kert 1f60f8c945 feat(comments): parallel walker + DuckDB view (refs #253) ...

- walker.walk_and_extract iterates docket/cid dirs under root,
  parallelizes per-comment extraction across ThreadPoolExecutor.
  PyMuPDF releases the GIL during decode so threads scale.
  Returns {written, skipped, failed} counts; honors --docket and
  --limit; resume rule via existing combined.md.
- view.rebuild_index walks combined.md files and writes _index.csv;
  view.register exposes it as a DuckDB view (comments_index) for
  filtering/aggregation. Bodies stay on disk; view.body_path is the
  canonical lookup helper for downstream consumers (#254/#255).

CLI + integration test land in the next batch.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-04-23 15:47:15 -04:00

1.7 KiB

Raw Blame History

View Raw