{"message":"POST /derive with {'graph': '<json_str>'}. Then use /getCode, /whereami, /logs, /result. Controls: POST /pause/{job_id}, /resume/{job_id}, /stop/{job_id}. Blocks: b_prompt (template rendering), b_llm (Ollama via LLM_API_URLS with automatic model routing, or OpenAI via stored key). LLM model endpoints: GET /llmModels, POST /openaiModels. Email folder endpoint: POST /emailFolders. LLM_API_URLS currently: ['http://host.docker.internal:18034/api/generate', 'http://host.docker.internal:18035/api/generate']"}