# Changelog

## Score

- CAI 60 → 60 (+0.7)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.3) — scores are not directly comparable.

## Lenses

- Code Health 75 → 74 (-1.3)
- Architecture 93 → 95 (+1.4)
- Maturity 55 → 60 (+4.8)
- Readiness 52 → 51 (-0.9)
- Security 68 → 68 (-0.3)
- Performance 100 → 100 (+0.0)

## Resolved (47)

- AutoTaggingProcessor.updatePages (cognitive 17) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/processors/AutoTaggingProcessor.java)
- Critical CVE: [GHSA redacted] (python/opendataloader-pdf/uv.lock)
- Critical vulnerability: [GHSA redacted] (python/opendataloader-pdf-mcp/uv.lock)
- Disclosure policy has no reporting contact
- Duplicate intent with divergent signatures: The library exposes two distinct entry points for PDF conversion ('convert' vs 'run') in both Python and Node.js. In Python, `convert_generated.convert` accepts a list of inputs and returns None (side-effect based), while `wrapper.run` accepts a single string input. In Node.js, `index.convert` accepts string|string[] and returns a Promise<string>, while `index.run` accepts a single string and returns a Promise<string>. The distinction between 'convert' and 'run' is semantically unclear, and the input handling (single vs list) is inconsistent across the parallel APIs.
- Duplicated block (11 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/markdown/MarkdownGenerator.java)
- Duplicated block (11 lines × 2) (python/opendataloader-pdf-mcp/src/opendataloader_pdf_mcp/server.py)
- Duplicated block (16–19 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/api/FilterConfig.java)
- Duplicated block (16–20 lines × 3) (verification/ci-verify.py)
- Duplicated block (17 lines × 2) (verification/ci-verify.py)
- Duplicated block (18–19 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/html/HtmlGenerator.java)
- Duplicated block (19–23 lines × 4) (verification/ci-verify.py)
- Duplicated block (21–25 lines × 2) (verification/ci-verify.py)
- Duplicated block (5 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/html/HtmlGenerator.java)
- Duplicated block (6 lines × 2) (build-scripts/set_version.py)
- Duplicated block (6 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/processors/ListProcessor.java)
- Duplicated block (8–9 lines × 2) (verification/ci-verify.py)
- Duplicated block (9 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/markdown/MarkdownGenerator.java)
- Duplicated block (9 lines × 2) (verification/ci-verify.py)
- Duplicated block (9 lines × 2) (verification/ci-verify.py)
- …and 27 more

## New (143)

- AutoTaggingProcessor.addMcidChildren (cognitive 22) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/processors/AutoTaggingProcessor.java)
- AutoTaggingProcessor.processAnnotsForTextChunk (cognitive 18) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/processors/AutoTaggingProcessor.java)
- AutoTaggingProcessor.releaseFormTagging (cognitive 18) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/processors/AutoTaggingProcessor.java)
- Change coupling: HtmlGenerator.java ↔ TextGenerator.java (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/html/HtmlGenerator.java)
- Change coupling: MarkdownGenerator.java ↔ TextGenerator.java (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/markdown/MarkdownGenerator.java)
- Change coupling: cli.ts ↔ wrapper.py (node/opendataloader-pdf/src/cli.ts)
- ChunksWriter.processOperator (cognitive 40) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/autotagging/ChunksWriter.java)
- ChunksWriter.processOperator (cyclomatic 51) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/autotagging/ChunksWriter.java)
- ClassTooLong: ParagraphProcessor (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/processors/ParagraphProcessor.java)
- ClassTooLong: TextLineProcessor (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/processors/TextLineProcessor.java)
- ClassTooLong: XYCutPlusPlusSorter (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/processors/readingorder/XYCutPlusPlusSorter.java)
- Disclosure policy has no reporting contact (.github/SECURITY.md)
- DocumentProcessor.generateOutputs (cyclomatic 17) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/processors/DocumentProcessor.java)
- Duplicated block (10–12 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/html/HtmlGenerator.java)
- Duplicated block (13–14 lines × 2) (.agents/maintainer/review/evals/run.py)
- Duplicated block (5 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/autotagging/ChunksWriter.java)
- Duplicated block (5 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/html/HtmlGenerator.java)
- Duplicated block (6 lines × 2) (.agents/maintainer/issue/evals/run.py)
- Duplicated block (8 lines × 2) (.agents/maintainer/issue/evals/run.py)
- Duplicated block (8 lines × 2) (java/opendataloader-pdf-core/src/main/java/org/opendataloader/pdf/html/HtmlGenerator.java)
- …and 123 more

## Changes since last survey

- 183 commits — 66 feature/other, 117 fixes

## By area

- java/opendataloader-pdf-core — 116 commits
- python/opendataloader-pdf — 31 commits
- .agents/skills — 24 commits
- node/opendataloader-pdf — 5 commits
- java/opendataloader-pdf-cli — 3 commits
- (root) — 1 commit
- examples/python — 1 commit
- python/opendataloader-pdf-mcp — 1 commit
- skills/odl-pdf — 1 commit

## Notable commits

- fix: Fix parser downgrade instructions in Type3 glyph regression test
- fix: fix(cli): refuse --to-stdout with a format that stdout cannot carry
- fix: fix(cli): say that --filter-hidden-text applies to the native pipeline only
- fix: fix(cli): test the --to-stdout check in core and document applyAllTo
- fix: fix(core): accept a PDF whatever its file name
- fix: fix(core): build the ID tree and structure destinations after the pages
- fix: fix(core): check the rest of a split line for a list label
- fix: fix(core): count form draws in one pass and treat the unknown as shared
- fix: fix(core): do not count U+0000 as extracted text
- fix: fix(core): do not give /StructParents to a form that is drawn more than once
- fix: fix(core): do not read an underline below a text as a leader
- fix: fix(core): do not unbox the page number of the lines cache
- fix: fix(core): do not use a structure tree without a parent tree
- fix: fix(core): end a list of bullets at a section title between its items
- fix: fix(core): end a numbered list at text that its items cannot take
- fix: fix(core): end the last list item of a page at an empty line
- fix: fix(core): enter forms that draw a shared form, and bind a key to the form it was made for
- fix: fix(core): give a page its next element when a release removes its first
- fix: fix(core): give up shared forms before the ID tree and destinations are built
- fix: fix(core): join a numbered table only when each row has all its cells
- …and 163 more

## Architecture

- Unchanged — 0 containers · 1 contexts · 0 edges
