How srinivas26k/zoom_poll_demo works
Project type and purpose
• This is a full-stack Python application (“Zoom Poll Automator”) combining:
• A Flask-based web interface (setup wizard and status pages)
• A CLI/daemon for continuous audio polling
• A modular “backend” that captures audio, transcribes it with Whisper, generates polls via LLaMA, and posts them to Zoom
• Purpose: Keep Zoom meetings engaging by automatically generating and posting context-aware polls in real time.
File structure
• Top-level scripts:
• app.py – Flask web server
• run.py / run_loop.py / cli.py – CLI entry-points for automated polling
• audio_capture.py – Captures live Zoom audio
• transcribe_whisper.py – Wrapper around OpenAI Whisper transcription
• poll_prompt.py – Constructs prompt for LLaMA
• poller.py – Orchestrates the end-to-end flow (capture → transcribe → prompt → Zoom API)
• config.py – Loads .env and centralizes configuration
• Assets & templates:
• templates/*.html – Frontend pages (setup, status, error)
• Tests & CI:
• tests/ – Unit, integration, performance, security tests
• pytest.ini, test-reports/ – Test configuration and output
• Setup & deployment:
• setup.sh/.bat, start.sh/.ps1/.bat – Installation and launch scripts
• requirements.txt, setup.py – Packaging and dependencies
Key insights from the README
• Core services: Zoom API, OpenAI Whisper, Ollama (LLaMA)
• Uses OAuth2 for Zoom integration, FFmpeg for audio capture
• Two user interfaces: Web (Flask) and CLI
• Zoom Meeting (Audio Stream)
• Audio Capture Module
• Transcription Service (OpenAI Whisper)
• Poll Generation Engine (LLaMA via Ollama)
• Zoom API Integration
• Configuration Store (.env via config.py)
• Web UI (Flask app.py + templates)
• CLI/Daemon (run.py, run_loop.py, cli.py)
• Logging & Error Handling
1. Zoom Meeting → Audio Capture
2. Audio Capture → Transcription Service
3. Transcription → Poll Generation
4. Generated Poll → Zoom API
5. Zoom API → Meeting (poll posted)
6. Web UI ↔ Configuration Store ↔ OAuth callback
7. CLI/Daemon invokes Poller (same backend modules)
8. All modules log to central error/log handler
c. Architectural patterns
• Modular “service-oriented monolith”: each step in its own module
• MVC-style separation in the Flask UI (templates vs. app logic)
• Single-responsibility: capture, transcribe, prompt, post are separate
d. Technologies & frameworks
• Python 3.8+
• Flask for web
• OpenAI Whisper (transcribe_whisper.py)
• Ollama/LLaMA for poll prompts via poll_prompt.py
• Zoom API (OAuth2 + REST calls in poller.py)
• FFmpeg for raw audio capture
• Pytest for testing