fix(review): order-swapped dual-AI tie-break stability

[RealDiligent]

Summary

• Add order-swapped dual-AI tie-break judge with stability gate (fix(review): add order-swapped stability check to the dual-AI tie-break pass #2997)
• Fall back to conservative combineReviews when order is unstable or judge is inconclusive
• Raise patch coverage for the new tie-break paths to satisfy the codecov/patch gate

Test plan

• npm run typecheck
• npx vitest run test/unit/ai-review.test.ts

Closes #2997

[superagent-security]

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

[loopover-orb]

Warning

🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨

⏸️ Gittensory review result - manual review recommended

_{Review updated: 2026-07-05 17:38:15 UTC}

2 files · 1 AI reviewer · no blockers · readiness 80/100 · CI pending · blocked

⏸️ Suggested Action - Manual Review

Review summary
This PR adds an order-swapped dual-AI tie-break judge for the `consensus` combiner: on reviewer disagreement it runs the judge twice with reviewer presentation order swapped, only accepts the resolution if both calls agree on the same physical reviewer (guarding against position bias), and falls back to the existing conservative `combineReviews` on any instability, judge failure, or unparseable output. The stability-matching logic in `dualAiTieBreakVerdictsOrderStable` and the reviewer-index remapping in `mapDualAiTieBreakVerdictToCombineResult` are internally consistent (verified index correspondence: normal-order `reviewer_0`/swapped-order `reviewer_1` both map to the same physical review object), and the fail-safe/fail-closed defaults (missing AI binding, thrown provider errors, unparseable JSON, mismatched swap results) all correctly reduce to the pre-existing conservative path. Test coverage is extensive, exercising both stable and unstable branches, the fallback-model path, and the full integration flow through `runGittensoryAiReview`.

Nits — 5 non-blocking

• src/services/ai-review.ts (buildDualAiTieBreakJudgeUserPrompt/runDualAiTieBreakJudgeCall): the retry count `3` and `max_tokens: 512` are unexplained magic numbers — consider naming constants like the existing `RELIABLE_FALLBACK_MODELS`/`BEST_REVIEW_MODELS` pattern elsewhere in this file.
• The `mapDualAiTieBreakVerdictToCombineResult` policy of fully dismissing a flagged reviewer's blocker when the judge favors the clean reviewer ("trust the pass even when the other reviewer flagged") is a meaningful gate-disposition change — worth a one-line call-out in the PR description since it changes the historical fail-closed-on-disagreement behavior for this specific case, even though the swap-stability gate mitigates false positives.
• Reviewer disagreement now doubles AI spend for that PR (up to 2 judge calls × 3 attempts × 2 models = up to 12 extra calls) — not a blocker since it's bounded and intentional, but worth confirming this is accounted for in daily neuron budget expectations.
• console.warn on tiebreak_order_unstable (ai-review.ts:1833) is fine as structured logging consistent with the file's existing pattern, but double-check it isn't spamming logs on noisy/flaky judge models in production.
• Consider extracting the retry-count (3) and max_tokens (512) into named constants near TIE_BREAK_JUDGE_SYSTEM_PROMPT for discoverability.

Signal	Result	Evidence
Code review	✅ No blockers	1 reviewer
Linked issue	✅ Linked	#2997
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 (1 linked issue).
Validation posture	❌ 5/25	Preflight is holding this PR: the review lane is unavailable, so it is not ready for automated review.
Contributor workload	✅ 10/10	Author activity: 181 registered-repo PR(s), 37 merged, 0 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	RealDiligent; Gittensor profile; 181 PR(s), 0 issue(s).
Gate result	✅ Passing	No configured blocker found.

Review context

• Author: RealDiligent
• Role context: outside_contributor
• Public audience mode: oss maintainer
• Lane context: Repository registration is not available in the local Gittensory cache.
• Public profile languages: JavaScript, Ruby, Svelte, TypeScript, Cuda, Markdown
• Official Gittensor activity: 181 PR(s), 0 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Await review-lane availability.
• Refresh registry data or choose a registered active repo.

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.

_{🟩 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 Gittensory, a quiet PR intelligence layer for OSS maintainers.

• Re-run Gittensory review

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.11%. Comparing base (eb8cb4f) to head (907897d).
⚠️ Report is 25 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3569      +/-   ##
==========================================
+ Coverage   93.09%   93.11%   +0.02%     
==========================================
  Files         301      301              
  Lines       31450    31546      +96     
  Branches    11483    11526      +43     
==========================================
+ Hits        29279    29375      +96     
  Misses       1517     1517              
  Partials      654      654              

Files with missing lines	Coverage Δ
src/services/ai-review.ts	96.90% <100.00%> (+0.65%)	⬆️

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.