Omniroute connections

[dantaverner-create]

I have Omniroute running locally, I've connected Nvidia NIM, Openrouter successfully. I'm trying to connect tokenharbor.ai, and whichever way i try, it fails on the prefix field in add connection. What prefix should i use for token harbor?

[diegosouzapw]

Hi @dantaverner-create. There is no built-in tokenharbor provider, so the path is Providers → Add OpenAI Compatible: their base URL and your API key from them.

About the Prefix field: it is just the name you will use to route (<prefix>/<model>), you choose it freely. It is required and there is no character restriction, so there are only two ways it fails:

1. Empty: "Prefix is required".
2. Collides with a built-in provider id or alias: "<x>" is a reserved provider prefix — choose a different prefix (src/shared/validation/schemas/provider.ts:409-425). Every registry id and alias is reserved, along with retired providers, because <prefix>/model would otherwise route to the built-in provider instead of your node.

tokenharbor and th are not reserved, so something else is being rejected. Paste the literal error message and the exact value you typed and I can tell you which of the two it is.

One note for the connection itself: give the base URL up to the version segment, for example https://<host>/v1, not the full /chat/completions path.

This rule is not documented in any guide, which is a fair part of why you hit it. That is now tracked in #13939.

[turbolego]

I've made a pull request to add tokenharbor.ai as new provider in OmniRoute: #16315

[diegosouzapw]

Thanks @turbolego. For anyone following: Token Harbor as a built-in provider is PR #16315, which is still open and under review, so it isn't in a release yet. Until it merges, the Add OpenAI Compatible route described above works with any prefix that isn't reserved.