Introduction

Deploy AI-powered speaking agents in video meetings

This small open-source API demonstrates the capabilities of MeetingBaas 🐟's video meeting APIs by integrating with Pipecat's Python framework for building voice and multimodal conversational agents:

curl -X POST https://speaking.meetingbaas.com/bots \
  -H "Content-Type: application/json" \
  -d '{
    "meeting_url": "https://us06web.zoom.us/j/123456789?pwd=example",
    "personas": ["baas_onboarder"],
    "meeting_baas_api_key": "your-api-key"
  }'

Speaking Meeting Bot

Clone to customize and deploy your own meeting agents

Our implementation creates AI meeting agents that can join and participate in Google Meet and Microsoft Teams meetings with distinct personalities and context defined in Markdown files.

It extends Pipecat's WebSocket server implementation to create:

  • Meeting agents that can join Google Meet, Zoom or Microsoft Teams through the MeetingBaas API
  • Customizable personas with unique context
  • Support for running multiple instances locally or at scale

This api is launched using Docker and fly.io.

API Access

The Speaking Bot API is accessible at speaking.meetingbaas.com. You can access the OpenAPI specification directly for your LLM here: speaking.meetingbaas.com/openapi.json

The API follows a minimalist design with sensible defaults while offering optional customization. A bot can be deployed with just a meeting URL and API key, but parameters are available for tailoring behavior:

The API includes features not explicitly defined in the routes documentation:

  • WebSocket infrastructure for bidirectional audio streaming with selectable quality (16/24kHz)
  • Persona system with custom voice selection, language preferences, and contextual knowledge
  • Voice Activity Detection with configurable parameters for natural conversation
  • Function calling tools (weather, time, etc.) that can be enabled or disabled
  • LLM context management for consistent, coherent conversations

The join route supports options like custom bot names, avatar images, entry messages, and specialized persona selection.

The API and implementation are open source. We welcome contributions and pull requests from the community. See our getting started guide for local development setup.

Directory Structure

On this page