feat: add minimal AI processing support (#133)

* transcoder: enable copy mode only the 'copy' is selected

* AI Processing Support: Real-ESRGAN model powered by BMF

- Add enahnce_module.py (Real-ESRGAN), support GPU acceleration.

- Both of GUI and CLI support call this AI feature.

**Python Runtime Support:**
- Add PythonManager for automatic Python environment setup
- Add PythonInstallDialog for user-friendly installation UI
- install Python runtime with PyTorch, Real-ESRGAN, OpenCV, NumPy
- Download Python in App Path

**BMF Integration:**
- Add runtime BMF library path detection
- Set BMF_MODULE_CONFIG_PATH and PYTHONPATH automatically
- Support both system BMF (Debug) and bundled BMF (Release)

* ui: only compile AI related component when bmf and gui enabled

* ai_processing_page: only set codec when user specific

* cd: use macos-14 runner for compatibility

enable bmf on linux

* cd: specific the python 3.9 for mac build

* cd: add linux arm build

* python_manager: set PATH and *_LIBRARY_PATH for python

* force require App Python for linux platform(release or not)

* cd: change the start entry to OpenConverter from AppRun

This avoid wrong App Data Path by Qt.

* transcoder_bmf: get module path from BMF_MODULE_PATH env

* cd: add libavdevices.so.59 for linux build

* cd: add build-linglong workflow job

add linglong.yaml and dynamically build default.desktop

* cd: get ll cache from docker

* python_manager: install cpu-only support torch

* cd: bundle the weight file for linux build

* python_manager: install bmf package from github for linux build

* cd: dynamically build the desktop file for ll

* cd: use defualt icon for ll build

* python_install_dialog: require user to restart app after python support installed

* ui: dynamically build the page list

Enable ai_processing page only if bmf is enabled

* cd: merge review.yaml and release.yaml into build.yaml

---------

Signed-off-by: Jack Lau <[email protected]>
Signed-off-by: Jack Lau <[email protected]>

Author: JackLau1222

.augment-guidelines

@@ -254,21 +254,110 @@ All source files (doesn't inlcude FFmpeg code) should include Apache 2.0 license
 
 ### Building the Project
 
+OpenConverter supports two build modes:
+
+#### **Debug Mode (Development)**
+Fast builds for development, uses system libraries:
+
 ```bash
 # Create build directory
 mkdir build && cd build
 
-# Configure with CMake
-cmake ../src -DENABLE_GUI=ON
+# Configure (Debug is default)
+cmake ../src -DENABLE_GUI=ON -DBMF_TRANSCODER=ON
 
 # Build
 make -j4
 
 # Run
-./OpenConverter              # GUI mode
-./OpenConverter --help       # CLI mode
+./OpenConverter.app/Contents/MacOS/OpenConverter  # macOS
+./OpenConverter                                    # Linux
+```
+
+**What's bundled:**
+- ✅ Python modules (`enhance_module.py`)
+- ✅ AI model weights (2.4 MB)
+- ❌ BMF libraries (uses system BMF from CMake path)
+- ❌ Qt frameworks (uses system Qt)
+- ❌ Python runtime (uses system Python)
+
+**Requirements:**
+- Python 3.9+ with PyTorch, Real-ESRGAN, OpenCV, NumPy
+- BMF framework (path configured in CMake)
+- Qt 5.15+ (for GUI)
+
+**Environment:** No environment variables needed! BMF path is detected from CMake configuration.
+
+#### **Release Mode (Distribution)**
+Standalone builds for distribution, bundles everything:
+
+```bash
+# Create separate build directory
+mkdir build-release && cd build-release
+
+# Configure with Release mode
+cmake ../src -DCMAKE_BUILD_TYPE=Release -DENABLE_GUI=ON -DBMF_TRANSCODER=ON
+
+# Build (takes longer due to bundling)
+make -j4
+
+# Run - NO environment variables needed!
+open OpenConverter.app  # macOS
+./OpenConverter         # Linux
 ```
 
+**What's bundled:**
+- ✅ Python modules
+- ✅ AI model weights (2.4 MB)
+- ✅ BMF libraries (25+ libraries, ~30 MB)
+- ✅ Qt frameworks (10+ frameworks, ~150 MB)
+- ✅ Python runtime with PyTorch, Real-ESRGAN, OpenCV, NumPy (~30 MB)
+- ✅ FFmpeg libraries (~25 MB)
+
+**Total Size:** ~800 MB - 1.2 GB (fully standalone)
+
+**Advantages:**
+- ✅ Fully standalone (no dependencies)
+- ✅ Works on any Mac (macOS 11+)
+- ✅ No environment variables needed
+- ✅ Ready for distribution (zip/DMG)
+
+### Library Bundling (macOS Release Mode)
+
+**Important:** All library bundling (Qt, FFmpeg, BMF) is handled by **`tool/fix_macos_libs.sh`**, NOT by CMake.
+
+#### **Workflow**
+
+1. **Build**: `cmake -B build-release -DCMAKE_BUILD_TYPE=Release && cd build-release && make -j4`
+2. **Bundle**: `cd .. && tool/fix_macos_libs.sh`
+
+The script auto-detects build directory, bundles Qt/FFmpeg/BMF libraries, fixes paths, and code signs.
+
+#### **BMF Library Structure**
+
+```
+OpenConverter.app/Contents/
+├── Frameworks/
+│   ├── lib/                      # Builtin modules (BMF hardcoded path)
+│   │   ├── libbuiltin_modules.dylib
+│   │   ├── libcopy_module.dylib
+│   │   └── libcvtcolor.dylib
+│   ├── libbmf_module_sdk.dylib   # Core BMF libraries
+│   ├── libhmp.dylib
+│   ├── _bmf.cpython-39-darwin.so
+│   └── BUILTIN_CONFIG.json
+└── Resources/bmf_python/         # BMF Python package
+```
+
+**Why `Frameworks/lib/`?** BMF expects builtin modules in `lib/` subdirectory relative to `BMF_MODULE_CONFIG_PATH`.
+
+**Script Logic:**
+- Auto-appends `/output/bmf` to `$BMF_ROOT_PATH` if needed (same as CMake)
+- Copies builtin modules to `Frameworks/lib/`
+- Copies core libraries to `Frameworks/`
+- Processes both `.dylib` and `.so` files
+- Fixes all paths to use `@executable_path`
+
 ### Translation Workflow
 
 OpenConverter uses Qt Linguist tools for internationalization (i18n). Translation files are located in `src/resources/`.