{"endpoint":"/api/llm","method":"POST","shape":{"subscription_id":"string (required) — same id from /api/subscribe","capability":"string (required) — one of: fast | reasoning | general | code | vision | embedding","messages":"array (required for chat capabilities) — OpenAI chat-message shape","embedding_input":"string|string[] (required for capability=embedding)","max_tokens":"number (optional, default 1024, hard cap 4096)","temperature":"number (optional, default 0.7)"},"relay_reachable":false,"capabilities":["fast","reasoning","general","code","vision","embedding"],"pricing":"1 credit per successful call (refill via /api/subscribe/credit). Future: token-weighted meter.","architecture":"Vercel shim (auth + credit gate) -> Mars-controlled relay (provider keys never leave that boundary).","privacy":"Requests are not logged beyond aggregate counters. Provider used is returned for transparency.","v":"0.2"}