# Changelog

## Score

- CAI 54 → 56 (+1.4)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.4) — scores are not directly comparable.

## Lenses

- Code Health 78 → 75 (-3.0)
- Architecture 85 → 86 (+1.2)
- Maturity 57 → 63 (+5.9)
- Readiness 42 → 42 (+0.0)
- Security 69 → 69 (+0.0)

## Resolved (4)

- Duplicated block (7 lines × 3) (cli/display.py)
- Duplicated block (8 lines × 2) (cli/display.py)
- Hotspot: tradingagents/agents/context.py (tradingagents/agents/context.py)
- No ADRs found

## New (8)

- Edited copy of a member (10 corresponding lines) (tradingagents/llm_clients/factory.py)
- Edited copy of a member (41 corresponding lines) (tradingagents/agents/analysts/fundamentals_analyst.py)
- Edited copy of a member (51 corresponding lines) (tradingagents/memory/log.py)
- Hotspot: tradingagents/memory/log.py (tradingagents/memory/log.py)
- No ADRs found (pyproject.toml)
- Repeated repair: tradingagents/dataflows/files.py (tradingagents/dataflows/files.py)
- Repeated repair: tradingagents/memory/settlement.py (tradingagents/memory/settlement.py)
- Scattered collaborators

## Changes since last survey

- 23 commits — 5 feature/other, 18 fixes

## By area

- (root) — 3 commits
- tests/test_tool_date_enforcement.py — 3 commits
- tests/test_backtest.py — 2 commits
- tests/test_memory_log.py — 2 commits
- tests/test_portfolio_context.py — 2 commits
- tests/test_sec_edgar.py — 2 commits
- (repo) — 1 commit
- cli/prompts.py — 1 commit
- tests/test_anthropic_effort.py — 1 commit
- tests/test_azure_client.py — 1 commit
- tests/test_cache_writes.py — 1 commit
- tests/test_polymarket.py — 1 commit
- tests/test_reporting.py — 1 commit
- tradingagents/__init__.py — 1 commit
- tradingagents/dataflows — 1 commit

## Notable commits

- fix: fix(backtest): give each sweep started without a run_id its own run (#1549)
- fix: fix(backtest): report the settlement failures a sweep's last pass records (#1557)
- fix: fix(cli): check a ticker with the data layer's own rule (#1604)
- fix: fix(llm): pass the output-token cap and a configured endpoint to Azure OpenAI (#1584, #1600)
- fix: fix(llm): read a dated Claude snapshot's version without its date (#1585)
- fix: fix(memory): settle decisions on final closes, across long market closures (#1548, #1539)
- fix: fix(memory): take the most recent lessons by decision date (#1603)
- fix: fix(polymarket): keep a long shot's odds from printing as 0% or 100% (#1598)
- fix: fix(portfolio): match a holding listed under a broker alias to the analyzed instrument (#1556)
- fix: fix(portfolio): refuse a portfolio with a number that is not finite (#1509)
- fix: fix(reporting): remove a previous run's sections when a report is saved to the same folder (#1566)
- fix: fix(sec_edgar): print a figure under one million instead of 0 (#1589)
- fix: fix(sec_edgar): serve a statement one row per period (#1517, #1542)
- fix: fix(tools): accept only the statement frequencies the vendors serve (#1593)
- fix: fix(tools): hand vendors the model's dates as YYYY-MM-DD (#1488)
- fix: fix(tools): send a date window that ends before it starts back to the model (#1595)
- fix: fix(yahoo): print prices with their significant digits (#1567)
- fix: test: fix tests that failed on Windows and macOS (#1506, #1485, #1541)
- change: Merge pull request #1613 from TauricResearch/v0.7.1
- change: announce Trading-R1 official release
- …and 3 more
