Skip to main content

Retrievers

A retriever is an interface that returns documents given an unstructured query. It is more general than a vector store. A retriever does not need to be able to store documents, only to return (or retrieve) them. Retrievers can be created from vector stores, but are also broad enough to include Wikipedia search and Amazon Kendra.

Retrievers accept a string query as input and return a list of Document's as output.

For specifics on how to use retrievers, see the relevant how-to guides here.


Was this page helpful?


You can also leave detailed feedback on GitHub.