Tool-Server LLM Override (Optional)

By default the tool server shares the backend LLM configuration, but you can point it at a dedicated OpenAI-compatible endpoint. This is useful when the tool server needs a cheaper or faster model for text compression, summarization, or tool routing.

Variable	Description
LLM_CONFIG__OPENAI_API_KEY	API key for the OpenAI-compatible endpoint the tool server should use (OpenAI, Azure OpenAI, Together, OpenRouter, etc.).
LLM_CONFIG__OPENAI_BASE_URL	Base URL for the API (leave blank to use api.openai.com).
LLM_CONFIG__OPENAI_MODEL	Model identifier to request (e.g., gpt-4o-mini, gpt-5-mini).

Populate the variables, restart the tool server, and check its logs to ensure it successfully loads the override configuration.