feat(review): capture per-reviewer AI votes with swap-proof attribution

[JSONbored]

Summary

Advances #8229 (stage 0 of 3 — the capture substrate; the would-have-routed report and flag-gated weighting follow once votes accumulate).

The dual-AI combiner never persisted which provider voted which way, and its disagreement tie-break judge re-runs the pair order-swapped — so any capture at slot-interpretation time would misattribute votes. This PR attaches votes where slot↔model is unambiguous by construction:

• LoopOverAiReviewResult ok-branch gains reviewerVotes: { reviewer, votedFail }[], populated at leg production time (a.review ↔ primary.model, b.review ↔ secondary.model) in the dual path and on the single-reviewer block path. Advisory-only runs and unparseable legs cast no votes — never a fabricated stance.
• Orchestration persists each vote as a best-effort reviewer_vote audit event (actor = model id, target = owner/repo#N, stance in metadata) after an ok result; a rejecting write is swallowed and never touches the review outcome.
• The stored shape feeds computeProviderTrackRecords (calibration: per-provider reviewer track records from the consensus corpus #8228) directly in stage 1.

Validation

• Full local gate npm run test:ci green (TEST_CI_EXIT=0) on a branch rebased onto current main, plus npm audit --audit-level=moderate.
• 100% line+branch coverage on every changed range in src/services/ai-review.ts and src/queue/ai-review-orchestration.ts (verified via lcov against the exact diff hunks): split-case attribution, unparseable-leg no-vote, advisory-empty, persistence rows, and the swallowed-rejection arm each have both sides exercised.
• New suite test/unit/reviewer-vote-capture.test.ts pins the persistence contract; runner-side invariants live beside the existing combiner tests in test/unit/ai-review.test.ts.

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[github-actions]

Logic backtest

Replayed 0 historical case(s) for linked_issue_scope_mismatch through the base (264f26a) and head (d3d913f) versions of its detection logic (corpus checksum 4f53cda18c2b).

Backtest comparison: linked_issue_scope_mismatch

Verdict: unchanged — no comparable axis moved.

Advisory only — this check never blocks merge (#8105).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.62%. Comparing base (904ec7d) to head (d3d913f).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8265      +/-   ##
==========================================
- Coverage   92.15%   91.62%   -0.53%     
==========================================
  Files         786      786              
  Lines       78814    78823       +9     
  Branches    23790    23796       +6     
==========================================
- Hits        72630    72224     -406     
- Misses       5062     5524     +462     
+ Partials     1122     1075      -47     

Flag	Coverage Δ
shard-1	56.10% <88.88%> (+2.24%)	⬆️
shard-2	49.48% <22.22%> (-0.79%)	⬇️
shard-3	54.66% <88.88%> (-2.64%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/queue/ai-review-orchestration.ts	100.00% <100.00%> (ø)
src/services/ai-review.ts	96.87% <100.00%> (+0.01%)	⬆️

... and 3 files with indirect coverage changes

[loopover-orb]

Warning

⏸️ LoopOver review result - manual review recommended

_{Review updated: 2026-07-23 17:06:25 UTC}

4 files · 1 AI reviewer · 2 blockers · CI green · clean

⏸️ Suggested Action - Manual Review

Review summary
This PR adds a `reviewerVotes` array to `LoopOverAiReviewResult`'s ok-branch, populated at leg-production time in `runLoopOverAiReview` (src/services/ai-review.ts:2381-2383, 2444-2445, 2509) so slot↔model attribution is fixed before the order-swapped tie-break judge could ever muddy it, and persists each vote as a best-effort audit event in `runAiReviewForAdvisory` (src/queue/ai-review-orchestration.ts:724-734) after the ok-result check, guarded by `.catch(() => undefined)` so a write failure never touches the review outcome. The attribution logic is correct: votes are only pushed when `a.review`/`b.review` is present (never fabricated for an unparseable leg), and stay empty in advisory-only runs since the vote-pushing code only runs inside the `mode === "block"` branch. Tests cover the split-attribution, no-vote-on-unparseable, advisory-empty, and swallowed-rejection cases against the real production paths, not fabricated states.

Nits — 4 non-blocking

• src/queue/ai-review-orchestration.ts:725 awaits each `recordAuditEvent` sequentially inside the `for` loop instead of firing both writes in parallel (e.g. `Promise.allSettled`) — with at most 2 votes this is a negligible cost, not worth blocking on.
• Both changed files are now well over the repo's stated 400-line long-file threshold per the external size-smell brief; consider flagging for a future extraction pass rather than growing them further.
• Consider batching the reviewer_vote writes with `Promise.allSettled` in src/queue/ai-review-orchestration.ts's new loop for a minor latency win when both reviewers vote.
• If `recordAuditEvent`'s signature doesn't already support this call shape elsewhere, add a short comment noting where an existing calibration write uses the identical pattern, to make the 'best-effort like every calibration write' claim easy to verify for the next reader.

Concerns raised — review before merging

• No linked issue detected: No closing reference or linked issue number was found in the PR metadata/body. — If this PR is intended to solve an issue, link it explicitly in the PR body.
• Maintainer requires a linked issue: This repo's maintainer focus manifest requires every PR to reference a tracked issue. — Link the relevant issue (for example Closes #123) before opening the PR.

📋 Copy for AI agents — paste into your coding agent

Fix the following blocker(s) from this PR review:

1. No linked issue detected: No closing reference or linked issue number was found in the PR metadata/body. — If this PR is intended to solve an issue, link it explicitly in the PR body.

2. Maintainer requires a linked issue: This repo's maintainer focus manifest requires every PR to reference a tracked issue. — Link the relevant issue (for example `Closes #123`) before opening the PR.

Decision drivers

• ❌ Code review — 2 blockers (1 reviewer)
• ❌ Gate result — Blocking (Repo-configured hard blocker found.)

Context & advisory signals — never blocks the verdict

Signal	Result	Evidence
Linked issue	⚠️ Missing	No linked issue or no-issue rationale found.
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Change scope	✅ 20/20	Low review scope from cached public metadata (no linked issue context).
Validation posture	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 14 registered-repo PR(s), 14 merged, 247 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 14 PR(s), 247 issue(s).
Improvement	✅ Minor	risk: clean · value: minor · LLM: moderate

Review context

• Author: JSONbored
• Role context: owner (maintainer lane)
• Public audience mode: oss maintainer
• Lane context: Repository is configured for direct PR review.
• Public profile languages: Python, TypeScript, Ruby, Go, MDX, Shell, Solidity, JavaScript
• Official Gittensor activity: 14 PR(s), 247 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Start here: Treat this as maintainer-lane context rather than normal contributor-lane activity.
• Then work through the remaining 2 steps in the Signals table above.

Signal definitions

• Related work = same linked issue, overlapping active PRs, or title/path similarity.
• Change scope = cached public metadata such as size labels, draft state, and review-burden hints.
• Validation posture = whether the PR provides enough public validation/test evidence for maintainer review.
• Contributor workload = public contributor activity and cleanup pressure, not a repo-wide quality failure.
• Contributor context = public GitHub/Gittensor identity context; non-Gittensor status is not a blocker.

🧪 Chat with LoopOver

Ask LoopOver a question about this PR directly in a comment — grounded only in the same cached, public-safe facts shown above, never a new claim.

• @loopover ask <question> answers contribution-quality Q&A with source citations and freshness.
• @loopover chat <question> answers in natural prose from cached decision-pack facts via local inference (maintainer/collaborator; read-only).
• A plain-language @loopover mention with a real question is routed to the closest matching read-only command automatically — no exact syntax required.

Full command reference: https://loopover.ai/docs/loopover-commands

🧪 Experimental — new and may change.

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed}

---

💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →.

Checked by LoopOver, a quiet PR intelligence layer for OSS maintainers.

• Re-run LoopOver review