feat(review): add an ordinal improvement/value judgment to the AI review schema

[JSONbored]

Summary

• Adds a valueAssessment field to ModelReview (src/services/ai-review.ts) — an ordinal
  unclear/minor/moderate/significant band plus a one-sentence rationale — populated by the
  same AI-review call the engine already makes. No new LLM call, no new provider/model-selection
  surface.
• This is a genuinely different axis from confidence: confidence is calibrated
  defect-certainty ("how sure am I my own blockers are real"); valueAssessment instead asks "does
  this change plausibly move the codebase forward, given the diff and its stated intent." It is also
  not a risk judgment — that stays the separate deterministic signals/slop.ts tier, untouched here.
• Gated behind a caller-resolved input.improvementSignal boolean, mirroring how
  inlineFindings/findingCategories/securityFocus already work in this file: off by default, zero
  extra prompt/output tokens spent when off.
• Dual-review combination (block mode, two reviewers): surfaces the more conservative (lower) of
  the two reviewers' bands, carrying that reviewer's own rationale (documented in
  composeImprovementSignal's doc comment).
• Sanitizer safety: the prompt explicitly steers the model toward "improvement/value/gain" wording and
  away from "score" and its sibling forbidden terms; new tests assert a representative set of
  rationale strings pass every independent public-comment sanitizer this repo relies on.

Closes #4743. Part of epic #4737 (advances it; the epic itself stays open until its remaining
sub-issues land).

Scope

• The PR title follows type(scope): short summary Conventional Commit format.
• This PR is focused and does not mix unrelated backend, UI, MCP, docs, dependency, and deploy
  changes (2 files: the service module and its test file).
• This follows CONTRIBUTING.md and does not touch GitHub Pages/VitePress/site//CNAME.
• I linked a currently open issue this PR resolves (Closes #4743) — Extend ModelReview with an ordinal improvement/value judgment #4743 is a complete,
  independently-resolved sub-issue of epic Epic: PR improvement signal — quality-delta scoring as the positive-axis counterpart to slop-risk #4737; the epic itself (Epic: PR improvement signal — quality-delta scoring as the positive-axis counterpart to slop-risk #4737) stays open separately
  since it has 8 other sub-issues still pending.

Validation

• git diff --check
• npm run actionlint
• npm run typecheck
• npm run test:coverage locally (unsharded) — every line and branch this PR changed is covered
  (verified with a v8 coverage-map cross-reference against the actual diff hunks, not just the
  aggregate repo percentage)
• npm run test:workers
• npm run build:mcp
• npm run test:mcp-pack
• npm run ui:openapi:check (unaffected — no OpenAPI/schema surface changed)
• npm run ui:lint
• npm run ui:typecheck
• npm run ui:build
• npm audit --audit-level=moderate — 0 vulnerabilities
• New/changed behavior has unit/integration tests for new branches, fallback paths, and sanitizer
  boundaries

Additionally ran the full npm run test:ci composite gate end-to-end (not just its pieces
individually): exit 0, 692 test files / 13,742 tests passed, 2/12 pre-existing skips, zero
failures.

Safety

• No secrets, wallet details, hotkeys, coldkeys, PATs, private keys, raw trust scores, private
  rankings, or private maintainer evidence anywhere in the diff (confirmed by reading the diff).
• Public GitHub text stays sanitized: the new prompt instructions steer the model away from
  "score" and its sibling forbidden terms toward "improvement/value/gain" wording; new tests
  assert a representative set of rationale strings pass isPublicSafeText
  (src/signals/redaction.ts) and both independently-implemented sanitizePublicComment
  functions (src/queue-intelligence.ts and src/github/commands.ts), plus a negative-control
  test proving those assertions are meaningful (a rationale using forbidden vocabulary does trip
  all three).
• Auth/cookie/CORS/GitHub App/Cloudflare/session: N/A — no such surface touched.
• API/OpenAPI/MCP: N/A — this extends an internal type + prompt only; no route/schema/tool changed.
• UI: N/A — no UI/frontend code touched.
• Docs/changelog: N/A — no changelog edit (not a release-prep PR).

UI Evidence

N/A — no visible UI/frontend/docs surface in this PR; it only changes
src/services/ai-review.ts (internal schema + prompt) and its test file.

Notes

• Depends on Config-as-code foundation: improvementSignal as a converged feature (global + per-repo) #4738 (adds improvementSignal as a real ConvergedFeatureKey) before any caller
  can actually resolve and flip this flag on in production. Config-as-code foundation: improvementSignal as a converged feature (global + per-repo) #4738 had not merged to main as of this
  PR. This PR does not block on that landing order: rather than an internal
  resolveConvergedFeature/convergedFeatureActive call (which would require "improvementSignal" to
  already be a valid ConvergedFeatureKey to compile), input.improvementSignal is a plain
  caller-resolved boolean — the exact same pattern this file already uses for
  inlineFindings/findingCategories/securityFocus. That keeps this PR independently compilable,
  testable, and mergeable regardless of when Config-as-code foundation: improvementSignal as a converged feature (global + per-repo) #4738 lands. A follow-up change at whichever call site
  constructs GittensoryAiReviewInput (out of scope for this issue, which is scoped to
  src/services/ai-review.ts only) will set
  improvementSignal: await resolveConvergedFeature(env, manifest, "improvementSignal", repoFullName)
  once Config-as-code foundation: improvementSignal as a converged feature (global + per-repo) #4738 merges.
• Per epic Epic: PR improvement signal — quality-delta scoring as the positive-axis counterpart to slop-risk #4737's design constraints: advisory-only, never wired into any gate/blocker decision path;
  the deterministic risk tier (slop.ts) is untouched by this change.
• Confirmed via a diff read: no new fetch/API call, no new *_AI_MODEL/*_AI_PROVIDER-style env
  var, no new provider-selection config. This rides whatever AI_REVIEW_PLAN/AI_PROVIDER the repo
  already has configured (including a self-hosted ollama entry).

[superagent-security]

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.22%. Comparing base (82ccb70) to head (b2a48ab).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4754   +/-   ##
=======================================
  Coverage   94.21%   94.22%           
=======================================
  Files         440      440           
  Lines       38712    38736   +24     
  Branches    14101    14112   +11     
=======================================
+ Hits        36474    36498   +24     
  Misses       1577     1577           
  Partials      661      661           

Files with missing lines	Coverage Δ
src/services/ai-review.ts	96.71% <100.00%> (+0.12%)	⬆️

🚀 New features to boost your workflow:

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

[loopover-orb]

Warning

🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨

⏸️ Gittensory review result - manual review recommended

_{Review updated: 2026-07-10 23:19:03 UTC}

2 files · no blockers · readiness 100/100 · CI green · clean

⏸️ Suggested Action - Manual Review

• AI review already in progress for this PR head: Another Gittensory pass is already running the AI review for this exact PR head. This pass is skipping to avoid a duplicate LLM call.

Review summary
AI review is already running for this PR head in another Gittensory pass. Gittensory is holding this PR for manual review until that pass completes.

Nits — 2 non-blocking

• PR author also opened the linked issue — Link an issue that was opened by a different contributor, or provide a rationale for why this self-authored issue represents genuine discovery work.
• AI review already in progress for this PR head — The gate is held for a human reviewer rather than passed automatically; it re-evaluates once the in-flight review completes or on the next update.

Signal	Result	Evidence
Code review	✅ No blockers	No AI review summary
Linked issue	✅ Linked	#4743
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	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 48 registered-repo PR(s), 40 merged, 285 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 48 PR(s), 285 issue(s).
Gate result	⚠️ Not blocking	Advisory; not blocking this PR.

Linked issue satisfaction

Addressed
The PR adds an ordinal `valueAssessment` field to `ModelReview` with explicit doc-comment and prompt language distinguishing it from `confidence`/risk, gates it behind a caller-resolved `improvementSignal` flag with byte-identical prompts when off, reuses the existing LLM call with no new provider surface, documents and implements a conservative-band dual-review combination in `composeImprovementS

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, JavaScript, Ruby, Go, Kotlin, MDX, Shell
• Official Gittensor activity: 48 PR(s), 285 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Treat this as maintainer-lane context rather than normal contributor-lane activity.

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