Hack LiveKit Voice Agent to Support Groq LLM & STT Models

---

---

AI Summary

Video Summary: Converting Live Kit Agent from OpenAI to Grock API

Setup:

• Start with a Live Kit voice agent template from the previous video.
• Open the voice agent’s folder in VS Code.

Configuration Changes:

• Add a system variable for the Grock API key in the .env.local file.
• Import the OS library in the agent.py file.
• Use the search function to find and modify the LLM engine code.

Understanding Dependencies:

• Use “Go to Definition” in VS Code to view Python dependency files.
• If using a virtual environment, navigate to the dependency file in the venv folder.

Modifying Live Kit Dependencies:

• Update the Live Kit Python dependency files to support Grock for speech-to-text.
• Modify the OpenAI STT class to send audio to the Grock API.
• Change the model to Grock’s distilwhisper large v3.
• Set the base URL to Grock’s OpenAI formatted API URL.
• Adjust the stt class and parameters to match Grock’s API requirements.

Finalizing Agent Configuration:

• Pass the Grock API key as an input parameter to the STT and LLM engines.

Running the Agent:

• Open a new terminal in VS Code.
• If using a virtual environment, activate it.
• Run the command to start the agent in development mode.

Future Improvements:

• The video hints at future content for enhancing the voice agent, including function calling, setting up the Live Kit front end and API locally, and more.

Note:

• The source code for the updates is available for Pro members on the creator’s Discord server.
• The video encourages likes and subscriptions for future updates and offers early access to tutorials for Pro members.