The retrieval plugin includes:
- Support for multiple vector databases providers
- All 4 different authentication methods
- Multiple different API features
- Docs
- Github
The code is a backend API built using FastAPI and Python. It defines endpoints for inserting and querying documents, as well as deleting them. The API uses a datastore that is defined in datastore.factory, and gets instantiated in the startup function. There is also an authentication middleware in place that checks for a bearer token passed in the header of each request. The token can be set as an environment variable, and is checked against the token in the validate_token function. If the token is missing or invalid, a 401 HTTP error is returned.
- Github
- Tech stack: node/js/ts/next.js