Inference APIs
Routero normalizes the request and response schema across every provider behind a single OpenAI-compatible API. Point any OpenAI SDK — or a plain HTTP client — at one base URL, and use the model string to choose what runs.
- Base URL:
https://api.routero.dcsmartvision.com/v1 - Authentication:
Authorization: Bearer YOUR_ROUTERO_KEY
Every endpoint accepts the standard OpenAI request shape and returns the standard OpenAI response shape; Routero translates to and from each upstream provider.
Endpoints
| Endpoint | Method | Purpose |
|---|---|---|
/chat/completions |
POST | Generate text and conversations |
/embeddings |
POST | Create vector embeddings |
/images/generations |
POST | Generate and edit images |
/models |
GET | List available models |
For chat, Routero also speaks the Anthropic-native Messages API at https://api.routero.dcsmartvision.com/anthropic/v1/messages. See Chat.
Response headers
Every inference response carries x-routero-* headers you can log or display:
| Header | Meaning |
|---|---|
x-routero-call-id |
Unique ID for the request |
x-routero-model-id |
The deployment that served the request |
x-routero-model-region |
Region of the serving deployment |
x-routero-response-cost |
Cost charged for the response |
x-routero-response-duration-ms |
End-to-end latency |
x-routero-attempted-fallbacks |
Deployments tried on failover, if any |
Streaming
Streaming endpoints use standard Server-Sent Events (SSE). On failover mid-stream, Routero replays only the tail, so the client receives one uninterrupted stream.
Related
→ Calling the API for a quick first request. → Routing & Load Balancing for how a model string resolves to a deployment.