API compatibility
Connect OpenAI and OpenRouter clients to Starport, and learn the routes, SDK versions, errors, and limits of each API family.
This area is for a developer who connects a client or an SDK to a Starport gateway. Starport serves two API families on one listener.
Two API families
| Family | Base URL | Error body |
|---|---|---|
| OpenAI | http://<gateway-host>:8080/v1 | OpenAI error object |
| OpenRouter | http://<gateway-host>:8080/api/v1 | OpenRouter error object |
Each request needs a gateway API key in the Authorization: Bearer header. A provider key does not work as a gateway key.
Catalog and callable models
A model can be in the Starmap catalog and still not be callable. The model routes list only the offerings that this gateway can route. Use a model ID from that list in a request.
What this area covers
- The routes of each family and the scope that each route needs.
- SDK setup and the SDK versions that the repository tests.
- Error bodies, rate limits, budgets, request limits, and guardrails.
Optional caches
Turn the response, discovery, extraction, and semantic caches on or off, learn their lifetimes and memory limits, and fix a cache conflict in development mode.
Supported API surfaces
Find each OpenAI and OpenRouter route that Starport serves, the scope that each route needs, and the routes that each family does not have.