docs: refresh stale items vs current code; archive completed hosting record

CHANGELOG.md

@@ -5,6 +5,10 @@ All notable changes to `asa` are documented here.
 ## Unreleased
 
 ### Added
+- **Optional MT3 polyphonic transcription stage** ([`apps/backend/mt3_transcription.py`](apps/backend/mt3_transcription.py), `apps/ui/src/services/mt3Client.ts`): Google MT3 (T5X) wired as an opt-in, additive staged backend — peer of pitch/note translation. Gated by run-level form field `mt3_mode='enabled'` on `POST /api/analysis-runs` (canonical) and env var `ASA_ENABLE_MT3=1` on the legacy `analyze.py` CLI. Emits per-stem MIDI as artifacts and an `mt3` namespace on the run snapshot. Never overrides measurement (PURPOSE.md invariant #1); heavy dependencies pinned separately in `apps/backend/requirements-mt3.txt`. New endpoint: `POST /api/analysis-runs/{run_id}/mt3-transcriptions`. See [`JSON_SCHEMA.md` "Optional MT3 Namespace"](apps/backend/JSON_SCHEMA.md#optional-mt3-namespace).
+- **Live 12 source-catalogue + Phase 2 output-validation gates** ([`apps/backend/live12_catalogue.py`](apps/backend/live12_catalogue.py), [`apps/backend/phase2_catalogue_gates.py`](apps/backend/phase2_catalogue_gates.py), [`data/live12_catalogue.json`](data/live12_catalogue.json) + schema, [`scripts/build_live12_catalogue.py`](scripts/build_live12_catalogue.py)): regenerable static device/parameter catalogue extracted from the upstream `gluon/AbletonLive12_MIDIRemoteScripts`. Backend cross-checks every `{device, parameter, value, phase1Fields}` record in `mixAndMasterChain`, `abletonRecommendations`, and `secretSauce.workflowSteps` against the catalogue and emits **warn-and-keep** `RECOMMENDATION_UNVERIFIED` events on `validationWarnings` (reasons: `device_unknown`, `parameter_unknown`, `value_out_of_range`, `citation_missing`). Never drops a recommendation, never rewrites a parameter — an earlier fuzzy-rewrite path produced confidently-wrong output, so the contract is warn-only.
+- **Transcription pianoroll endpoint + symusic consolidation** ([`apps/backend/transcription_pianoroll.py`](apps/backend/transcription_pianoroll.py), `apps/ui/src/services/transcriptionPianorollClient.ts`, `apps/ui/src/components/TranscriptionPianoroll.tsx` / `TranscriptionPianorollBlock.tsx`): velocity-encoded `(pitch, time)` uint8 matrix backed by `symusic`. Backs `GET /api/analysis-runs/{run_id}/transcription/pianoroll`; mounted in the Session Musician suite when `transcriptionDetail.noteCount > 0`. Derived view — Phase 1 stays authoritative; response cites the Phase 1 `bpm` and `timeSignature` so every cell traces back to a measurement.
+- **Recommendation-proof campaign** ([`GOAL.md`](GOAL.md) + [`apps/backend/NEEDS.md`](apps/backend/NEEDS.md) + [`apps/backend/RECOMMENDATION_VERDICT.md`](apps/backend/RECOMMENDATION_VERDICT.md)): recommendation-quality scorer ([`apps/backend/recommendation_evaluation.py`](apps/backend/recommendation_evaluation.py) + `scripts/evaluate_recommendations.py`), Node 23+ deterministic source bridge (`scripts/emit_deterministic_recs.ts`), per-recommendation UI corpus-verification badge (`apps/ui/src/services/recommendationVerification.ts`, `apps/ui/src/data/recommendationVerification.ts`, `apps/ui/src/components/RecommendationVerificationBadge.tsx`), and the known-settings fixture corpus at `apps/backend/tests/fixtures/recommendation_tracks/`. Off the product path; badge hidden until corpus has scored fixtures.
 - **Backend Phase 2 citation-path verification** (defense-in-depth): [`_validate_phase2_citation_paths`](apps/backend/server_phase2.py) (called from [`server.py`](apps/backend/server.py)) flags any cited `phase1Fields` path that doesn't resolve against the authoritative measurement payload, mirroring the frontend's `validatePhase1FieldCitations`. WARNING-only — rides the existing `validationWarnings` channel and surfaces in the UI's interpretation-caution section; Phase 1 stays authoritative and citations are never rejected.
 - **`packages/loudness-spectro-wasm`** (Phase 1): browser-first Rust→WASM DSP that lifts openmeters' ITU-R BS.1770-5 / EBU R128 loudness (K-weighting, momentary/short-term, 4× true-peak) and adds integrated loudness + gating + LRA, plus an A-weighted whole-file spectrum and a spectral-reassignment spectrogram. Loudness is validated against absolute EBU Tech 3341/3342 conformance signals, the `ebur128` crate, and a pyloudnorm cross-check, with CI. Standalone — not yet imported by `apps/ui` or `apps/backend`; the Essentia path remains the authoritative Phase 1 loudness source and the product reassigned-spectrogram endpoint runs on librosa. GPL-3.0-or-later (inherited from openmeters). See [`packages/loudness-spectro-wasm/README.md`](packages/loudness-spectro-wasm/README.md).
 - **Loudness-action guardrail** in Phase 2 validation: a `MISSING_LOUDNESS_ACTION` consistency check (`validateLoudnessActionPresence`) backed by [`apps/ui/src/services/loudnessGuardrails.ts`](apps/ui/src/services/loudnessGuardrails.ts). When Phase 1 measures an objective loudness defect (digital clipping via `saturationDetail.clippedSampleCount`, or a true-peak over via `truePeak`), Phase 2 must include a MASTERING/DYNAMICS card citing the triggering measurement. The Phase 2 system prompt was updated to require it.
@@ -41,7 +45,7 @@ All notable changes to `asa` are documented here.
 - **UI design-system migration ("D-series").** Introduced a shared primitive layer in `apps/ui/src/components/ui/` (`Button`, `Panel`, `DeviceRack`, `SectionHeader`, `MetricBar`/`MetricBarRow`/`MetricTile`, `DataTable`, `EmptyState`, `LedIndicator`, `Pill`, `SignalChain`, `ChainSeparator`, `TimeReadout`, `Checkbox`, `Tooltip`) with Storybook stories and semantic design tokens in `src/index.css`, then migrated feature components onto it — inline hex colors replaced with tokens and bespoke layout boxes replaced with primitives. Also: TypeScript strict mode, `prefers-reduced-motion` support, and a source-SR-preserving STFT spectrogram as the default spectrogram tab.
 - **Backend monolith split** into domain modules (commit `5c40dd44`): `analyze_core.py`, `analyze_audio_io.py`, `analyze_detection.py`, `analyze_estimate.py`, `analyze_rhythm.py`, `analyze_segments.py`, `analyze_structure.py`, `analyze_transcription.py`, `analyze_fast.py`. Server routes likewise split into `server_phase1.py`, `server_phase2.py`, `server_upload.py`, `server_samples.py`.
 - **`BatchedBandpass` centralization** ([`apps/backend/dsp_bandbank.py`](apps/backend/dsp_bandbank.py)): per-band 4th-order Butterworth filtering with zero-phase `filtfilt` now lives in one place instead of being duplicated across detectors.
-- **Hosted runtime foundation** landed without disturbing local mode: [`runtime_profile.py`](apps/backend/runtime_profile.py), [`worker.py`](apps/backend/worker.py), [`artifact_storage.py`](apps/backend/artifact_storage.py), [`auth_context.py`](apps/backend/auth_context.py). See [`docs/PUBLIC_HOSTING_FOUNDATION.md`](docs/PUBLIC_HOSTING_FOUNDATION.md).
+- **Hosted runtime foundation** landed without disturbing local mode: [`runtime_profile.py`](apps/backend/runtime_profile.py), [`worker.py`](apps/backend/worker.py), [`artifact_storage.py`](apps/backend/artifact_storage.py), [`auth_context.py`](apps/backend/auth_context.py). See [`docs/history/public-hosting-foundation-2026-04-01.md`](docs/history/public-hosting-foundation-2026-04-01.md).
 - Synthesis character labels aligned to phase2 prompt thresholds (three-tier: clean subtractive / FM-acid / wavetable-noise)
 - Removed citation instructions from Phase 2 system prompt (citations added noise, not value)
 - `dynamicCharacter` forwarded through `_build_phase1()` to Gemini Phase 2