Skip to content

Latest commit

 

History

History
53 lines (32 loc) · 2.46 KB

File metadata and controls

53 lines (32 loc) · 2.46 KB

Hey Vexa 👋

License: Apache 2.0

Join Discord

Vexa Main Project Repo Parent repo stars

The first real‑time meeting co‑pilot that listens, thinks, and speaks to actively assist your Google Meet / Zoom / MS Teams calls.


Most "AI note‑takers" just sit in a call and dump a doc later. Hey Vexa is the first open‑source bot that actively joins the conversation. Say "Hey Vexa, research this problem and come back with the report" (or any custom command) and the agent will

1. Invoke an agent with access to your favorite MVP tools (previous meetings, RAG, company docs, web search, etc)

2. Response with a natural voice that answers back into the meeting — instantly.

Built on top of the Vexa real‑time transcription API, which only requires the addition of an extra endpoint. It adds just one missing piece: a lightweight playback endpoint so the bot can talk.


🗺️ Project Roadmap

Phase 0 – Vexa Core Upgrade

effort: light

  • update @vexa-bot to playback sound, obtained through Redis MQ so an external service can stream audio/wav back into the call.
  • Extend @bot-manager with /playback endpoint to queue the audio playback jobs via Redis

MVP 1 – Hey Vexa MVP

effort: easy

  1. Transcript Listener – poll Vexa's real‑time caption stream.
  2. Wake‑word Detector – regex for "Hey, Vexa" triggers an agent.
  3. Simple LLM one pass response
  4. TTS speech generation
  5. Agent Invocation – send recent context → LLM → TTS → POST /playback.

Product – MCP‑Powered Agent

effort: should be easy

  • Replace LLM call with the MCP-enabled agent

Join Discord

🤝 Call for Collaboration

We're looking for backend hackers, prompt engineers, TTS/LLM enthusiasts, DevRel storytellers, and testers to help build this in the open. 👉 Hop into the Discord, pick an issue, and say "Hey Vexa, I'm in!"