fix(stats): stop misreporting broken citations for unreadable source meta

[RealDiligent]

Summary

Root cause: citation_summary() (used by vouch stats / kb.stats) built sources_present from store.list_sources(), which skips unreadable meta.yaml files via _load_or_skip. When a source directory exists but its meta is corrupt (hand edit, mojibake), claims citing that source were counted as broken_citation even though the source is on disk.

health.lint() already fixed this asymmetry via _collect_sources_for_lint() — the source id is its directory name, so citations stay honest and lint reports unreadable_source instead of broken_citation.

Fix: Use _collect_sources_for_lint() for the presence set in citation_summary().

Impact: kb.stats citation coverage metrics no longer lie when a source meta is unreadable; operators aren't sent to delete citations for sources that still exist.

Risk / tradeoffs

• broken_citation may drop for KBs that previously looked worse than lint said — that was the bug. vouch lint remains the repair surface for unreadable metas.

Test plan

• test_citation_summary_honest_when_source_meta_unreadable (new)
• tests/test_stats.py pass

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: cfd36e4e-b2ad-4c79-b75b-8d01e97555db

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[RealDiligent]

CI note: the py3.13 failure was an unrelated flaky timing test (\ est_websocket_broadcasts_on_approve\ — broadcast took 1.17s vs 1.0s budget on a loaded runner). py3.11 and py3.12 passed; retriggered CI with an empty commit.