Running on your own model

In Settings, a workspace can use its own model instead of ours: an OpenAI-compatible endpoint, the Claude API, Azure OpenAI, Vertex AI or Amazon Bedrock, with the credential that kind takes. It answers the questions a run puts to a model. With an OpenAI-compatible or Azure provider it can make your search embeddings too, with text-embedding-3-small or text-embedding-3-large; otherwise those are ours. Either way, a release is always compared with queries embedded by the same model that built it. Its calls don't spend the workspace's starter credits, the calls to our model it is given once to start with.

The credential stays where you put it. It is encrypted, never shown again, and only sent to the address it was saved for: change the address and the credential has to be entered again. An address on a private or internal network is refused.