fix(ui): finish confidence-band unification + close estimatePlr audit item

apps/ui/src/components/analysisResultsViewModel.ts

@@ -1,17 +1,10 @@
 import { AbletonRecommendation, Phase1Result, Phase2Result } from "../types";
 import {
   type ConfidenceBand,
+  getConfidenceBand,
   toConfidenceBand,
 } from "../services/sessionMusician/confidenceBand";
 
-// Audit Finding #4: the three-level `ConfidenceLevel` vocabulary stays alive
-// for the melody-insights key/value detail rows in Sonic Element cards
-// (rendered as plain text like "High (78%)", not as a colored pill). The
-// confidence-badge pill path migrated to the canonical four-band ladder
-// via `ConfidenceBand` below. Follow-up: migrate the melody rows too once
-// producers see the new bands in context.
-export type ConfidenceLevel = "High" | "Moderate" | "Low";
-
 export interface ConfidenceBadgeViewModel {
   label: string;
   /**
@@ -213,7 +206,6 @@ export interface MelodyInsightsViewModel {
   dominantNotes: string[];
   rangeLabel: string;
   confidence: number;
-  confidenceLabel: ConfidenceLevel;
   isDraft: boolean;
   source: MelodyInsightsSource;
 }
@@ -473,14 +465,12 @@ export function buildMelodyInsights(phase1: Phase1Result): MelodyInsightsViewMod
       ? `${transcriptionDetail.pitchRange.minName as string} - ${transcriptionDetail.pitchRange.maxName as string}`
       : "n/a";
     const confidence = Math.max(0, Math.min(1, transcriptionDetail.averageConfidence ?? 0));
-    const confidenceLabel: ConfidenceLevel = confidence >= 0.8 ? "High" : confidence >= 0.5 ? "Moderate" : "Low";
 
     return {
       noteCount,
       dominantNotes,
       rangeLabel,
       confidence,
-      confidenceLabel,
       isDraft: confidence < LOW_TRANSCRIPTION_CONFIDENCE_THRESHOLD,
       source: "transcription",
     };
@@ -496,14 +486,12 @@ export function buildMelodyInsights(phase1: Phase1Result): MelodyInsightsViewMod
     ? `${midiToNoteName(detail.pitchRange.min as number)} - ${midiToNoteName(detail.pitchRange.max as number)}`
     : "n/a";
   const confidence = Math.max(0, Math.min(1, detail.pitchConfidence ?? 0));
-  const confidenceLabel: ConfidenceLevel = confidence >= 0.8 ? "High" : confidence >= 0.5 ? "Moderate" : "Low";
 
   return {
     noteCount,
     dominantNotes,
     rangeLabel,
     confidence,
-    confidenceLabel,
     isDraft: confidence < LOW_MELODY_CONFIDENCE_THRESHOLD,
     source: "melody",
   };
@@ -581,7 +569,7 @@ function getSonicMeasurements(
                 melodyInsights.source === "transcription"
                   ? "Transcription"
                   : "Melody Confidence",
-              value: `${melodyInsights.confidenceLabel} (${Math.round(melodyInsights.confidence * 100)}%)`,
+              value: `${getConfidenceBand(melodyInsights.confidence).label} (${Math.round(melodyInsights.confidence * 100)}%)`,
             },
           ]
         : []),
@@ -1109,7 +1097,7 @@ function buildMelodyPatchParameters(insights: MelodyInsightsViewModel): ChainPar
     { label: noteLabel, value: `${insights.noteCount}` },
     { label: "Note Range", value: insights.rangeLabel },
     { label: "Dominant Notes", value: insights.dominantNotes.slice(0, 3).join(", ") || "n/a" },
-    { label: confidenceLabel, value: `${Math.round(insights.confidence * 100)}% (${insights.confidenceLabel})` },
+    { label: confidenceLabel, value: `${Math.round(insights.confidence * 100)}% (${getConfidenceBand(insights.confidence).label})` },
   ];
 }
 

apps/ui/src/services/mixDoctor.ts

@@ -99,6 +99,16 @@ export function resolveMixDoctorGenreId(phase1: Phase1Result): string {
   return DEFAULT_GENRE_ID;
 }
 
+// Audit 2026-05-16 §4 (boundary review): the truePeak − lufsIntegrated branch
+// below is *not* a competing source of truth with Phase 1. The backend's
+// analyze_plr (apps/backend/analyze_core.py) computes
+// `round(true_peak − lufs_integrated, 2)` — byte-identical to the fallback
+// here — and emits `plr` in both --fast (analyze.py:1385) and full
+// (analyze.py:1775) paths. The fallback only fires when `phase1.plr` is
+// null/undefined, which today means: a legacy snapshot replayed from runtime
+// SQLite that predates the field. When it fires it lands on the same number
+// the backend would have. Result is consumed locally for advisory scoring;
+// never written back into the Phase 1 snapshot.
 function estimatePlr(phase1: Phase1Result): number | null {
   const explicit = asFiniteNumber(phase1.plr);
   if (explicit !== null) return roundTo(explicit, 2);

apps/ui/tests/services/analysisResultsViewModel.test.ts

@@ -9,6 +9,7 @@ import {
   truncateAtSentenceBoundary,
   truncateBySentenceCount,
 } from '../../src/components/analysisResultsViewModel';
+import { getConfidenceBand } from '../../src/services/sessionMusician/confidenceBand';
 import { MeasurementResult, Phase2Result, TranscriptionDetail } from '../../src/types';
 
 const measurement: MeasurementResult = {
@@ -747,7 +748,43 @@ describe('analysisResultsViewModel helpers', () => {
     expect(insights?.noteCount).toBe(6);
     expect(insights?.rangeLabel).toBe('C3 - G5');
     expect(insights?.dominantNotes).toEqual(['C3', 'G3']);
-    expect(insights?.confidenceLabel).toBe('High');
     expect(insights?.isDraft).toBe(false);
+    expect(insights?.confidence).toBeCloseTo(0.83);
+    expect(getConfidenceBand(insights!.confidence).label).toBe('Solid scaffold');
+  });
+
+  // PR #60: cover the two render sites that interpolate the canonical band
+  // label, not just the upstream getConfidenceBand mapping. A regression in
+  // either string-interpolation site would otherwise hide until manual QA.
+  it('renders the canonical band label inside the Sonic Element melody row (transcription path)', () => {
+    const sonicCards = buildSonicElementCards(
+      { ...measurement, transcriptionDetail: pitchNote },
+      {
+        kick: 'Kick.',
+        bass: 'Bass.',
+        melodicArp: 'Arp.',
+        grooveAndTiming: 'Groove.',
+        effectsAndTexture: 'Fx.',
+        widthAndStereo: 'Width.',
+        harmonicContent: 'Harmony.',
+      },
+    );
+    const melodic = sonicCards.find((card) => card.id === 'melodicArp');
+    const confidenceRow = melodic?.measurements.find((m) => m.label === 'Transcription');
+    expect(confidenceRow?.value).toBe('Solid scaffold (83%)');
+  });
+
+  it('renders the canonical band label inside the MIDI Clip Guide patch parameters (melody fallback path)', () => {
+    // Phase 2 with no MIDI-focused recommendations forces the synthetic
+    // "MIDI Clip Guide" card, which calls buildMelodyPatchParameters with
+    // the melodyDetail-derived insights (pitchConfidence: 0.72 → workable).
+    const phase2 = {
+      mixAndMasterChain: [],
+      synthAndDrumPatches: [],
+    } as unknown as Phase2Result;
+    const cards = buildPatchCards(measurement, phase2);
+    const guide = cards.find((c) => c.device === 'MIDI Clip Guide');
+    const confidenceRow = guide?.parameters.find((p) => p.label === 'Melody Confidence');
+    expect(confidenceRow?.value).toBe('72% (Workable draft)');
   });
 });

audits/nightly-2026-05-16.md

@@ -160,6 +160,16 @@ and the safer-but-stricter reading would still flag this.
 emit `plr`, or keep the fallback and document it as an intentional derived
 display rather than a Phase 1 override.
 
+**Resolution (post-merge follow-up):** keep the fallback, document why.
+Verified `analyze_plr` in [apps/backend/analyze_core.py:624-635](../apps/backend/analyze_core.py) computes
+`round(true_peak − lufs_integrated, 2)` — byte-identical to the UI fallback —
+and that `plr` is emitted in both `--fast` ([analyze.py:1385](../apps/backend/analyze.py))
+and full ([analyze.py:1775](../apps/backend/analyze.py)) paths. The fallback fires only
+when `phase1.plr` is null/undefined (legacy snapshots replayed from runtime
+SQLite that predate the field) and lands on the same number the backend would
+have. Documented inline above `estimatePlr` so the next audit doesn't re-flag
+this. See [apps/ui/src/services/mixDoctor.ts:102](../apps/ui/src/services/mixDoctor.ts).
+
 ---
 
 ### What was *not* found