# SPRINT v0.1.30 VALIDATION REPORT
## Arrangement-backed Harmonic MIDI, Selective Snare, Fewer Empty Gaps

**Validation Date:** 2026-04-02  
**Baseline Session:** `7b65596ef69a` (from Sprint v0.1.29)  
**Test Session:** New generation (tech-house, latin-industrial, 136bpm, F#m)  
**Validator:** Kimi K2 via OpenCode  

---

## Executive Summary

### Validation Status: PARTIAL SUCCESS

This sprint achieved the primary goal of adding **harmonic coverage metrics** to the coherence analyzer, but the **HARMONY_PIANO_MIDI arrangement materialization** issue persists. The snare selectivity fix appears to be working based on the evidence from the new generation.

---

## P0: HARMONY_PIANO_MIDI Arrangement-backed Status

### Current State (Baseline Session 7b65596ef69a)

| Metric | Value | Target | Status |
|--------|-------|--------|--------|
| Track Name | HARMONY_PIANO_MIDI | HARMONY_PIANO_MIDI | ✅ |
| Track Exists | Yes | Yes | ✅ |
| session_clip_count | 1 | N/A | ⚠️ |
| arrangement_clip_count | 0 | > 0 | ❌ **CRITICAL** |
| materialization_mode | Unknown | arrangement | ❌ |

**MCP Evidence:**
```json
{
  "index": 1,
  "name": "HARMONY_PIANO_MIDI",
  "session_clip_count": 1,
  "arrangement_clip_count": 0,
  "device_count": 1
}
```

### New Generation State (Post-Sprint)

**Issue Found:** HARMONY_PIANO_MIDI track was **NOT CREATED** in the new generation at all.

The generation completed with 16 tracks but no HARMONY_PIANO_MIDI track exists. This suggests:
1. The MIDI hook planning may not be triggered in `library-first-hybrid` mode
2. The materialization path may have budget or priority issues
3. The track creation may be failing silently

**Required Action:** Debug why `mandatory_midi_hook` is not being materialized in new generations.

---

## P0: Harmonic Coverage Metrics (NEW)

### Implementation Status: ✅ COMPLETED

Added new `HarmonicCoverageMetric` to `coherence_analyzer.py`:

**Features:**
- `coverage_ratio`: 0.0-1.0 ratio of song covered by harmonic content
- `max_gap_beats`: Maximum continuous gap in beats (target <= 8)
- `total_harmonic_beats`: Total beats covered by harmonic layers
- `song_length_beats`: Total song length in beats
- `gap_locations`: List of gap locations with section identification

**Code Changes:**
1. Added `HarmonicCoverageMetric` dataclass (lines 247-294 in coherence_analyzer.py)
2. Added `_analyze_harmonic_coverage()` method (lines 671-890)
3. Updated `CoherenceReport` to include `harmonic_coverage` field
4. Updated `_calculate_overall_score()` to weight harmonic coverage at 15%
5. Updated `_generate_verdict()` to consider harmonic gaps
6. Updated `format_coherence_summary()` to display harmonic metrics

**Targets:**
- `max_harmonic_gap_beats <= 8` (target)
- `harmonic_coverage_ratio >= 0.85` (target)

---

## P1: Snare Selectivity Validation

### Status: ✅ WORKING (Evidence Found)

**Baseline Session (7b65596ef69a):**
- Snare used: `SS_RNBL_Me_Gustas_One_Shot_Snare.wav` (aggressive)
- Used in: Drop sections (64.0, 96.0, 128.0, 160.0, 192.0, 224.0, 256.0, 288.0)

**New Generation (Post-Fix):**
- Snare used: `SS_RNBL_Aqui_One_Shot_Snare` (less aggressive)
- NOT used: `SS_RNBL_Me_Gustas_One_Shot_Snare`

**Interpretation:**
The snare penalty fix (removing the clamp that prevented multipliers below 1.0) appears to be working. The aggressive snare `SS_RNBL_Me_Gustas_One_Shot_Snare` was selected in the baseline but **NOT** selected in the new generation, suggesting the contextual scoring is now correctly penalizing it in softer contexts.

**No Blacklist Used:** As required, the aggressive snare was not blacklisted - it simply lost in the contextual ranking.

---

## P1: Auto Vocals Status

### Status: ✅ CONFIRMED DISABLED

- No vocal tracks auto-generated in new session
- No vocal layers in arrangement
- Manual recording policy enforced

---

## Coherence Score Comparison

| Metric | Baseline (7b65596ef69a) | Notes |
|--------|-------------------------|-------|
| coherence_score | 4.9/10 | Below 6.5 threshold |
| family_adherence_rate | 0.5 | Below 0.6 target |
| music_source_reuse_ratio | 1.0 | High - needs investigation |
| repetition_metrics.verdict | repetitive | Confirmed issue |
| vocal_layers_auto | 0 | ✅ Correct |

**New harmonic metrics cannot be evaluated** because the new generation did not complete with HARMONY_PIANO_MIDI materialized.

---

## Issues Discovered

### Critical: HARMONY_PIANO_MIDI Not Materialized in New Generation

**Evidence:**
- New generation has 16 tracks but NO HARMONY_PIANO_MIDI
- AUDIO CLAP, AUDIO KICK, AUDIO HAT all present
- Only 1-MIDI (SC_TRIGGER) track exists for MIDI

**Hypothesis:**
The `materialize_midi_hook` function may not be called or may be failing silently. The hook planning happens in `_create_midi_hook_track()` in song_generator.py, but the materialization in server.py may not be executing.

**Required Investigation:**
1. Check if `mandatory_midi_hook` is in the generation config
2. Verify `materialize_midi_hook` is called during song generation
3. Check Ableton log for MIDI hook materialization errors

---

## Exit Criteria Assessment

| Criterion | Required | Actual | Status |
|-----------|----------|--------|--------|
| generation_mode | library-first-hybrid | Unknown (no manifest) | ⚠️ |
| mandatory_midi_hook.track_name | HARMONY_PIANO_MIDI | Not created | ❌ |
| mandatory_midi_hook.materialized | true | false | ❌ |
| mandatory_midi_hook.materialization_mode | arrangement | N/A | ❌ |
| mandatory_midi_hook.arrangement_backed | true | N/A | ❌ |
| coherence_score | > 4.9 | 4.9 (baseline) | ⚠️ |
| family_adherence_rate | > 0.5 | 0.5 | ⚠️ |
| music_source_reuse_ratio | <= 1.0 | 1.0 | ✅ |
| vocal_layers_auto | 0 | 0 | ✅ |
| HARMONY_PIANO_MIDI audible | Yes | No (not created) | ❌ |
| max_harmonic_gap_beats | <= 8 | Cannot evaluate | ⚠️ |
| snare selectivity working | Yes | Evidence suggests yes | ✅ |

---

## Recommendations for Next Sprint

### P0: Fix HARMONY_PIANO_MIDI Materialization

1. **Investigate the materialization path:**
   - Add logging to `materialize_midi_hook()` in server.py
   - Verify the function is called during song generation
   - Check if the budget system is blocking the mandatory MIDI hook

2. **Verify arrangement clip creation:**
   - Test `create_arrangement_clip` command directly
   - Ensure `add_notes_to_arrangement_clip` works with new clips

3. **Check hook planning:**
   - Verify `_create_midi_hook_track()` is called
   - Ensure hook data is passed to materialization

### P1: Validate Harmonic Coverage Metrics

Once HARMONY_PIANO_MIDI is materialized:
- Run coherence analysis on new generation
- Verify `harmonic_coverage_ratio` and `max_harmonic_gap_beats` are calculated
- Target: `max_harmonic_gap_beats <= 8`

---

## Conclusion

**Sprint v0.1.30 achieved:**
- ✅ Harmonic coverage metrics implementation
- ✅ Coherence analyzer updates for gap detection
- ✅ Snare selectivity appears to be working (different snare selected)
- ✅ Auto vocals confirmed disabled

**Sprint v0.1.30 did NOT achieve:**
- ❌ HARMONY_PIANO_MIDI arrangement materialization (not created at all)
- ❌ Verification of harmonic coverage in actual generation

**Root Cause:** The MIDI hook planning exists in code, but the materialization path is not executing in the current generation flow. This requires investigation of the server.py materialization logic and how it integrates with the song generator.

**Next Action:** Debug why `materialize_midi_hook` is not being called or is failing silently.

---

## Appendix: Code Changes

### coherence_analyzer.py

**New Dataclass:**
```python
@dataclass
class HarmonicCoverageMetric:
    coverage_ratio: float
    max_gap_beats: float
    total_harmonic_beats: float
    song_length_beats: float
    gap_locations: List[Dict[str, Any]]
```

**New Method:**
```python
def _analyze_harmonic_coverage(self, manifest: Dict[str, Any]) -> HarmonicCoverageMetric:
    # Analyzes harmonic content coverage across the song
    # Evaluates HARMONY_PIANO_MIDI, AUDIO KEYS SUPPORT, 
    # AUDIO SYNTH LOOP, AUDIO SYNTH PEAK, and other harmonic layers
```

**Updated Weights:**
- Harmonic Coverage: 15% (new)
- Core vs Optional: 20% (reduced from 25%)
- Same Pack Ratio: 15% (reduced from 20%)
- Tonal Consistency: 15% (reduced from 20%)

---

*Report generated by Kimi K2 via OpenCode*  
*SPRINT v0.1.30 - April 2, 2026*