Playground

This talks to the same public endpoint any OpenAI client would use. Each answer carries the routing decision, cost against the premium baseline and a trace id you can open.

Conversation

No messages yet

Send something short, then something demanding — the router should pick different tiers.

Sending anonymously — allowed on this demo.

Same call from a client

Any OpenAI SDK works unchanged.

curl -N https://sentinel-gateway-83i9.onrender.com/v1/chat/completions \
  -H "content-type: application/json" \
  -H "authorization: Bearer $SENTINEL_API_KEY" \
  -d '{"messages":[{"role":"user","content":"..."}],"stream":true}'

Extra fields policy, cache and capabilities are additive; the response adds a sentinel block. Standard clients ignore both.

Request options

ModelLeave on router to let the policy decide.Policy

TemperatureMax tokens

CacheAuto caches only low-temperature calls.

Stream tokens

Server-sent events, identical framing to OpenAI.