OpenAI-compatible chat completions (abliterated LLM). POST the standard {"model","messages"} body; answered in <= 512 tokens.