Support persistent custom model options in openai.LLM.withTelnyx() / OpenAI-compatible LLMs

[robbedgdev]

Your question

Hi LiveKit team,

I’m using the Node.js Agents SDK with @livekit/agents-plugin-openai and Telnyx via:

openai.LLM.withTelnyx({
apiKey,
model: "moonshotai/Kimi-K2.6",
temperature: 0.1,
})

Telnyx exposes an OpenAI-compatible request body parameter called enable_thinking, for example:

{
"enable_thinking": false
}

Currently, withTelnyx() only accepts a fixed set of init options (model, apiKey, baseURL, user, temperature, client). The plugin’s chat() method does support arbitrary extraKwargs, but the default Agent.llmNode() path does not pass any extraKwargs, so there doesn’t seem to be a clean way to set this once at LLM initialization for normal AgentSession usage.

Feature request: add a persistent modelOptions / extraKwargs option to openai.LLM and provider helpers like withTelnyx(), so OpenAI-compatible provider-specific fields can be forwarded on every chat completion request.

Example desired API:

openai.LLM.withTelnyx({
apiKey,
model: "moonshotai/Kimi-K2.6",
temperature: 0.1,
modelOptions: {
enable_thinking: false,
},
})

This would also help with other provider-specific OpenAI-compatible parameters without needing to override llmNode() or wrap the LLM manually.

Thanks.