cannot use a custom LLM client model

[czz253861763]

The earlier version (0.50) allowed users to use a custom LLM client model via an OpenAI‑compatible API (e.g., openai/custom_llm_xxx), but this feature appears to be unavailable in the latest version.