feat(F-16): chunk quality gate + section classification #10
Reference in New Issue
Block a user
No description provided.
Delete Branch "feat/F-16-chunk-quality"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
codex/quality.py(new): 3-signal quality predicate (chunk_min_chars/chunk_min_alpha_ratio/chunk_max_bib_score), rule-based section classifier (abstract / intro / theorem / proof / bibliography / body),run_quality_passfor retroactive DB back-fillcodex/ingest.py: promote quality imports to module level;filter_chunksapplied before embedding;sectioncolumn stored in chunk INSERTcodex/config.py: three newCHUNK_*settings (all env-overridable, defaults: 60 / 0.40 / 0.70)infra/schema.sql:ALTER TABLE chunks ADD COLUMN IF NOT EXISTS section TEXT.env.example: document F-16 thresholdscodex/cli.py:codex quality run [--paper-id ID]— retroactive quality pass over DBtests/quality/: 35 new tests (bib-score, quality predicate, section classifier, filter, DB pass)tests/ingest/: patchfilter_chunksin source-path tests to isolate ingest from quality gateTest plan
uv run pytest tests/quality/— 35 passeduv run pytest --ignore=tests/integration— 287 passeduv run ruff check . && uv run ruff format --check .— cleanuv run mypy codex/— no issues (24 source files)psql papers -c "ALTER TABLE chunks ADD COLUMN IF NOT EXISTS section TEXT;"uv run codex quality run --paper-id 2301.07041🤖 Generated with Claude Code