How romeo-eng/meetingmind- works
MeetingMind turns transcript inputs into structured, source-grounded meeting notes and follow-up Q&A through a FastAPI backend and Next.js workspace. Its main path validates and normalizes a source, chunks it, runs local-model extraction and structured parsing, aggregates a meeting record, and returns results to the UI. A separate Q&A path retrieves transcript excerpts before answering; sample analysis, runtime health, and CSV/Markdown exports are additional product capabilities. Edges are limited to relationships shown in the excerpts, SOURCE INDEX references, or README; unsampled frontend interactions are represented only where documented.
Main entry points: service.py Meeting service · routes.py HTTP routes · grounded_qa.py Grounded Q&A · workspace.tsx Workspace UI
Components 20
Meeting workspace
- Workspace UI
frontend/components/workspace.tsx - API client
frontend/lib/api.ts - Runtime status
frontend/components/runtime-status.tsx
API and sessions
- FastAPI app
backend/app/api/main.py - HTTP routes
backend/app/api/routes.py - Typed records
backend/app/models.py - Meeting service
backend/app/pipeline/service.py - Sample meeting
backend/app/pipeline/sample.py - Meeting exports
backend/app/api/routes.py
Transcript analysis
- Transcript loading
backend/app/loaders/transcripts.py - Transcript chunking
backend/app/preprocessing/chunking.py - Chunk extraction
backend/app/chains/prompts.py - Structured parsing
backend/app/parsers/structured.py - Merge and summary
backend/app/pipeline/service.py
Local AI and retrieval
- Mistral runtime
backend/app/chains/local_llm.py - FAISS retrieval
backend/app/embeddings/vector_store.py - Grounded Q&A
backend/app/rag/grounded_qa.py
Other components
- Meeting user
- YouTube captions
- Transcript file
Connections 24
- Meeting user → Workspace UI: submits transcript
README.md - Workspace UI → API client: requests API
- API client → FastAPI app: calls API
- Runtime status → FastAPI app: checks health
README.md - FastAPI app → HTTP routes: registers routes
backend/app/api/main.py - HTTP routes → Transcript loading: loads source
backend/app/api/routes.py - HTTP routes → Meeting service: invokes analysis
backend/app/api/routes.py - HTTP routes → Sample meeting: registers sample
backend/app/api/routes.py - HTTP routes → Meeting exports: streams export
backend/app/api/routes.py - HTTP routes → Typed records: validates responses
backend/app/api/routes.py - YouTube captions → Transcript loading: fetches captions
backend/app/loaders/transcripts.py - Transcript file → Transcript loading: reads file
backend/app/loaders/transcripts.py - Meeting service → Transcript chunking: splits transcript
backend/app/pipeline/service.py - Meeting service → Chunk extraction: builds extraction chains
backend/app/pipeline/service.py - Meeting service → Structured parsing: parses model output
backend/app/pipeline/service.py - Meeting service → Merge and summary: merges results
backend/app/pipeline/service.py - Meeting service → Mistral runtime: loads runtime
backend/app/pipeline/service.py - Meeting service → FAISS retrieval: builds meeting index
backend/app/pipeline/service.py - Meeting service → Grounded Q&A: requests answer
backend/app/pipeline/service.py - FAISS retrieval → Transcript chunking: indexes chunks
backend/app/embeddings/vector_store.py - Grounded Q&A → FAISS retrieval: retrieves chunks
backend/app/rag/grounded_qa.py - Grounded Q&A → Chunk extraction: uses QA prompts
backend/app/chains/prompts.py - Grounded Q&A → Typed records: returns sourced answer
backend/app/rag/grounded_qa.py - Meeting user → Workspace UI: asks meeting question