Introducing the KITT V2 Project
KITT V2 offers a modern approach to integrating artificial intelligence within video conferencing. Built on the Agents Framework, the new version of KITT enhances the user experience by embedding a ChatGPT-powered AI within WebRTC conference calls. With the introduction of this updated version, KITT continues to innovate in the realm of virtual communication.
What is KITT?
KITT is an AI solution capable of enhancing video call interactions. By using ChatGPT, KITT can engage participants in real-time discussions during a WebRTC video conference, offering a dynamic addition to traditional video conferencing tools.
Experience KITT
For those eager to see KITT in action, an online demo is accessible at LiveKit. This demo provides a hands-on opportunity to explore how KITT operates in a virtual meeting environment.
How Does KITT Work?
The KITT project comprises two main services:
- Meet Service: This is a NextJS application that sets up a conventional video call interface.
- lkgpt-service: This service embodies KITT, integrating features like Google Cloud Platform's (GCP) speech-to-text, ChatGPT for conversation, and GCP's text-to-speech for responses.
When a room is created, a webhook triggers the lkgpt-service, adding KITT as a participant to the conference. KITT processes conversations using speech recognition and artificial intelligence to provide seamless interaction.
Architecture Overview
The architecture of KITT is straightforward yet effective, consisting of a pipeline that processes audio input to generate intelligent responses, as illustrated below:
Getting Started with KITT
Necessary Tools and Credentials
Before running KITT, ensure you have the following:
- Google Application Credentials for authenticating GCP services.
- OpenAI API key for accessing ChatGPT.
- LiveKit API key, secret, and URL from LiveKit Cloud.
- Go programming language (version 1.19 or newer) and Node.js installed.
Running KITT Locally
To try KITT on a local machine, both the meet and lkgpt-service need to be operational.
Running the Meet Service
-
Navigate to the
meet/directory. -
Copy the example environment file to a local environment configuration and fill out LiveKit connection details.
-
Execute the command:
yarn install && yarn dev
Running the lkgpt-Service
-
Navigate to the
lkgpt-service/directory. -
Copy and configure the YAML configuration file with necessary credentials.
-
Start the service using:
go run /cmd/server/main.go --config config.yaml --gcp-credentials-path gcp-credentials.json
After initiating both services, access the service at http://localhost:3000. Due to the local setup not having webhook access to spawn KITT, manually invoke an API call using the room name from the Meet interface to activate KITT:
curl -XPOST http://localhost:3001/join/<room_name>
Through this setup, users can explore the capabilities and experience the intelligent interactions KITT V2 brings to online conferencing.