How akarshvyas/ai-video-assistant- works
This is an AI video/meeting assistant with a Streamlit interface and a CLI workflow. Users provide a YouTube URL or local media file; the CLI converts and chunks its audio, transcribes it, then generates a title, summary, action items, decisions, and open questions, and builds transcript-grounded chat. The Streamlit app also invokes analysis and chat components. The graph preserves the distinct ingestion, transcription, analysis, and retrieval stages. Source excerpts show the CLI orchestration and key UI calls; omitted UI lines limit evidence for its ingestion and transcription wiring.
Main entry points: main.py CLI pipeline · rag_engine.py RAG question answering · app.py Streamlit workflow · transcriber.py Transcription
Components 16
Media ingestion
- Audio preparation
utils/audio_processor.py - YouTube
Transcription
- Transcription
core/transcriber.py - Whisper model
- Sarvam API
Meeting analysis
- Streamlit workflow
app.py - CLI pipeline
main.py - Title and summary
core/summarizer.py - Meeting insights
core/extractor.py
Transcript chat
- RAG question answering
core/rag_engine.py - Transcript vector store
core/vector_store.py - Mistral service
- Chat response
Other components
- User
- Streamlit app
- CLI user
Connections 17
- User → Streamlit app: uses
- Streamlit app → Streamlit workflow: runs
app.py - CLI user → CLI pipeline: starts
main.py - CLI pipeline → Audio preparation: processes input
main.py - Audio preparation → YouTube: downloads audio
utils/audio_processor.py - CLI pipeline → Transcription: transcribes chunks
main.py - Transcription → Whisper model: transcribes locally
core/transcriber.py - Transcription → Sarvam API: sends audio
core/transcriber.py - CLI pipeline → Title and summary: generates title and summary
main.py - CLI pipeline → Meeting insights: extracts insights
main.py - CLI pipeline → RAG question answering: builds chat chain
main.py - Streamlit workflow → Title and summary: generates title and summary
app.py - Streamlit workflow → Meeting insights: extracts insights
app.py - Streamlit workflow → RAG question answering: builds and queries
app.py - RAG question answering → Transcript vector store: builds and retrieves
core/rag_engine.py - RAG question answering → Mistral service: invokes model
- RAG question answering → Chat response: returns answer
core/rag_engine.py