quattro router

6 brains, one endpoint. pick a model or let it roll through all of them until one answers. phoenix router dey inside as last fallback.

• nemotron-3-ultra-550b — the big smart one
• nemotron-3.5-super-vl — thorough agent
• nemotron-3.5-lightning — speed demon
• cohere north-mini-code — precise + fast
• laguna-xs-2.1 — thorough agent
• nemotron-3-nano-omni-30b — reasoning + agent

chat: POST /chat body {"message": "hi"} + header X-API-Key

stream: POST /chat/stream same body, server-sent events back

health: GET /health · api docs: /docs

running on quattro · auto-fallback · safety-filtered