v0.2.0: comprehensive codebase review fixes

.gitignore

@@ -40,6 +40,7 @@ src-tauri/sidecar/__pycache__/
 src-tauri/sidecar/dist/
 src-tauri/sidecar/build/
 src-tauri/sidecar/.venv/
+src-tauri/sidecar/data/
 .sidecar-venv/
 
 # Superpowers

CHANGELOG.md

@@ -1,5 +1,38 @@
 # Changelog
 
+## [0.2.0] - 2026-04-08
+
+**57 files | 3,086 additions | 608 deletions | 3 PRs | 29 commits**
+
+### 🌟 New Features
+
+- **Model Settings panel** — dedicated sidebar page to browse, download, and manage AI models with category tags, recommended badges, and Whisper model selection (#21)
+- **Video rename** — rename recordings directly from the library view (#20)
+- **Auto-detect `ELEVENLABS_API_KEY`** from environment variables — no manual entry needed if already exported (#20)
+- **Version bump tooling** — `bump.sh` script for consistent release versioning
+
+### 🐛 Bug Fixes
+
+- **Slow/deep voice on first generation** after switching voices — TTS engine now reinitializes properly on voice change
+- **Video clipping** in both TTS and voice conversion modes — output duration now matches source (#20)
+- **TTS timing and transcription accuracy** improvements — better duration probing and segment alignment (#20)
+- **Model cache not refreshing** after download completes — cache invalidation added (#21)
+- **CosyVoice3 missing** from prerequisites UI and download JSON parsing error (#20)
+- **Sidecar code signing** added to PyInstaller build for macOS Gatekeeper (#20)
+- **84 findings** from comprehensive codebase review resolved — spanning error handling, edge cases, and code quality
+- **22 model-related bugs** fixed in consolidated review (#21)
+- **`whisper-small` removed** — too low quality for production use (#21)
+- **Models nav alignment** fixed in sidebar (#21)
+- **`yt-dlp` options** corrected for voice extraction
+
+### 🛠️ Other Improvements
+
+- Models page moved from settings sub-tab to dedicated sidebar route with `ModelInfo` struct enriched with `category` and `recommended` fields (#21)
+- `cleanup_stale_recordings` test made deterministic (#21)
+- README updated: Linux references removed (macOS-only app), visual architecture diagram replaces ASCII, TTS audio cutoff added to known issues (#19, #20)
+- Integration tests updated for CosyVoice3 three-model setup (#20)
+- Unused CI workflow files cleaned up (#20)
+
 ## [0.1.0] - 2026-04-02
 
 **86 files | 12,367 lines | 18 PRs | 99 commits**

README.md

@@ -368,8 +368,8 @@ These are the IPC commands exposed by the Rust backend, invoked from the fronten
 | `sidecar_fetch` | Proxy a GET/POST request to the sidecar |
 | `sidecar_upload` | Proxy a file upload to the sidecar |
 | `get_sidecar_status` | Get sidecar process status |
-| `check_models_downloaded` | Check if required models are available |
 | `download_model` | Download a model from HuggingFace |
+| `delete_model` | Delete a downloaded model from the cache |
 | `get_models_disk_usage` | Get disk usage of downloaded models |
 | `google_drive_connect` | Start Google Drive OAuth2 flow |
 | `google_drive_disconnect` | Disconnect Google Drive |
@@ -404,6 +404,7 @@ HTTP endpoints served by the Python FastAPI sidecar on localhost.
 | `/profiles/{id}` | DELETE | Delete a voice profile |
 | `/models/status` | GET | Get model download/load status |
 | `/models/download` | POST | Download a model (streams progress as SSE) |
+| `/models/{model_name}` | DELETE | Delete a downloaded model |
 
 ## Debugging
 

docs/BUILD.md

@@ -129,6 +129,6 @@ cd src-tauri/sidecar && pyinstaller voiceover-tts.spec
 
 ## Notes
 
-- **Unsigned app**: The .dmg is not code-signed. Users will see a macOS Gatekeeper warning on first launch. To open: right-click the app → "Open" → click "Open" in the dialog. This only needs to be done once.
+- **Code signing**: The build is configured with `signingIdentity: "Developer ID Application"` and hardened runtime. If a valid Developer ID certificate is in your keychain, the app will be signed automatically. Without the certificate, the app will be unsigned and users will see a macOS Gatekeeper warning on first launch (right-click → "Open" → "Open").
 - **ffmpeg is bundled**: The static ffmpeg binary is included inside the .app so users don't need to install it separately. It's ~60MB and excluded from git via `.gitignore`.
 - **Settings location**: App config is stored at `~/Library/Application Support/com.voiceover.app/config.json`, not inside the .app bundle.

package.json

@@ -1,7 +1,7 @@
 {
 	"name": "voiceover",
 	"private": true,
-	"version": "0.1.0",
+	"version": "0.2.0",
 	"type": "module",
 	"scripts": {
 		"dev": "vite dev",
@@ -19,14 +19,13 @@
 		"test:all:coverage": "pnpm test:coverage && pnpm test:rust:coverage"
 	},
 	"devDependencies": {
-		"@sveltejs/adapter-auto": "^7.0.0",
 		"@sveltejs/adapter-static": "^3.0.10",
 		"@sveltejs/kit": "^2.50.2",
 		"@sveltejs/vite-plugin-svelte": "^6.2.4",
 		"@tauri-apps/cli": "^2.10.1",
 		"@testing-library/jest-dom": "^6.9.1",
 		"@testing-library/svelte": "^5.3.1",
-		"@types/node": "^25.5.0",
+		"@types/node": "^22.0.0",
 		"@vitest/coverage-v8": "^4.1.0",
 		"jsdom": "^29.0.1",
 		"lefthook": "^2.1.4",

scripts/bump-version.sh

@@ -0,0 +1,40 @@
+#!/usr/bin/env bash
+# Bump version across all three config files simultaneously.
+# Usage: ./scripts/bump-version.sh 0.2.0
+
+set -euo pipefail
+
+if [ $# -ne 1 ]; then
+    echo "Usage: $0 <version>"
+    echo "Example: $0 0.2.0"
+    exit 1
+fi
+
+VERSION="$1"
+
+# Validate semver format
+if ! echo "$VERSION" | grep -qE '^[0-9]+\.[0-9]+\.[0-9]+(-[a-zA-Z0-9.]+)?$'; then
+    echo "Error: '$VERSION' is not a valid semver (expected X.Y.Z or X.Y.Z-pre.N)"
+    exit 1
+fi
+
+ROOT="$(cd "$(dirname "$0")/.." && pwd)"
+
+# 1. package.json
+sed -i '' "s/\"version\": \"[^\"]*\"/\"version\": \"$VERSION\"/" "$ROOT/package.json"
+
+# 2. Cargo.toml (only the package version, line 3)
+sed -i '' "3s/version = \"[^\"]*\"/version = \"$VERSION\"/" "$ROOT/src-tauri/Cargo.toml"
+
+# 3. tauri.conf.json
+sed -i '' "s/\"version\": \"[^\"]*\"/\"version\": \"$VERSION\"/" "$ROOT/src-tauri/tauri.conf.json"
+
+echo "Bumped to $VERSION in:"
+echo "  package.json"
+echo "  src-tauri/Cargo.toml"
+echo "  src-tauri/tauri.conf.json"
+
+# Show the changes
+grep -n "\"version\"" "$ROOT/package.json" | head -1
+grep -n "^version" "$ROOT/src-tauri/Cargo.toml" | head -1
+grep -n "\"version\"" "$ROOT/src-tauri/tauri.conf.json" | head -1

src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "voiceover"
-version = "0.1.0"
+version = "0.2.0"
 description = "Screen recorder with ElevenLabs voice replacement"
 authors = ["dave"]
 license = "MIT"
@@ -28,10 +28,10 @@ tauri-plugin-dialog = "2"
 tauri-plugin-fs = "2"
 reqwest = { version = "0.12", features = ["multipart", "json", "stream"] }
 futures-util = "0.3"
-tokio = { version = "1", features = ["full"] }
+tokio = { version = "1", features = ["rt-multi-thread", "macros", "time", "process", "sync", "fs"] }
 dirs = "6"
 tauri-plugin-global-shortcut = "2"
-rand = "0.8"
+rand = "0.9"
 sha2 = "0.10"
 base64 = "0.22"
 url = "2"

src-tauri/entitlements.plist

@@ -2,16 +2,22 @@
 <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
 <plist version="1.0">
 <dict>
+    <!-- Required for MLX/PyTorch JIT compilation on Apple Silicon -->
     <key>com.apple.security.cs.allow-unsigned-executable-memory</key>
     <true/>
+    <!-- Required for MLX Metal compute kernels -->
     <key>com.apple.security.cs.allow-jit</key>
     <true/>
+    <!-- Required for PyInstaller frozen binaries (sidecar) -->
     <key>com.apple.security.cs.allow-dyld-environment-variables</key>
     <true/>
+    <!-- Required for microphone recording -->
     <key>com.apple.security.device.audio-input</key>
     <true/>
+    <!-- Required for webcam overlay -->
     <key>com.apple.security.device.camera</key>
     <true/>
+    <!-- Required for Python sidecar's dynamic .so loading -->
     <key>com.apple.security.cs.disable-library-validation</key>
     <true/>
 </dict>

src-tauri/sidecar/chunked_tts.py

@@ -9,9 +9,10 @@
 Short text (≤ max_chunk_chars) uses the single-shot fast path with zero overhead.
 """
 
+from __future__ import annotations
+
 import logging
 import re
-from typing import List, Tuple
 
 import numpy as np
 
@@ -30,15 +31,15 @@
 _PARA_TAG_RE = re.compile(r"\[[^\]]*\]")
 
 
-def split_text_into_chunks(text: str, max_chars: int = DEFAULT_MAX_CHUNK_CHARS) -> List[str]:
+def split_text_into_chunks(text: str, max_chars: int = DEFAULT_MAX_CHUNK_CHARS) -> list[str]:
     """Split text at natural boundaries into chunks of at most max_chars."""
     text = text.strip()
     if not text:
         return []
     if len(text) <= max_chars:
         return [text]
 
-    chunks: List[str] = []
+    chunks: list[str] = []
     remaining = text
 
     while remaining:
@@ -115,7 +116,7 @@ def _safe_hard_cut(segment: str, max_chars: int) -> int:
 
 
 def concatenate_audio_chunks(
-    chunks: List[np.ndarray],
+    chunks: list[np.ndarray],
     sample_rate: int,
     crossfade_ms: int = 50,
 ) -> np.ndarray:
@@ -282,13 +283,35 @@ def pad_audio_to_match_timing(
     return np.concatenate(result_pieces)
 
 
+# ---------------------------------------------------------------------------
+# Pad / truncate to target duration
+# ---------------------------------------------------------------------------
+
+
+def pad_or_truncate_to_duration(audio: np.ndarray, sample_rate: int, target_duration: float, fade_ms: float = 300) -> np.ndarray:
+    """Pad (with fade-out) or truncate audio to match target duration."""
+    if target_duration <= 0 or not sample_rate:
+        return audio
+    target_samples = int(target_duration * sample_rate)
+    if len(audio) < target_samples:
+        fade_samples = min(int(fade_ms / 1000 * sample_rate), len(audio) // 2)
+        if fade_samples > 0:
+            fade = np.linspace(1.0, 0.0, fade_samples, dtype=np.float32)
+            audio = audio.copy()
+            audio[-fade_samples:] *= fade
+        audio = np.pad(audio, (0, target_samples - len(audio)))
+    elif len(audio) > target_samples:
+        audio = audio[:target_samples]
+    return audio
+
+
 # ---------------------------------------------------------------------------
 # Timestamp-aware audio assembly
 # ---------------------------------------------------------------------------
 
 
 def assemble_timed_segments(
-    tts_segments: List[Tuple[np.ndarray, float, float]],
+    tts_segments: list[tuple[np.ndarray, float, float]],
     original_duration: float,
     sample_rate: int,
 ) -> np.ndarray:

src-tauri/sidecar/entitlements.plist

@@ -2,12 +2,16 @@
 <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
 <plist version="1.0">
 <dict>
+    <!-- Required for NumPy/PyTorch JIT on Apple Silicon -->
     <key>com.apple.security.cs.allow-unsigned-executable-memory</key>
     <true/>
+    <!-- Required for MLX Metal compute kernels -->
     <key>com.apple.security.cs.allow-jit</key>
     <true/>
+    <!-- Required for PyInstaller frozen binary loader -->
     <key>com.apple.security.cs.allow-dyld-environment-variables</key>
     <true/>
+    <!-- Required for loading Python extension .so modules -->
     <key>com.apple.security.cs.disable-library-validation</key>
     <true/>
 </dict>

src-tauri/sidecar/profiles.py

@@ -6,17 +6,23 @@
 - samples/{uuid}.txt: transcript for each audio file
 """
 
+from __future__ import annotations
+
 import json
 import logging
 import re
 import shutil
 import uuid
 from datetime import datetime, timezone
 from pathlib import Path
-from typing import Optional
 
 logger = logging.getLogger("voiceover-tts.profiles")
 
+__all__ = [
+    "list_profiles", "get_profile", "create_profile", "delete_profile",
+    "add_sample", "get_samples",
+]
+
 
 def _validate_profile_id(profile_id: str) -> None:
     """Validate profile_id is safe for use as a path component."""
@@ -45,7 +51,7 @@ def list_profiles(data_dir: Path) -> list[dict]:
     return profiles
 
 
-def get_profile(data_dir: Path, profile_id: str) -> Optional[dict]:
+def get_profile(data_dir: Path, profile_id: str) -> dict | None:
     """Get a single profile by ID."""
     _validate_profile_id(profile_id)
     meta_file = _profiles_dir(data_dir) / profile_id / "profile.json"