Deploy v1.39

Dockerfile

@@ -1,6 +1,6 @@
-# [Forge Log #77] Build v1.38 - Pre-Emptive No-Deps Strategy
-# Bypassing crepe setup.py network crawl by using --no-deps during isolation.
-ARG CACHE_BUSTER=20251219143000
+# [Forge Log #78] Build v1.39 - Fix Shell Syntax
+# Quoting version constraints to prevent shell redirection errors.
+ARG CACHE_BUSTER=20251219183000
 
 FROM python:3.8-slim-bullseye
 
@@ -26,7 +26,7 @@ RUN mkdir /models /models_tmp && \
     rm /models/model.zip
 
 RUN pip install --no-cache-dir --upgrade pip setuptools wheel && \
-    pip install --no-cache-dir numpy<1.24 scipy h5py resampy==0.2.2 && \
+    pip install --no-cache-dir "numpy<1.24" scipy h5py resampy==0.2.2 && \
     pip install --no-cache-dir --no-deps crepe==0.0.16
 
 COPY requirements.txt requirements.txt

app.py

@@ -1,7 +1,7 @@
-# [Forge Log #77] The Canary Banner v1.38
-# 'Vanguard Lambda' Protocol - Deep-Core Synthesis Discovery.
-# Exhaustive search for synthesis entry points in complex Keras/Gin proxy chains.
-print("--- [CANARY] DecentSampler DDSP Server v1.38 ---")
+# [Forge Log #78] The Canary Banner v1.39
+# 'Vanguard Mu' Protocol - Deep-Dive Synthesis Discovery.
+# Prioritizes explicit paths found in common DDSP 1.6.5 Gin exports.
+print("--- [CANARY] DecentSampler DDSP Server v1.39 ---")
 print("--- [CANARY] Executing Verified Production App ---")
 
 import os
@@ -34,17 +34,17 @@ import io
 def find_tae():
     try:
         import ddsp.training.models.midi_autoencoder as ma
-        return ma.MidiAutoencoder
+        if hasattr(ma, 'MidiAutoencoder'): return ma.MidiAutoencoder
     except: pass
     try:
         import ddsp.training.models as m
-        return m.TranscribingAutoencoder
+        if hasattr(m, 'TranscribingAutoencoder'): return m.TranscribingAutoencoder
     except: pass
     return None
 
 TAE_CLASS = find_tae()
 
-# --- Deep-Core Synth 'Vanguard' Ladder [Forge Log #77] ---
+# --- Deep-Dive Synth 'Vanguard' Ladder [Forge Log #78] ---
 def find_note_synth():
     target = 'note_expression_synthesis'
     try:
@@ -53,8 +53,8 @@ def find_note_synth():
             return lambda model, pitches, note_lengths, **kwargs: getattr(ma, target)(model, pitches, note_lengths, **kwargs)
     except: pass
 
-    def deep_core_synth_wrapper(model, pitches, note_lengths, **kwargs):
-        """Lambda Protocol: Exhaustive proxy-aware probe."""
+    def deep_dive_synth_wrapper(model, pitches, note_lengths, **kwargs):
+        """Mu Protocol: Explicit proxy mapping and atomic probing."""
         p_tensor = tf.convert_to_tensor(pitches, dtype=tf.float32)
         l_tensor = tf.convert_to_tensor(note_lengths, dtype=tf.float32)
         midi_dicts = [
@@ -62,16 +62,19 @@ def find_note_synth():
             {'midi': {'pitch': p_tensor, 'duration': l_tensor}},
             {'inputs': {'pitch': p_tensor, 'duration': l_tensor}}
         ]
-        methods_to_try = ['midi_to_audio', 'decode', 'synthesize', 'synthesize_audio', 'call', '__call__', 'predict']
-        objs = [model]
-        for l1 in ['ae', 'decoder', 'processor_group', 'nn', 'processor']:
-            if hasattr(model, l1):
-                o1 = getattr(model, l1)
-                objs.append(o1)
-                for l2 in ['decoder', 'processor_group', 'ae', 'processor']:
-                    if hasattr(o1, l2): objs.append(getattr(o1, l2))
-
-        for obj in objs:
+        methods_to_try = ['decode', 'midi_to_audio', 'synthesize', 'synthesize_audio', 'call', '__call__', 'predict']
+        probe_queue = []
+        if model:
+            probe_queue.append(model)
+            for path in ['ae.decoder', 'processor_group', 'ae', 'nn', 'decoder', 'processor']:
+                try:
+                    obj = model
+                    for part in path.split('.'):
+                        obj = getattr(obj, part)
+                    probe_queue.append(obj)
+                except AttributeError: pass
+
+        for obj in probe_queue:
             for name in methods_to_try:
                 method = getattr(obj, name, None)
                 if method and callable(method):
@@ -88,7 +91,7 @@ def find_note_synth():
             for m_dict in midi_dicts:
                 try: return model(m_dict, training=False)
                 except: pass
-        for obj in objs:
+        for obj in probe_queue:
             for attr_name in dir(obj):
                 if attr_name.startswith('_'): continue
                 attr = getattr(obj, attr_name)
@@ -96,9 +99,9 @@ def find_note_synth():
                     for m_dict in midi_dicts:
                         try: return attr(m_dict)
                         except: pass
-        raise AttributeError(f"Vanguard Lambda Failure: Exhausted all synthesis paths.")
+        raise AttributeError(f"Vanguard Mu Failure: Exhausted all synthesis paths.")
 
-    return deep_core_synth_wrapper
+    return deep_dive_synth_wrapper
 
 NOTE_SYNTH_FN = find_note_synth()
 
@@ -140,7 +143,7 @@ def note_to_midi(note_name: str) -> float:
 
 @app.get("/")
 def read_root():
-    return {"message": "DDSP Server Online", "canary": "v1.38"}
+    return {"message": "DDSP Server Online", "canary": "v1.39"}
 
 @app.post("/generate-note")
 async def generate_note(request: NoteRequest):
@@ -158,4 +161,4 @@ async def generate_note(request: NoteRequest):
         wav_io.seek(0)
         return Response(content=wav_io.getvalue(), media_type="audio/wav")
     except Exception as e:
-        raise HTTPException(status_code=500, detail=f"[Vanguard V1.38] {str(e)}")
+        raise HTTPException(status_code=500, detail=f"[Vanguard V1.39] {str(e)}")