push

.gitignore

@@ -1,2 +1,4 @@
-./verification_outputs
-./test_assets
+verification_outputs/
+test_assets/*.wav
+__pycache__/
+*.pyc

README.md

@@ -39,7 +39,7 @@ The UI exposes three explicit inference runtime modes:
 
 ## E2E Smoke Test
 
-A small reproducible test sample is stored at `test_assets/public_domain_tone.wav`.
+A small reproducible test sample is generated on demand at `test_assets/public_domain_tone.wav`.
 
 Run one-file smoke tests across all runtime modes and write file-backed artifacts:
 

scripts/e2e_smoke.py

@@ -1,7 +1,10 @@
 import argparse
+import math
 import os
+import struct
 import sys
 import time
+import wave
 from datetime import datetime, timezone
 
 ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
@@ -20,6 +23,45 @@ def _slug(text):
     return text.lower().replace(" + ", "_plus_").replace(" ", "_")
 
 
+def _ensure_synthetic_sample(audio_path):
+    if os.path.exists(audio_path):
+        return
+
+    if os.path.basename(audio_path) != "public_domain_tone.wav":
+        return
+
+    os.makedirs(os.path.dirname(audio_path), exist_ok=True)
+    sample_rate = 22050
+    segment_sec = 2.0
+    progression = [
+        (261.63, 329.63, 392.00),
+        (293.66, 369.99, 440.00),
+        (349.23, 440.00, 523.25),
+        (392.00, 493.88, 587.33),
+    ]
+
+    frames = []
+    for chord in progression:
+        total = int(segment_sec * sample_rate)
+        for i in range(total):
+            t = i / sample_rate
+            attack = min(1.0, i / (0.05 * sample_rate))
+            release = min(1.0, (total - i) / (0.05 * sample_rate))
+            envelope = attack * release
+            mixed = 0.0
+            for freq in chord:
+                mixed += math.sin(2.0 * math.pi * freq * t)
+            mixed = (mixed / len(chord)) * 0.35 * envelope
+            sample = int(max(-1.0, min(1.0, mixed)) * 32767)
+            frames.append(struct.pack("<h", sample))
+
+    with wave.open(audio_path, "wb") as wf:
+        wf.setnchannels(1)
+        wf.setsampwidth(2)
+        wf.setframerate(sample_rate)
+        wf.writeframes(b"".join(frames))
+
+
 def run_mode(audio_path, extraction_mode, inference_mode, out_dir):
     started = time.time()
     print(f"[start] mode={inference_mode}", flush=True)
@@ -101,6 +143,9 @@ def main():
         if os.path.exists(repo_relative):
             audio_path = repo_relative
 
+    if not os.path.exists(audio_path):
+        _ensure_synthetic_sample(audio_path)
+
     if not os.path.exists(audio_path):
         raise SystemExit(f"Audio file not found: {audio_path}")
 

test_assets/README.md

@@ -1,4 +1,4 @@
 # Test Assets
 
-- `public_domain_tone.wav`: synthetic audio generated in-repo for reproducible smoke tests.
-- Rights: dedicated to the public domain by repository authorship (no third-party copyrighted source audio).
+- `public_domain_tone.wav` is generated on demand by `scripts/e2e_smoke.py` if missing.
+- Rights: generated in-repo from synthesized sine tones (no third-party copyrighted source audio).