autorag.retrieve

Vector-similarity retrieval over an ingested corpus.

Composes an Embedder (to encode the query) with a VectorStore (to do the nearest-neighbour lookup) so that the choice of embedding model and backing store can vary independently.

class autorag.retrieve.Retriever(store, embedder)

Bases: object

Embed a question and pull the top_k most similar chunks.

Parameters:

• store (VectorStore)

• embedder (Embedder)

retrieve(question, top_k)

Return the top_k chunks most similar to question.

Parameters:

• question (str)

• top_k (int)

Return type:

list[Retrieved]