fix(embeddings): preserve unknown status counts

[Sinity]

Summary

Preserve unknownness in bounded embedding debt and suppress message coverage when its candidate denominator is only an analyzed-index estimate.

Ref polylogue-b2r9.

Problem

Live source-v4 dogfooding reported 430 pending sessions as Pending messages: 0 even though the bounded status payload explicitly returned pending_messages=null and pending_messages_exact=false. Detailed status also divided 675,724 embedded messages by a stale sqlite_stat1 estimate of 652,760 candidate messages, publishing an impossible 103.5% coverage value.

The archive contains 11,348 retained embedding metadata rows whose message IDs are absent from the rebuilt index, so an approximate denominator and historical numerator do not even describe the same set.

Solution

• Derive message_coverage_percent only when candidate_prose_messages_exact is true.
• Keep the candidate estimate visible with its existing exactness flag, but do not turn it into false precision.
• Render pending/stale message counts as unknown in bounded archive debt and point operators to the detailed status attempt instead of coercing None to zero.
• Add a retained-orphan synthetic fixture that would produce coverage above 100% before this change.

Terminal HTTP-400 exclusions and orphan reconciliation remain separate lifecycle work; this PR does not change the no-retry behavior shipped for polylogue-n846.

Verification

• devtools test tests/unit/cli/test_embed_status_fast.py tests/unit/operations/test_archive_debt.py -> 48 passed in 6.80s.
• devtools verify --quick -> 13/13 steps passed, run 20260710T164654Z-quick-1311820-e24d14c7.
• Pre-push quick gate -> 13/13 steps passed, run 20260710T164750Z-quick-1312562-bed58221.

Summary by CodeRabbit

• Bug Fixes

  • Prevented message coverage from being reported when candidate message counts are estimates rather than exact values.
  • Avoided coverage percentages exceeding 100% in embedding status details.
  • Clarified embedding backlog details when pending or stale message counts cannot be determined exactly.

• CLI

  • Added clearer guidance to run detailed status checks when exact embedding counts are unavailable.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.

[coderabbitai]

Warning

Review limit reached

@Sinity, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 57 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: ca2b6480-592b-444d-9242-74b6638891cc

📥 Commits

Reviewing files that changed from the base of the PR and between a0492fd and f5ef223.

📒 Files selected for processing (2)

• polylogue/storage/embeddings/status_payload.py
• tests/unit/cli/test_embed_status_fast.py

📝 Walkthrough

Walkthrough

Embedding status coverage now requires exact candidate-message counts. Archive-debt projections now distinguish exact pending/stale counts from bounded unknown values and expose corresponding caveats.

Changes

Embedding count exactness

Layer / File(s)	Summary
Exact candidate-count coverage polylogue/storage/embeddings/status_payload.py, tests/unit/cli/test_embed_status_fast.py	message_coverage_percent is suppressed for unavailable or estimated candidate prose counts, including retained embedding-row scenarios.
Bounded archive-debt counts polylogue/operations/archive_debt.py, tests/unit/operations/test_archive_debt.py	Embedding backlog debt rows report pending/stale counts only when exact; otherwise they report unknown bounded counts and direct readers to exact-count details.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• Sinity/polylogue#1523: Introduces related embedding pending-message exactness and catch-up status fields.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title matches the main change: preserving unknown counts and avoiding false precision in embedding status output.
Description check	✅ Passed	The description follows the template well with Summary, Problem, Solution, and Verification filled in clearly.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feature/fix/embedding-status-truth

---

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.}