Weaviate Retriever¶

Use WeaviateRetriever to evaluate objects in Weaviate collections.

Install¶

pip install "evret[weaviate]"

Basic Usage¶

from evret.retrievers import WeaviateRetriever

def encode_query(query: str) -> list[float]:
    return embedding_model.embed_query(query)

retriever = WeaviateRetriever(
    collection_name="Docs",
    query_encoder=encode_query,
    url="http://localhost:8080",
    id_field="doc_id",
    return_properties=["doc_id", "text", "source"],
)

results = retriever.retrieve("how to tune ndcg", k=5)
for item in results:
    print(item.doc_id, item.score, item.metadata)

With an Existing Weaviate Client¶

import weaviate

client = weaviate.connect_to_local()

retriever = WeaviateRetriever(
    collection_name="Docs",
    query_encoder=encode_query,
    client=client,
    query_filter=my_filter,
)

Notes¶

query_encoder must return a non-empty vector
Scores are derived from metadata (certainty, distance, or score)
Works with clients that expose collections.use/get