Connect Another OpenAI-Compatible API
An unlisted runtime can use the advanced text-worker path when it exposes the OpenAI models and chat-completions contract the worker tests locally. A similar URL shape is not enough; the setup canary must complete successfully.
Setup
- Start the backend and confirm its
/v1/modelsendpoint reports the model. - Download the verified text worker from aipowergrid.io/run.
- Choose the manual OpenAI-compatible path and enter the backend base URL.
- Enter any backend credential only in the protected local worker UI.
- Select the exact model and run the local generation test.
- Set conservative output, context, schedule, and concurrency limits.
- Approve the worker in Console, wait for Online, and verify it on
/run.
⚠️
Use only an endpoint you control or are authorized to resell. Connecting a third-party hosted API can violate its terms and exposes Grid workloads to an additional provider.
Operator-entered endpoints may be loopback, LAN, or public HTTPS services, but the worker rejects malformed URLs, embedded credentials, and known metadata service targets. Public backend credentials remain on the worker host.