py:110— backends/python/server/text_embeddings_server/models/__init__.py:110-117 | backends/python/server/text_embeddings_server/models/__init__.py:129-137 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
End-of-life runtime: Rust 1.92 rust-toolchain.toml:1— rust-toolchain.toml declares Rust 1.92 as this project's toolchain file, and Rust 1.92, superseded by 1.93 on 2026-01-22 (the Rust project patches only the current stable). An unsupported runtime receives no security patches, so every vulnerability disclosed in it since 2026-01-22 is present and unfixable without moving off it. This is a migration rather than an upgrade: there is no newer release of a runtime that has ended.
Interpolated log message defeats structured logging backends/python/server/text_embeddings_server/utils/kernels.py:25— `info` at line 25 is handed an f-string as its message, so the logger receives one finished string — the values in it cannot be queried, filtered or aggregated as fields, every occurrence reads as a different message, and the string is built even when the record is filtered out. Pass the values beside a constant message instead: `logger.info("order %s moved to %s", order_id, status)` with the standard library's `logging`, which keeps the message and its arguments apart on the record, or `log.info("order moved", order_id=order_id, status=status)` with structlog.
Interpolated log message defeats structured logging backends/python/server/text_embeddings_server/utils/kernels.py:34— `warning` at line 34 is handed an f-string as its message, so the logger receives one finished string — the values in it cannot be queried, filtered or aggregated as fields, every occurrence reads as a different message, and the string is built even when the record is filtered out. Pass the values beside a constant message instead: `logger.warning("order %s moved to %s", order_id, status)` with the standard library's `logging`, which keeps the message and its arguments apart on the record, or `log.warning("order moved", order_id=order_id, status=status)` with structlog.
Interpolated log message defeats structured logging router/src/lib.rs:229— `warn!` at line 229 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `warn!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/lib.rs:236— `info!` at line 236 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/lib.rs:319— `warn!` at line 319 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `warn!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/lib.rs:320— `warn!` at line 320 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `warn!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/lib.rs:360— `info!` at line 360 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/main.rs:235— `info!` at line 235 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `info!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:208— `error!` at line 208 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:320— `error!` at line 320 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:342— `error!` at line 342 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:384— `error!` at line 384 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:506— `error!` at line 506 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:522— `error!` at line 522 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:639— `error!` at line 639 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:655— `error!` at line 655 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:826— `error!` at line 826 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:842— `error!` at line 842 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:1004— `error!` at line 1004 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:1020— `error!` at line 1020 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:1159— `warn!` at line 1159 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `warn!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:1215— `error!` at line 1215 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:1231— `error!` at line 1231 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:1391— `error!` at line 1391 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Interpolated log message defeats structured logging router/src/http/server.rs:1407— `error!` at line 1407 is handed a format string with `{…}` values in it, so tracing records one finished `message` field — the values in it cannot be queried, filtered or aggregated as fields, and every occurrence reads as a different message. Record the values as fields beside a constant message instead: `error!(order_id = %order_id, status = ?status, "order status updated")` (`%` records a value by its `Display`, `?` by its `Debug`, a bare name as itself), which a JSON subscriber, OpenTelemetry or a log store keeps as queryable keys. An error goes in as a field too: `error = %err`.
Duplicated predicate backends/python/server/text_embeddings_server/models/flash_mistral.py:98— `_triton_layer_norm is not None and hidden_states.is_cuda` appears character-identically in 3 files — backends/python/server/text_embeddings_server/models/flash_mistral.py, backends/python/server/text_embeddings_server/models/flash_qwen3.py, backends/python/server/text_embeddings_server/models/jinaBert_model.py. It is one line, so the duplication detector's token window never sees it; the copies drift when only one is corrected. Give the condition a name and one home.
Duplicated predicate backends/python/server/text_embeddings_server/models/flash_mistral.py:57— `getattr(config, "head_dim", None) or config.hidden_size // config.num_attention_heads` appears character-identically in 2 files — backends/python/server/text_embeddings_server/models/flash_mistral.py, backends/python/server/text_embeddings_server/models/flash_qwen3.py. It is one line, so the duplication detector's token window never sees it; the copies drift when only one is corrected. Give the condition a name and one home.
Duplicated predicate backends/python/server/text_embeddings_server/models/flash_mistral.py:134— `isinstance(x.device.type, str) and x.device.type != "mps"` appears character-identically in 2 files — backends/python/server/text_embeddings_server/models/flash_mistral.py, backends/python/server/text_embeddings_server/models/flash_qwen3.py. It is one line, so the duplication detector's token window never sees it; the copies drift when only one is corrected. Give the condition a name and one home.
Duplicated predicate backends/python/server/text_embeddings_server/models/flash_bert.py:349— `self._is_rocm and not ROCM_HAS_FA_VARLEN` appears character-identically in 3 files — backends/python/server/text_embeddings_server/models/flash_bert.py, backends/python/server/text_embeddings_server/models/flash_mistral.py, backends/python/server/text_embeddings_server/models/flash_qwen3.py. It is one line, so the duplication detector's token window never sees it; the copies drift when only one is corrected. Give the condition a name and one home.
Duplicated predicate backends/candle/src/models/bert.rs:903— `(attention_mask.is_some() || has_pooling_requests) && batch_size > 1` appears character-identically in 9 files — backends/candle/src/models/bert.rs, backends/candle/src/models/debertav2.rs, backends/candle/src/models/distilbert.rs, backends/candle/src/models/gte.rs. It is one line, so the duplication detector's token window never sees it; the copies drift when only one is corrected. Give the condition a name and one home.
Duplicated predicate backends/candle/src/models/flash_nomic.rs:288— `self.scaled_rotary_cache.is_some() && batch.max_length > self.max_position_embeddings` appears character-identically in 2 files — backends/candle/src/models/flash_nomic.rs, backends/candle/src/models/nomic.rs. It is one line, so the duplication detector's token window never sees it; the copies drift when only one is corrected. Give the condition a name and one home.
Off-boarding risk: anonymized user #1 — If anonymized user #1 becomes unavailable, 1 significant file(s) lose their only recent owner: router/src/prometheus.rs. Pair on, review, or document these before any departure.
D16 · Bus Factor· Further sole-owners (lower concentration) · ×1
Further sole-owners (lower concentration) — 1 other contributor(s) are each the sole owner of a small amount of code below the off-boarding threshold — folded into the bus-factor score and metrics (2 single-owned of 62 analysed files in total, counted over production source files of roughly 2,400 bytes or more, excluding vendored, generated, example/demo and scaffolding-template trees and test files identified by path convention, largest first; 62 of the 84 production source files in this repository met that bar). They are anonymized user #2 (1 file(s)) — spread or document their files in the same way, at lower priority than the named off-boarding risks above. The files folded here: backends/python/server/text_embeddings_server/utils/tracing.py.
No ADRs found router/Cargo.toml:13— No ADRs found. No recognised ADR directory (`docs/adr/`, `docs/decisions/`, `adr/`, `docs/rfcs/`, an `ADR0001/` folder, or their siblings) exists anywhere in this tree. What was searched, so you can tell an empty log from a search that missed one: every directory under the tree (build output, dependencies and VCS metadata excepted), for a document that is either any non-index page inside a recognised ADR directory, whatever its name and however deeply nested (`docs/adr/use-postgres.md`, `docs/adr/2024/0001-x.md`); or a file anywhere whose name is ADR-shaped (`0001-use-postgres.md`, `adr-012-caching.md`); or, when neither turned anything up, a document carrying the decision-record signature (an "Architecture Decision Record" heading, or Status / Context / Decision / Consequences as section headings). A decision log that clears none of these — unnumbered files outside any recognised directory, without those headings — is not seen by this check and this row is then wrong. If that is your case, say so rather than renaming anything; otherwise, consider recording architectural decisions in `docs/adr/`. ADRs are expected here because `router/Cargo.toml` makes this repository a deployable product (line 13).
No ADRs — No Architecture Decision Records found — no conventional ADR directory, no numbered `NNNN-title` documents in any markup this check reads, and nothing ADR-shaped by content. Design rationale recorded elsewhere (a design-notes tree, a mailing list, pull-request discussion) is not visible to this check and is not re-findable per decision, so a future maintainer cannot ask why one choice was made and get an answer.
P3 · Security & performance tooling· Dependabot does not watch cargo/uv/gitsubmodule · ×1
Dependabot does not watch cargo/uv/gitsubmodule .github/dependabot.yml:2— `.github/dependabot.yml` schedules updates, but no `package-ecosystem` entry covers `cargo`, `uv`, `gitsubmodule`, which this tree depends on — so those dependencies get no update pull requests and no advisory pressure, however complete the dep-scan credit looks. Add the missing entries to the `updates:` list.
No changelog — No CHANGELOG/HISTORY/RELEASES file — what shipped when isn't easy to reconstruct for support or audit. (Versioning/tagging makes releases traceable, but a changelog records the what.)
Outdated: clap — `clap` is locked at 4.6.6 but 4.6.7 is the current stable release on crates.io, and it already satisfies the `"4.1"` requirement declared in backends/Cargo.toml — so the lockfile is behind this repository's own declared range. Run `cargo update -p clap` and commit the updated REDACTED.
Outdated: libc — `libc` is locked at 0.2.189 but 0.2.191 is the current stable release on crates.io, and it already satisfies the `"0.2.149"` requirement declared in router/Cargo.toml — so the lockfile is behind this repository's own declared range. Run `cargo update -p libc` and commit the updated REDACTED.
Outdated: memmap2 — `memmap2` is locked at 0.9.11 but 0.9.12 is the current stable release on crates.io, and it already satisfies the `"^0.9"` requirement declared in backends/candle/Cargo.toml — so the lockfile is behind this repository's own declared range. Run `cargo update -p memmap2` and commit the updated REDACTED.
Outdated: serde_json — `serde_json` is locked at 1.0.151 but 1.0.154 is the current stable release on crates.io, and it already satisfies the `"1.0"` requirement declared in backends/Cargo.toml — so the lockfile is behind this repository's own declared range. Run `cargo update -p serde_json` and commit the updated REDACTED.
Outdated: tokio — `tokio` is locked at 1.53.1 but 1.54.1 is the current stable release on crates.io, and it already satisfies the `"1.25"` requirement declared in backends/Cargo.toml — so the lockfile is behind this repository's own declared range. Run `cargo update -p tokio` and commit the updated REDACTED.
Appendix B — Reproduction & audit trail
Every external tool invocation behind a deep-scan dimension — the tool, its captured version, the exact command, how many findings it yielded, and a link to the retained raw output. To reproduce any finding: check out the same commit and run the command shown (repo-relative — never an absolute scratch path). The complete raw scanner output is retained verbatim under artifacts/raw/ (indexed in artifacts/raw/index.json); per-invocation exit codes and wall-clock durations are in sidecar.json — kept out of this table so the rendered report stays byte-identical across runs of the same commit.
semgrep: not applicable — No personal data was found crossing a boundary the PII/GDPR ruleset checks — nothing written to a log or console sink, placed in a URL or query string, or persisted to browser storage. That is a clean result for the LEAK surface only: this ruleset detects personal data escaping, it does not inventory the personal data a repository holds, so it is not evidence that this repository has no personal-data surface. The personal-data map (Appendix C) and the C1-C5 compliance cards are what speak to that.
disclosure: not applicable — No vulnerability-disclosure policy file found (SECURITY.md/.markdown/.rst/.txt at root or under .github/.forgejo/.gitea/docs, .well-known/security.txt). A coordinated-disclosure policy may live off-repo, so this is not evidenced rather than failed.
runtime-hardening: not applicable — No Kubernetes/orchestration workloads found in the repository manifests; network egress policy is a cluster-native control that may live at the platform/firewall layer, so there is nothing to assess here.
runtime-hardening: not applicable — No Kubernetes/orchestration workloads found in the repository manifests; seccomp/AppArmor/SELinux confinement is a workload-level control, so there is nothing to assess here.
runtime-hardening: not applicable — No Kubernetes/orchestration workloads found in the repository manifests; runtime threat-detection and admission-control policy are cluster-level controls, so there is nothing to assess here.
Run 01a1db84-7497-76d0-a326-1d76ac48669e · every finding is also locatable in findings.md, and the complete scoring record (with exit codes + durations) in sidecar.json.
Issues: 61 · Warnings: 471 · Recommendations: 56 · Info: 5 — Appendix A · all findings · full markdown report.
Generated by the Canine Code Analyzer: deterministic code-health analysis. 14-11-2026 @ 19:24 UTC.
Downloadable artifacts
Machine-readable and reproducible from this commit + frozen rubric — drop them straight into a contract appendix, a CRA dossier, or a downstream SCA / VEX tool.