One request shape
Use the familiar chat-completions endpoint with streaming and model selection. The auto class lets the router choose inside your policy.
OpenAI-compatible LLM API
Keep the client. Change who competes for the request.
Omnious exposes 327 catalogued model classes through one API. Point an OpenAI-compatible client at the Omnious base URL, fund a bounded credit key, and let live providers compete on price and measured service.
A complete first request
POST /v1/chat/completionscurl https://api.omnious.xyz/v1/chat/completions \
-H "x-credit-key: $OMNIOUS_CREDIT_KEY" \
-H "content-type: application/json" \
-d '{
"model": "auto",
"messages": [{"role": "user", "content": "Explain this trace"}],
"stream": true
}'Credit keys are spend capabilities. Put one in a secret manager, set a narrow budget, and never ship it in browser code.
What changes behind the endpoint
A live price
Providers submit executable asks. The winning route and clear are decided for the request instead of copied from a static reseller price list.
Actual-use settlement
A prepaid credit key reserves a bounded amount and captures the metered token cost. The final charge follows the receipt, not the reservation.
A receipt, not just a log
The response records the measured usage and economic path needed to reconstruct what was charged and why.