How to Use Weaviate with Langchain for Vector Search

To use Weaviate with Langchain, install the weaviate-client and langchain Python packages, then create a Weaviate vector store instance in Langchain pointing to your Weaviate server. Use Langchain's Weaviate vector store class to add, query, and manage embeddings seamlessly.

Syntax

The main steps to use Weaviate with Langchain are:

• Import the Weaviate vector store from Langchain.
• Initialize the Weaviate client with your server URL and API key if needed.
• Create a Langchain Weaviate vector store instance by passing the client and index name.
• Use the vector store methods like add_texts to add documents and similarity_search to query.

python

from langchain.vectorstores import Weaviate
import weaviate

client = weaviate.Client(url="http://localhost:8080", auth_client_secret=None)

vector_store = Weaviate(client, index_name="MyIndex", embedding=None)

# Add documents
vector_store.add_texts(["Hello world", "Langchain with Weaviate"])

# Query
results = vector_store.similarity_search("Hello")

Example

This example shows how to connect Langchain to a local Weaviate instance, add text documents, and perform a similarity search.

python

from langchain.vectorstores import Weaviate
from langchain.embeddings.openai import OpenAIEmbeddings
import weaviate

# Initialize Weaviate client
client = weaviate.Client(url="http://localhost:8080")

# Initialize embeddings (requires OPENAI_API_KEY set in environment)
embeddings = OpenAIEmbeddings()

# Create Weaviate vector store instance
vector_store = Weaviate(client, index_name="LangchainDemo", embedding=embeddings)

# Add texts to Weaviate
texts = ["Langchain makes building LLM apps easy.", "Weaviate is a vector search engine."]
vector_store.add_texts(texts)

# Perform similarity search
query = "What is Langchain?"
results = vector_store.similarity_search(query, k=1)

print(results)

Output

[Document(page_content='Langchain makes building LLM apps easy.', metadata={})]

Common Pitfalls

• Not running a Weaviate server or incorrect URL causes connection errors.
• Missing or incorrect API keys if your Weaviate instance requires authentication.
• Forgetting to specify the embedding parameter when creating the vector store, which is needed to convert text to vectors.
• Not creating the Weaviate schema/index before adding data can cause failures.

python

from langchain.vectorstores import Weaviate
import weaviate

# Wrong: Missing embedding parameter
client = weaviate.Client(url="http://localhost:8080")
vector_store = Weaviate(client, index_name="Demo")  # Missing embedding causes error

# Right: Include embedding
from langchain.embeddings.openai import OpenAIEmbeddings
embeddings = OpenAIEmbeddings()
vector_store = Weaviate(client, index_name="Demo", embedding=embeddings)

Quick Reference

Key points to remember when using Weaviate with Langchain:

• Install weaviate-client and langchain packages.
• Run a Weaviate server accessible to your code.
• Use Langchain's Weaviate vector store class with an embedding model.
• Always create or verify the Weaviate schema before adding data.
• Use add_texts to add documents and similarity_search to query.

Key Takeaways

Use Langchain's Weaviate vector store class with a Weaviate client and an embedding model.

Ensure your Weaviate server is running and accessible before connecting.

Always specify the embedding parameter to convert text into vectors.

Create the Weaviate schema/index before adding documents to avoid errors.

Use add_texts to add data and similarity_search to find relevant documents.