Custom OpenAI-compatible API endpoint (external server: LM Studio, Ollama, llama-server, vLLM). Configured via .