langchain-weaviate

0.0.6 verified Fri May 01 auth: no python

An integration package connecting Weaviate vector database with LangChain, providing vector store, retriever, and document loader. Current version 0.0.6 supports Python >=3.10,<4.0. Release cadence is monthly.

pip install langchain-weaviate

Common errors

error ModuleNotFoundError: No module named 'langchain_weaviate' ↓

cause Package not installed or installed in wrong environment.

fix

pip install langchain-weaviate

error AttributeError: 'str' object has no attribute 'get' ↓

cause Passing a Weaviate client instance instead of URL string to `weaviate_url`.

fix

Use weaviate_url="http://localhost:8080" instead of passing a client.

error weaviate.exceptions.WeaviateClosedError: Failed to connect ↓

cause Weaviate server not running or wrong URL/port.

fix

Start Weaviate via Docker: docker run -p 8080:8080 semitechnologies/weaviate:1.25.0 and verify URL.

Warnings

breaking Version 0.0.6 requires langchain-core>=0.3.0, which is a breaking change from langchain<0.3. Ensure you upgrade langchain-core. ↓

fix pip install "langchain-core>=0.3.0"

gotcha The `weaviate_url` parameter expects a string like 'http://localhost:8080', not a Weaviate client instance. Using client instance will raise AttributeError. ↓

fix Pass the URL as a string; client is created internally.

gotcha When connecting to Weaviate Cloud (WCD), you must provide `auth_client_secret` in `client_kwargs`, not as a top-level parameter. Common mistake: passing `auth_client_secret` as a separate argument. ↓

fix vectorstore = WeaviateVectorStore(..., client_kwargs={"auth_client_secret": weaviate.auth.AuthApiKey(api_key="your-key")})

Imports

WeaviateVectorStore

wrong

from langchain.vectorstores import Weaviate

correct

from langchain_weaviate import WeaviateVectorStore

Deprecated in langchain-community, use this dedicated package

WeaviateHybridSearchRetriever

wrong

from langchain.retrievers import WeaviateHybridSearchRetriever

correct

from langchain_weaviate import WeaviateHybridSearchRetriever

Wrong import path

WeaviateDocumentLoader

wrong

from langchain.document_loaders import WeaviateDocumentLoader

correct

from langchain_weaviate.document_loaders import WeaviateDocumentLoader

Cannot find this class in langchain-community

Quickstart

Initialize a Weaviate vector store with embeddings.

import os
from langchain_weaviate import WeaviateVectorStore
from langchain_community.embeddings import FakeEmbeddings

embeddings = FakeEmbeddings(size=512)
vectorstore = WeaviateVectorStore.from_documents(
    documents=[],
    embedding=embeddings,
    weaviate_url=os.environ.get("WEAVIATE_URL", "http://localhost:8080"),
    index_name="MyDocument",
    client_kwargs={"auth_client_secret": None},
)
print(vectorstore._client.is_ready())