You can use your own LLM (Large Language Model) with

AgentSession

in several ways, depending on how your LLM is deployed and accessible: • Ollama for Local Models: If your model runs locally and exposes an OpenAI-compatible API (such as via Ollama), you can use the OpenAI plugin. Specify your model name and the base URL for your local API endpoint in the configuration. • LangChain Integration: If you have a custom LLM workflow using LangChain or LangGraph, use the LangChain plugin. Create your workflow graph and pass it to the

LLMAdapter

for integration with

AgentSession

. • Provider with Custom Model Support: Some providers, like Amazon Bedrock, support custom models. You can specify your custom model name when configuring the provider plugin for

AgentSession

. • Custom LLM Node: For full customization, implement your own LLM node by creating a class that follows the LLM interface pattern and pass it to

AgentSession

. Choose the approach that matches your LLM deployment. The Ollama and LangChain integrations are the most common for local and custom workflows, respectively. Sources: Ollama integration guide | LiveKit Docs | livekit-plugins/livekit-plugins-langchain/README.md | Large language model (LLM) integrations | LiveKit Docs