Fix R2 bucket name: voicebox (not voicebox-releases)

.github/workflows/release.yml

@@ -205,7 +205,7 @@ jobs:
 
           # Upload to R2
           aws s3 cp backend/cuda-release/voicebox-server-cuda-${PLATFORM}.exe \
-            s3://voicebox-releases/cuda/${VERSION}/voicebox-server-cuda-${PLATFORM}.exe \
+            s3://voicebox/cuda/${VERSION}/voicebox-server-cuda-${PLATFORM}.exe \
             --endpoint-url $R2_ENDPOINT \
             --acl public-read
 

backend/test_r2_upload.sh

@@ -48,7 +48,7 @@ echo "Test upload configuration:"
 echo "  Version: $VERSION"
 echo "  Platform: $PLATFORM"
 echo "  Endpoint: $R2_ENDPOINT"
-echo "  Bucket: voicebox-releases"
+echo "  Bucket: voicebox"
 echo "  Path: cuda/$VERSION/$FILENAME"
 echo ""
 
@@ -63,7 +63,7 @@ echo ""
 echo "Uploading to R2..."
 
 aws s3 cp "$CUDA_BINARY" \
-    "s3://voicebox-releases/cuda/${VERSION}/${FILENAME}" \
+    "s3://voicebox/cuda/${VERSION}/${FILENAME}" \
     --endpoint-url "$R2_ENDPOINT" \
     --acl public-read
 

backend/voicebox-server-cuda.spec

@@ -0,0 +1,48 @@
+# -*- mode: python ; coding: utf-8 -*-
+from PyInstaller.utils.hooks import collect_data_files
+from PyInstaller.utils.hooks import collect_submodules
+from PyInstaller.utils.hooks import copy_metadata
+
+datas = []
+hiddenimports = ['backend', 'backend.main', 'backend.config', 'backend.database', 'backend.models', 'backend.profiles', 'backend.history', 'backend.tts', 'backend.transcribe', 'backend.platform_detect', 'backend.backends', 'backend.backends.pytorch_backend', 'backend.utils.audio', 'backend.utils.cache', 'backend.utils.progress', 'backend.utils.hf_progress', 'backend.utils.validation', 'torch', 'transformers', 'fastapi', 'uvicorn', 'sqlalchemy', 'librosa', 'soundfile', 'qwen_tts', 'qwen_tts.inference', 'qwen_tts.inference.qwen3_tts_model', 'qwen_tts.inference.qwen3_tts_tokenizer', 'qwen_tts.core', 'qwen_tts.cli', 'pkg_resources.extern']
+datas += collect_data_files('qwen_tts')
+datas += copy_metadata('qwen-tts')
+hiddenimports += collect_submodules('qwen_tts')
+hiddenimports += collect_submodules('jaraco')
+
+
+a = Analysis(
+    ['server.py'],
+    pathex=[],
+    binaries=[],
+    datas=datas,
+    hiddenimports=hiddenimports,
+    hookspath=[],
+    hooksconfig={},
+    runtime_hooks=[],
+    excludes=[],
+    noarchive=False,
+    optimize=0,
+)
+pyz = PYZ(a.pure)
+
+exe = EXE(
+    pyz,
+    a.scripts,
+    a.binaries,
+    a.datas,
+    [],
+    name='voicebox-server-cuda',
+    debug=False,
+    bootloader_ignore_signals=False,
+    strip=False,
+    upx=True,
+    upx_exclude=[],
+    runtime_tmpdir=None,
+    console=True,
+    disable_windowed_traceback=False,
+    argv_emulation=False,
+    target_arch=None,
+    codesign_identity=None,
+    entitlements_file=None,
+)

backend/voicebox-server.spec

@@ -4,15 +4,11 @@ from PyInstaller.utils.hooks import collect_submodules
 from PyInstaller.utils.hooks import copy_metadata
 
 datas = []
-hiddenimports = ['backend', 'backend.main', 'backend.config', 'backend.database', 'backend.models', 'backend.profiles', 'backend.history', 'backend.tts', 'backend.transcribe', 'backend.platform_detect', 'backend.backends', 'backend.backends.pytorch_backend', 'backend.utils.audio', 'backend.utils.cache', 'backend.utils.progress', 'backend.utils.hf_progress', 'backend.utils.validation', 'torch', 'transformers', 'fastapi', 'uvicorn', 'sqlalchemy', 'librosa', 'soundfile', 'qwen_tts', 'qwen_tts.inference', 'qwen_tts.inference.qwen3_tts_model', 'qwen_tts.inference.qwen3_tts_tokenizer', 'qwen_tts.core', 'qwen_tts.cli', 'pkg_resources.extern', 'backend.backends.mlx_backend', 'mlx', 'mlx.core', 'mlx.nn', 'mlx_audio', 'mlx_audio.tts', 'mlx_audio.stt']
+hiddenimports = ['backend', 'backend.main', 'backend.config', 'backend.database', 'backend.models', 'backend.profiles', 'backend.history', 'backend.tts', 'backend.transcribe', 'backend.platform_detect', 'backend.backends', 'backend.backends.pytorch_backend', 'backend.utils.audio', 'backend.utils.cache', 'backend.utils.progress', 'backend.utils.hf_progress', 'backend.utils.validation', 'torch', 'transformers', 'fastapi', 'uvicorn', 'sqlalchemy', 'librosa', 'soundfile', 'qwen_tts', 'qwen_tts.inference', 'qwen_tts.inference.qwen3_tts_model', 'qwen_tts.inference.qwen3_tts_tokenizer', 'qwen_tts.core', 'qwen_tts.cli', 'pkg_resources.extern']
 datas += collect_data_files('qwen_tts')
-datas += collect_data_files('mlx')
-datas += collect_data_files('mlx_audio')
 datas += copy_metadata('qwen-tts')
 hiddenimports += collect_submodules('qwen_tts')
 hiddenimports += collect_submodules('jaraco')
-hiddenimports += collect_submodules('mlx')
-hiddenimports += collect_submodules('mlx_audio')
 
 
 a = Analysis(

docs/r2-setup.md

@@ -7,7 +7,7 @@ The CUDA binary (2.4GB) is hosted on Cloudflare R2 at `downloads.voicebox.sh` in
 ## R2 Bucket Configuration
 
 ✅ **Completed:**
-- Bucket created: `voicebox-releases`
+- Bucket created: `voicebox`
 - Custom domain configured: `downloads.voicebox.sh`
 
 ## GitHub Secrets Required
@@ -71,7 +71,7 @@ Value: https://<your-account-id>.r2.cloudflarestorage.com
 After CI uploads, the bucket will have this structure:
 
 ```
-voicebox-releases/
+voicebox/
 └── cuda/
     ├── v0.1.12/
     │   └── voicebox-server-cuda-x86_64-pc-windows-msvc.exe
@@ -112,7 +112,7 @@ pip install awscli
 # Test upload (use a small test file first)
 echo "test" > test.txt
 aws s3 cp test.txt \
-  s3://voicebox-releases/test/test.txt \
+  s3://voicebox/test/test.txt \
   --endpoint-url $R2_ENDPOINT \
   --acl public-read
 
@@ -121,7 +121,7 @@ curl https://downloads.voicebox.sh/test/test.txt
 
 # If successful, try the actual CUDA binary
 aws s3 cp backend/dist/voicebox-server-cuda.exe \
-  s3://voicebox-releases/cuda/v0.1.12-test/voicebox-server-cuda-x86_64-pc-windows-msvc.exe \
+  s3://voicebox/cuda/v0.1.12-test/voicebox-server-cuda-x86_64-pc-windows-msvc.exe \
   --endpoint-url $R2_ENDPOINT \
   --acl public-read
 ```
@@ -163,7 +163,7 @@ The workflow now:
     R2_ENDPOINT: ${{ secrets.R2_ENDPOINT }}
   run: |
     aws s3 cp backend/cuda-release/voicebox-server-cuda-*.exe \
-      s3://voicebox-releases/cuda/${VERSION}/... \
+      s3://voicebox/cuda/${VERSION}/... \
       --endpoint-url $R2_ENDPOINT \
       --acl public-read
 ```
@@ -172,7 +172,7 @@ The workflow now:
 
 Monitor your R2 usage:
 
-**Cloudflare Dashboard → R2 → voicebox-releases → Metrics**
+**Cloudflare Dashboard → R2 → voicebox → Metrics**
 
 Expected costs (per month):
 - Storage: 2.4GB × $0.015/GB = **$0.036**
@@ -219,7 +219,7 @@ Or set bucket default permissions in R2 Dashboard.
 ✅ **Recommended:**
 - Object Read & Write only
 - No admin permissions needed
-- Scoped to `voicebox-releases` bucket only
+- Scoped to `voicebox` bucket only
 
 ❌ **Avoid:**
 - Account-wide permissions
@@ -242,11 +242,11 @@ Optional: Delete old CUDA binaries to save storage costs
 
 ```bash
 # List all versions
-aws s3 ls s3://voicebox-releases/cuda/ \
+aws s3 ls s3://voicebox/cuda/ \
   --endpoint-url $R2_ENDPOINT
 
 # Delete old version
-aws s3 rm s3://voicebox-releases/cuda/v0.1.0/ \
+aws s3 rm s3://voicebox/cuda/v0.1.0/ \
   --recursive \
   --endpoint-url $R2_ENDPOINT
 ```

tauri/src-tauri/Cargo.lock

@@ -5041,7 +5041,7 @@ checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
 
 [[package]]
 name = "voicebox"
-version = "0.1.11"
+version = "0.1.12"
 dependencies = [
  "base64 0.22.1",
  "core-foundation-sys",

test_cuda.py

@@ -0,0 +1,77 @@
+"""Test CUDA detection in voicebox backend"""
+import sys
+import torch
+
+print("=" * 60)
+print("PyTorch CUDA Detection Test")
+print("=" * 60)
+
+# Basic torch info
+print(f"\nPyTorch version: {torch.__version__}")
+print(f"CUDA available: {torch.cuda.is_available()}")
+
+if torch.cuda.is_available():
+    print(f"CUDA version: {torch.version.cuda}")
+    print(f"GPU count: {torch.cuda.device_count()}")
+    print(f"Current GPU: {torch.cuda.current_device()}")
+    print(f"GPU name: {torch.cuda.get_device_name(0)}")
+    print(f"GPU memory: {torch.cuda.get_device_properties(0).total_memory / 1024**3:.2f} GB")
+else:
+    print("\nNo CUDA available - would run on CPU")
+
+# Test backend device selection
+print("\n" + "=" * 60)
+print("Backend Device Selection")
+print("=" * 60)
+
+# Simulate the _get_device method from pytorch_backend.py
+def _get_device() -> str:
+    """Get the best available device."""
+    if torch.cuda.is_available():
+        return "cuda"
+    elif hasattr(torch.backends, 'mps') and torch.backends.mps.is_available():
+        # MPS can have issues, use CPU for stability
+        return "cpu"
+    return "cpu"
+
+selected_device = _get_device()
+print(f"\nSelected device: {selected_device}")
+print(f"Would use dtype: {'torch.bfloat16' if selected_device != 'cpu' else 'torch.float32'}")
+
+# Test actual tensor creation on device
+print("\n" + "=" * 60)
+print("Testing Tensor Creation on Device")
+print("=" * 60)
+
+try:
+    test_tensor = torch.randn(1000, 1000).to(selected_device)
+    print(f"\n[OK] Successfully created tensor on {selected_device}")
+    print(f"  Tensor device: {test_tensor.device}")
+    print(f"  Tensor dtype: {test_tensor.dtype}")
+
+    # Test computation
+    result = test_tensor @ test_tensor.T
+    print(f"[OK] Successfully performed computation on {selected_device}")
+
+    if selected_device == "cuda":
+        print(f"\nCUDA memory allocated: {torch.cuda.memory_allocated() / 1024**2:.2f} MB")
+        print(f"CUDA memory reserved: {torch.cuda.memory_reserved() / 1024**2:.2f} MB")
+
+except Exception as e:
+    print(f"\n[ERROR] {e}")
+
+print("\n" + "=" * 60)
+print("Summary")
+print("=" * 60)
+
+if selected_device == "cuda":
+    print("\n[SUCCESS] CUDA IS WORKING!")
+    print("  The backend will use your NVIDIA GPU for inference")
+    print(f"  GPU: {torch.cuda.get_device_name(0)}")
+    print(f"  This will be significantly faster than CPU")
+else:
+    print("\n[FAIL] CUDA is not available")
+    print("  The backend will use CPU for inference")
+    print("  This will be slower than GPU")
+
+print("\n" + "=" * 60)