Embed API Client

The EMbed API client holds the credentials required to connect to the server or service that provides the embeddings for a given query.

In this section of the UI you can configure:

1. The Name of the Embed API Client
2. The Provacy Tier of the Embed API Client: this is the sensitivity of data that this embedding client can handle. This is used to ensure that the Embed API Client is only used for embedding content that is of the same or lower sensitivity than the Embed API Client.
3. Type: The type of client, Montag supports two types: “OpenAI” and “HuggingFace”.

OpenAI represents the OpenAI embeddings endpoint, and HuggingFace represents a self-hosted HuggingFace model.

Montag provides a service called embed_server that can run any transformers capable hugging-face model and will download the model automatically based on the value set in the Embedding Settings.

To use Montag’s local embedding service, simply set the endpoint URl to http://embed_server:8000 and the type to HuggingFace.