Overview
This project focuses on reliability and structure. Instead of a generic summary, it extracts meeting artifacts that are directly useful for teams, especially action items and decisions.
LangGraphLLMStructured OutputAutomation
AI / LLM Project
Meeting Minutes Generator
A LangGraph-based workflow that converts meeting transcripts (or audio/video via transcription) into structured minutes: participants, agenda, key topics, action items, minutes and an executive summary.
GitHub RepoLangGraph workflow
Key outputs
- Participants and roles when available.
- Agenda and meeting structure.
- Topics and key discussion points.
- Action items with owner and deadline when present.
- Minutes and executive summary.
Workflow
- Input: transcript (paste text) or audio/video (transcribe first).
- LangGraph orchestration: step-by-step extraction nodes.
- Validation: enforce schema and formatting for consistent outputs.
- Export: results ready for docs, tickets or follow-up emails.
Why LangGraph
- Modular nodes allow easy iteration (for example adding decisions or risks).
- Stateful workflow ensures predictable order and intermediate checkpoints.
- Provides better control compared to a single long prompt.
Next improvements
- Web UI (Streamlit or FastAPI) for a public demo.
- Calendar and ticket integrations (Jira / Trello) from action items.
- Speaker diarization support to improve participant extraction.
- Cost and latency optimizations (chunking and caching).