review: remove the legacy comment renderer (#6103)

src/env.d.ts

@@ -302,9 +302,10 @@ declare global {
      *  secret, never a public var. When absent, BYOK is unavailable and review uses the configured instance
      *  reviewer when available. */
     TOKEN_ENCRYPTION_SECRET?: string;
-    /** Convergence (Stage D): when truthy, the public PR comment is rendered by the unified-comment bridge
-     *  (ONE in-place comment in the converged shape) instead of the legacy `buildPublicPrIntelligenceComment`
-     *  panel. Default OFF — unset/false keeps the legacy panel byte-identical. */
+    /** #6103: retired -- the unified-comment bridge is now the ONLY PR-comment renderer unconditionally (the
+     *  legacy `buildPublicPrIntelligenceComment` panel this flag used to switch away from was deleted, having
+     *  no remaining production caller). No longer read; kept only so an operator's existing deployment config
+     *  setting it doesn't error. Safe to remove from any env once noticed. */
     LOOPOVER_REVIEW_UNIFIED_COMMENT?: string;
     /** Inline comments (#inline-comments): when truthy (AND the repo is in LOOPOVER_REVIEW_REPOS AND the repo's
      *  `.loopover.yml` sets `review.inline_comments: true`), the AI reviewer ALSO leaves quiet, NON-BLOCKING

src/queue/processors.ts

@@ -301,7 +301,6 @@ import {
   buildContributorStrategy,
   buildDuplicateWinnerRelatedWorkView,
   buildPreflightResult,
-  buildPublicPrIntelligenceComment,
   buildPublicPrPanelSignalRows,
   buildPublicReadinessScore,
   buildPublicSafeCollapsibles,
@@ -7700,22 +7699,13 @@ async function maybePublishPrPublicSurface(
   // a dry-run / pause / global-freeze publishes NOTHING (check-run, comment, label) — the gate verdict is still
   // computed + returned for the disposition logic, the writes are just suppressed + audited. (#dry-run-chokepoint)
   const mode = await resolveRepoActionMode(env, settings);
-  // Per-repo feature override (phase 2): the unified converged comment renders for THIS repo when the global
-  // LOOPOVER_REVIEW_UNIFIED_COMMENT kill-switch is ON and the repo's container-private `.loopover.yml`
-  // `features.unifiedComment` opts in — falling back to the LOOPOVER_REVIEW_REPOS allowlist when the manifest
-  // says nothing (byte-identical default). Computed once and used by both unified-comment sites below.
-  const unifiedCommentAllowed = await convergedFeatureActive(
-    env,
-    repoFullName,
-    "unifiedComment",
-  );
   // improvementSignal (#4744): the first real caller of #4738's activation wiring (epic #4737's config-as-code
   // foundation) -- nothing resolved this feature before this PR (see signals/improvement.ts's own header
-  // comment). Resolved once, independent of unifiedCommentAllowed above: it gates BOTH the deterministic
-  // tier's own computation further below (which has no AI dependency at all -- a paused repo, a non-reviewable
-  // author, or aiReviewMode: "off" still gets it) and, threaded into runAiReviewForAdvisory, the LLM tier's
-  // prompt addition (#4743). loadRepoFocusManifest is cached, so this second manifest resolution costs no
-  // extra fetch in the common case where something else already resolved it this pass.
+  // comment). Gates BOTH the deterministic tier's own computation further below (which has no AI dependency
+  // at all -- a paused repo, a non-reviewable author, or aiReviewMode: "off" still gets it) and, threaded
+  // into runAiReviewForAdvisory, the LLM tier's prompt addition (#4743). loadRepoFocusManifest is cached, so
+  // this second manifest resolution costs no extra fetch in the common case where something else already
+  // resolved it this pass.
   const improvementSignalAllowed = await convergedFeatureActive(
     env,
     repoFullName,
@@ -9242,9 +9232,8 @@ async function maybePublishPrPublicSurface(
               reviewSelfHostAiModel,
               reviewImpactMap,
               reviewCultureProfile,
-              // improvementSignal (#4744): resolved once above (independent of unifiedCommentAllowed), reused
-              // here so the LLM tier's value-assessment prompt addition (#4743) only fires when this repo has
-              // actually opted in.
+              // improvementSignal (#4744): resolved once above, reused here so the LLM tier's value-assessment
+              // prompt addition (#4743) only fires when this repo has actually opted in.
               improvementSignal: improvementSignalAllowed,
               // #regate-dup-prep: this call's own advisory lock is already claimed (by aiReviewCacheReadDecideAndRun's
               // caller, above) — pass it through so runAiReviewForAdvisory trusts it instead of re-claiming (and
@@ -9941,30 +9930,12 @@ async function maybePublishPrPublicSurface(
           })),
         })
       : undefined;
-    const commentArgs = {
-      repo,
-      pr,
-      profile,
-      detection,
-      queueHealth,
-      collisions,
-      preflight,
-      settings,
-      gate: gateEvaluation,
-      review: reviewConfig,
-      aiReview,
-      improvementSignal: structuralImprovementAssessment,
-      // #4745: the risk × value quadrant's risk half -- reuses the slop band already computed above (if any);
-      // never a second buildSlopAssessment call.
-      slopBand: slopBand ?? undefined,
-      duplicateWinnerEnabled,
-      env,
-    };
     let deterministicBody: string;
-    // Convergence (Stage D): when the unified-review-comment flag is ON, render the single converged comment
-    // (loopover shape + reviewbot's review folded in). The gate stays authoritative (passed as `decision`),
-    // and the body carries the SAME panel marker so the upsert updates in place. Flag-OFF (default) keeps the
-    // legacy panel byte-identical. Only the comment lane is affected; the gate check-run/labels/audit are not.
+    // Convergence (Stage D, #6103): the converged comment (loopover shape + reviewbot's review folded in) is
+    // the only comment path -- the legacy buildPublicPrIntelligenceComment panel was retired once it had no
+    // remaining production caller (settings-preview.ts's sample preview was migrated to this same renderer).
+    // The gate stays authoritative (passed as `decision`), and the body carries the SAME panel marker so the
+    // upsert updates in place.
     //
     // RECONCILIATION INVARIANT (#1016 — two-gate → one authoritative path; pinned by
     // test/unit/unified-comment-bridge.test.ts "reconciliation invariant"):
@@ -9979,7 +9950,21 @@ async function maybePublishPrPublicSurface(
     //      contradict the review-agent check-run conclusion.
     //   3. The `ai_consensus_defect` surfaces exactly ONCE — as the Code-review blocker — never also in the
     //      gate signal row (which renders only the conclusion-derived status text, not the defect string).
-    if (unifiedCommentAllowed && gateEvaluation) {
+    {
+      // #6103: the converged renderer is the only comment path (the legacy buildPublicPrIntelligenceComment
+      // panel was retired). When the gate was never evaluated for this repo (reviewCheckMode disabled AND no
+      // autonomy configured -- see shouldEvaluateGate above), synthesize a "skipped" gate for rendering
+      // purposes only, mirroring buildClosedUnifiedCommentBody's own pattern for the identical situation.
+      // This never touches the check-run/label/audit/disposition lanes (those still read the real,
+      // possibly-undefined `gateEvaluation`) -- only what this comment's verdict/"Gate result" row says.
+      const commentGateEvaluation: NonNullable<typeof gateEvaluation> = gateEvaluation ?? {
+        enabled: false,
+        conclusion: "skipped",
+        title: `${LOOPOVER_GATE_CHECK_NAME} skipped`,
+        summary: "Gate evaluation is not configured for this repository.",
+        blockers: [],
+        warnings: [],
+      };
       // FIX B: the unified comment's file count + visual-capture path filter need the real diff — reuse the
       // shared resolver (one resolve per review; inline-fetches when stored is still empty pre-detail-sync).
       const unifiedFiles = await getReviewFiles();
@@ -10025,7 +10010,7 @@ async function maybePublishPrPublicSurface(
         repoFullName,
       });
       // The public comment must match the authoritative Gate check-run conclusion.
-      const commentGate = gateEvaluation;
+      const commentGate = commentGateEvaluation;
       // Observability (#reviews-dashboard): record the would-be gate verdict so the Grafana panel shows the
       // merge/close/hold mix — the "are we rubber-stamping?" signal — even in advisory/dryRun (this is the rendered verdict).
       incr("loopover_gate_decisions_total", {
@@ -10337,8 +10322,6 @@ async function maybePublishPrPublicSurface(
         // render warnings from the full/private manifest here.
         manifestWarnings: publicRepoFocusManifestForComment?.warnings ?? [],
       });
-    } else {
-      deterministicBody = buildPublicPrIntelligenceComment(commentArgs);
     }
     try {
       await withReviewPipelineSpan(

src/review/unified-comment-bridge.ts

@@ -2,13 +2,14 @@
 //
 // A PURE, testable mapping from loopover's live PR-review data (the gate `GateCheckEvaluation`, the AI
 // `advisoryNotes` + consensus defect, the readiness signal rows + total, the footer) onto the ported
-// unified renderer (`renderUnifiedReviewComment`). Flag-gated and default-OFF in the processor; flag-OFF
-// keeps the legacy `buildPublicPrIntelligenceComment` path byte-identical.
+// unified renderer (`renderUnifiedReviewComment`). #6103: this is now the ONLY PR-comment renderer --
+// the legacy `buildPublicPrIntelligenceComment` path this used to sit alongside was deleted once it had
+// no remaining production caller.
 //
 // loopover's GATE stays authoritative: we pass the gate-derived `decision` into `buildUnifiedReviewInput`
 // so `deriveUnifiedStatus` lets it override the reviewer recommendations (the renderer already enforces
-// this). The output PREPENDS the exact panel marker the legacy body carries, so the existing in-place
-// upsert (`createOrUpdatePrIntelligenceComment`) updates the same comment instead of posting a duplicate.
+// this). The output PREPENDS the panel marker, so the existing in-place upsert
+// (`createOrUpdatePrIntelligenceComment`) updates the same comment instead of posting a duplicate.
 //
 // Public-safe: most inputs are already safe by construction — the AI notes via
 // `composeAdvisoryNotes`→`toPublicSafe`; the consensus-defect blocker via `toPublicSafe` (in
@@ -64,8 +65,7 @@ export { splitAiReviewNits } from "./ai-notes";
 // `warnings` (turned into Nits): they carry an AdvisoryFinding's raw title/action. The gate/check-run
 // path sanitizes those strings (sanitizeForCheckRun) before they reach GitHub, but this comment path
 // did not. Rather than trust that every present and FUTURE warning finding is benign, scrub Nits with a
-// boundary mirroring the check-run sanitizer + the legacy panel's private-term guard, and DROP a Nit
-// that still trips the guard. This never alters flag-OFF (the legacy panel keeps its own filtering).
+// boundary mirroring the check-run sanitizer's own private-term guard, and DROP a Nit that still trips it.
 //
 // Mirrors src/rules/advisory.ts CHECK_RUN_FORBIDDEN_TERMS (scrubbed → "[context]") and
 // src/signals/engine.ts containsPrivatePublicTerm (drop if still present). Kept inline so this module
@@ -114,7 +114,7 @@ export function verdictToRecommendation(verdict: Verdict): ReviewRecommendation
   }
 }
 
-/** Derive an ok/warn/fail state from a legacy panel result cell's leading status icon (✅/⚠️/❌). */
+/** Derive an ok/warn/fail state from a panel result cell's leading status icon (✅/⚠️/❌). */
 function rowState(resultCell: string): UnifiedSignalRow["state"] {
   if (resultCell.startsWith("✅")) return "ok";
   if (resultCell.startsWith("❌")) return "fail";
@@ -129,11 +129,12 @@ function rowResultText(resultCell: string): string {
   return resultCell.replace(/^[✅⚠️❌ℹ️]+\s*/u, "").trim();
 }
 
-/** Map the legacy panel signal rows → the unified table's rows (label/state/result/evidence). The
- *  unified renderer adds its own "Code review" row first; these follow it (loopover's gate row included).
- *  `gates: true` only for the "Gate result" row (#6067) -- the ONLY row among these that can actually move
- *  the verdict; every other row's own Evidence/Action text already says it's advisory-only. Drives the split
- *  between the renderer's always-visible "Decision drivers" list and its collapsed advisory-signals fold. */
+/** Map the panel signal rows (buildPublicPrPanelSignalRows) → the unified table's rows
+ *  (label/state/result/evidence). The unified renderer adds its own "Code review" row first; these follow
+ *  it (loopover's gate row included). `gates: true` only for the "Gate result" row (#6067) -- the ONLY row
+ *  among these that can actually move the verdict; every other row's own Evidence/Action text already says
+ *  it's advisory-only. Drives the split between the renderer's always-visible "Decision drivers" list and
+ *  its collapsed advisory-signals fold. */
 export function panelRowsToSignalRows(rows: PublicPrPanelSignalRow[]): UnifiedSignalRow[] {
   return rows.map((row) => {
     const [label, result, evidence] = row.cells;
@@ -322,7 +323,7 @@ export type UnifiedCommentBridgeArgs = {
   aiReview?: { notes: string } | undefined;
   /** The advisory findings — the bridge recovers the `ai_consensus_defect` consensus blocker from here. */
   advisoryFindings?: AdvisoryFinding[] | undefined;
-  /** The legacy panel readiness signal rows (from `buildPublicPrPanelSignalRows`). */
+  /** The readiness signal rows (from `buildPublicPrPanelSignalRows`). */
   panelRows: PublicPrPanelSignalRow[];
   /** Which rows the maintainer kept visible (`.loopover.yml review.fields`); a key set to `false` is hidden. */
   reviewFields?: Partial<Record<PublicPrPanelSignalRow["key"], boolean>> | undefined;
@@ -822,7 +823,7 @@ export function buildUnifiedCommentBody(args: UnifiedCommentBridgeArgs): string
         : input.reviewerCount
       : 0;
 
-  // Honor `.loopover.yml review.fields` row visibility, exactly as the legacy panel does.
+  // Honor `.loopover.yml review.fields` row visibility.
   const visibleRows = args.panelRows.filter((row) => args.reviewFields?.[row.key] !== false);
   const signals = panelRowsToSignalRows(visibleRows);
 
@@ -921,7 +922,7 @@ export function buildUnifiedCommentBody(args: UnifiedCommentBridgeArgs): string
  * routed through `buildUnifiedCommentBody` so a comment that started life as a unified OPEN-PR comment keeps
  * its unified shape (and the SAME marker) when the PR closes, instead of being overwritten by the legacy
  * panel under the shared marker. A synthetic `skipped` gate maps (via `gateConclusionToVerdict`) to the
- * `comment` verdict → `advisory` status, matching the legacy panel's non-blocking NOTE tone. No AI review,
+ * `comment` verdict → `advisory` status (a non-blocking NOTE tone). No AI review,
  * no findings, and a single synthetic "Gate result — Skipped" signal row (the only signal we can assert for
  * a PR we never finished evaluating). Public-safe by construction: every string here is a static literal.
  */

src/signals/settings-preview.ts

@@ -9,17 +9,20 @@ import {
   buildCollisionReport,
   buildContributorProfile,
   buildPreflightResult,
-  buildPublicPrIntelligenceComment,
+  buildPublicPrPanelSignalRows,
   buildPublicReadinessScore,
   buildQueueHealth,
   type ContributorDetection,
 } from "./engine";
 import { buildExtensionPrStatus, type ExtensionPrStatus } from "./extension-contributor-context";
 import { REQUIRED_INSTALLATION_PERMISSIONS } from "../github/backfill";
-import type { LoopOverFooterEnv } from "../github/footer";
+import { loopoverFooter, type LoopOverFooterEnv } from "../github/footer";
+import type { GateCheckConclusion, GateCheckEvaluation } from "../rules/advisory";
 import { LOOPOVER_GATE_CHECK_NAME, shouldPublishReviewCheck } from "../review/check-names";
 import { decideReviewEligibility } from "../review/review-eligibility";
+import { buildUnifiedCommentBody } from "../review/unified-comment-bridge";
 import { requiredAgentActionPermissions } from "../settings/agent-execution";
+import { isAgentConfigured } from "../settings/autonomy";
 
 export function hasVisiblePrSurface(settings: RepositorySettings): boolean {
   return settings.publicSurface !== "off" || settings.checkRunMode === "enabled" || shouldPublishReviewCheck(settings.reviewCheckMode);
@@ -722,5 +725,40 @@ function buildSamplePreviewComment(args: {
     args.issues,
     args.pullRequests,
   );
-  return buildPublicPrIntelligenceComment({ repo: args.repo, pr: samplePr, profile, detection, queueHealth, collisions, preflight, settings: args.settings, env: args.env });
+
+  // Simulated gate verdict for this sample PR (#6103: migrated off the retired legacy renderer). Mirrors
+  // the same enabled / hard-linked-issue-block heuristic the shared panel builder used to fall back on
+  // internally when no real gate had run -- a duplicate-PR block is never simulated here since the
+  // synthetic PR #0 can't realistically collide with anything in this repo's real open PRs.
+  const gateEnabled = shouldPublishReviewCheck(args.settings.reviewCheckMode) || isAgentConfigured(args.settings.autonomy);
+  const hardLinkedIssueBlock = args.settings.linkedIssueGateMode === "block" && samplePr.linkedIssues.length === 0;
+  const gateConclusion: GateCheckConclusion = !gateEnabled ? "success" : hardLinkedIssueBlock ? "failure" : "success";
+  const gate: GateCheckEvaluation = {
+    enabled: gateEnabled,
+    conclusion: gateConclusion,
+    title: !gateEnabled ? `${LOOPOVER_GATE_CHECK_NAME} not configured` : gateConclusion === "failure" ? `${LOOPOVER_GATE_CHECK_NAME} failed` : `${LOOPOVER_GATE_CHECK_NAME} passed`,
+    summary: "Simulated for this settings preview — no live gate evaluation ran.",
+    blockers: [],
+    warnings: [],
+  };
+
+  const { rows, readinessTotal } = buildPublicPrPanelSignalRows({
+    repo: args.repo,
+    pr: samplePr,
+    profile,
+    detection,
+    queueHealth,
+    collisions,
+    preflight,
+    settings: args.settings,
+    gate: { conclusion: gateConclusion, summary: gate.summary },
+  });
+
+  return buildUnifiedCommentBody({
+    gate,
+    panelRows: rows,
+    readinessTotal,
+    changedFiles: 0,
+    footerMarkdown: loopoverFooter(args.env, {}),
+  });
 }