WOTANN treats the LLM as just the inference call. Memory, tools, sandbox, orchestration, learning — all come from the harness. Inference comes from a remote subscription or API-key provider you ...