How rr28yosizumi/ai_meeting_recorder works
Project type and purpose
• This is a desktop “full-stack” (single-executable) meeting-recorder application with a GUI front end, local audio capture & processing, Whisper-based transcription, and Google Gemini API-based summarization.
File structure
• Top level:
• ai_meeting_recorder.py / src/main.py → entry points
• setup.py, pyproject.toml, MANIFEST.in → packaging/build
• src/ → all application modules
• Inside src/:
• view.py, gui_control.py, controller.py → GUI and MVC controller
• model.py, ai_control.py, sound_control.py → core business logic
• resource_util.py, resource/ → static assets (logo) and I/O helpers
• setting.py → load/save YAML config
• main.py → application bootstrap
• No microservices—everything runs in one process, with background threads for recording and AI tasks.
README insights
• Uses sounddevice + FFMpeg for audio capture/mixing
• Uses Whisper (local) and Google Gemini API (remote)
• Data flow: user config → record WAV → (optionally) transcribe → summarize → save text
• Persistent state: init.yml (YAML)
• Multi-threaded operations for UI responsiveness
Main components
a. Frontend (GUI layer)
• view.py: Tkinter windows, splash, main UI
• gui_control.py: binds UI events to controller
b. Application Controller
• controller.py: orchestrates user commands, spawns threads
c. Core Logic / Model
• model.py: data structures (RecordingSession, Transcript, Summary)
• sound_control.py: handles audio capture via sounddevice, FFMpeg mix
• ai_control.py: calls Whisper library and Google-GenerativeAI SDK
d. Persistence & Resources
• setting.py: load/save init.yml
• resource_util.py + src/resource/: manage logos and static files
e. External Services & Libraries
• Sounddevice / FFMpeg (audio I/O)
• OpenAI Whisper (local transcription)
• Google Gemini API (remote summarization)
f. File System
• WAV output folder
• Text summary output folder
• init.yml (config)
Relationships & interactions
• User clicks “Record” in GUI → gui_control → controller
• controller → spawns RecordingThread (sound_control) → writes WAV
• User clicks “Transcribe & Summarize” → controller → spawns AIThread
• AIThread → ai_control.transcribe(wav) → Whisper
• AIThread → ai_control.summarize(text) → Gemini API
• On complete → controller → view shows/save dialog
• Persistence: setting.py saves/loads on startup/shutdown
Architectural patterns & design principles
• MVC: view / gui_control (View + Presenter) → controller (Controller) → model & services
• Single process with worker threads for I/O-bound tasks
• Dependency injection via module boundaries (e.g., ai_control hides API details)
• Configuration via YAML, not hard-coded
Elements
• Blue for internal modules
• Green for external services/APIs
• Yellow for file system / config files
• Threads: RecordingThread, AIThread
Detailed capture of essential elements
• Data flow end-to-end: User → GUI → Controller → Audio/AI → File System → User