MeetMind
An AI meeting assistant that sends a bot to your calls, transcribes them, and turns the transcript into summaries, action items and a searchable chat over everything that was ever said.
Built with
Architecture
A Turborepo monorepo: a Next.js 15 dashboard, a separate Express 5 API that owns the bot, transcription and RAG pipeline, and a shared Prisma package so both sides read the same PostgreSQL schema. Transcripts are chunked, embedded through Hugging Face and indexed in ChromaDB; Groq serves the chat and summarisation passes. Recordings live in S3-compatible storage, and webhooks drive both the note-taker lifecycle and Stripe billing.
Modules
Meetings
A note-taker bot joins the call on schedule, records it, and returns a diarised transcript. Every meeting lands in one library with playback, full transcript and an AI summary.
Ask
RAG chat over your entire meeting history. Transcripts are chunked, embedded with a sentence-transformers model and stored in ChromaDB, so answers cite the meetings they came from.
Semantic Search
Search by meaning rather than keyword — find the call where pricing was renegotiated without remembering who said it or when.
Action Items
The LLM pass pulls owners, tasks and due dates out of the transcript, then pushes them straight into Jira, Trello or Asana through per-user OAuth connections.
Calendar & Scheduling
Connected calendars drive the bot. A scheduler dispatches it ahead of each event's start time, respecting the user's plan allowance and per-meeting opt-outs.
Integrations & Billing
Slack app for posting recaps into channels, OAuth token refresh for every issue tracker, and Stripe-backed plans with usage metering on meetings and chat.