/v0/providers
), in which case the “provider” argument
must be set accordingly. Otherwise, the endpoint must support the OpenAI
/chat/completions
format. To query your custom endpoint, replace your endpoint
string with <endpoint_name>@custom
when querying any general custom endpoint. You
can show all custom endpoints by querying /v0/endpoints
and passing custom
as
the provider
argument.
Bearer <token>
, where <token>
is your auth token.@custom
, otherwise if it’s a fine-tuned model from one of the existing providers it can be specified with a prepending custom-
, i.e. @custom-anthropic
.llama-3-baseten@custom
to distinguish the custom endpoint within Unify, but under the hood need to pass llama-3.2-90b-chat
to the Baseten endpoint.