This repository provides KITT, a real-time AI assistant powered by ChatGPT that integrates into WebRTC conference calls. It's designed for users who want to add intelligent voice-based interaction to their video conferencing experiences.

How It Works

KITT comprises two services: a Next.js-based video conferencing frontend (meet) and a Go-based backend (lkgpt-service). The backend leverages GCP's speech-to-text and text-to-speech, along with ChatGPT, to process audio, generate responses, and speak them back to participants in a conference call. This architecture enables seamless, real-time conversational AI within a familiar video call interface.

Quick Start & Requirements

• Installation: Clone the repository and set up environment variables for LiveKit, Google Cloud, and OpenAI.
• Prerequisites:
  • Google Cloud account with GOOGLE_APPLICATION_CREDENTIALS JSON.
  • OpenAI API key.
  • LiveKit API key, secret, and URL.
  • Go 1.19+ and Node.js.
• Running Locally:
  1. meet service: cd meet/, copy .env.example to .env.local, fill details, run yarn install && yarn dev.
  2. lkgpt-service: cd lkgpt-service/, copy config-sample.yaml to config.yaml, fill details, run go run /cmd/server/main.go --config config.yaml --gcp-credentials-path gcp-credentials.json.
  3. Manually spawn KITT: curl -XPOST http://localhost:3001/join/<room_name>.
• Demo: An online demo is available at https://livekit.io/kitt.

Highlighted Details

• Real-time ChatGPT integration within WebRTC calls.
• Utilizes GCP Speech-to-Text and Text-to-Speech for natural interaction.
• Next.js frontend for a standard video call experience.

Maintenance & Community

The repository is marked as archived, indicating it is no longer actively maintained. The description mentions "KITT V2 is here! We launched a new version of KITT built with Agents Framework," suggesting a successor project may exist.

Licensing & Compatibility

The repository does not explicitly state a license in the provided README.

Limitations & Caveats

The project is archived, meaning it is not actively maintained or updated. The local setup requires manual intervention to spawn the AI participant, as webhooks are not configured for local development environments.