LangChain Snowflake Vector Store

A LangChain integration for Snowflake's vector database capabilities, enabling semantic search and similarity matching using Snowflake's native VECTOR data type and VECTOR_COSINE_SIMILARITY function.

Features

• 🏔️ Native Snowflake Integration: Uses Snowflake's built-in vector capabilities
• 🔍 Semantic Search: Powered by VECTOR_COSINE_SIMILARITY function
• 📊 Scalable: Leverages Snowflake's cloud-native architecture
• 🔒 Secure: Enterprise-grade security and compliance
• 🚀 High Performance: Optimized for large-scale vector operations

Installation

pip install langchain-snowflake-vectorstore

Quick Start

from langchain_snowflake_vectorstore import SnowflakeVectorStore
from langchain_openai import OpenAIEmbeddings

# Initialize embeddings
embeddings = OpenAIEmbeddings()

# Create vector store
vector_store = SnowflakeVectorStore(
    account="your-account",
    user="your-username", 
    password="your-password",
    database="your-database",
    schema="your-schema",
    warehouse="your-warehouse",
    role="your-role",
    table_name="vector_documents",
    embedding_function=embeddings,
    embedding_dimension=1536
)

# Add documents
texts = [
    "LangChain is a framework for developing applications powered by language models.",
    "Snowflake is a cloud-based data warehousing platform.",
    "Vector databases enable semantic search and similarity matching."
]

ids = vector_store.add_texts(texts)

# Search for similar documents
results = vector_store.similarity_search("What is LangChain?", k=2)
for doc in results:
    print(doc.page_content)

Configuration

Environment Variables

You can set your Snowflake credentials using environment variables:

export SNOWFLAKE_ACCOUNT=your-account
export SNOWFLAKE_USER=your-username
export SNOWFLAKE_PASSWORD=your-password
export SNOWFLAKE_DATABASE=your-database
export SNOWFLAKE_SCHEMA=your-schema
export SNOWFLAKE_WAREHOUSE=your-warehouse
export SNOWFLAKE_ROLE=your-role

Connection Parameters

Parameter	Description	Required
account	Snowflake account identifier	Yes
user	Username for authentication	Yes
password	Password for authentication	Yes
database	Database name	Yes
schema	Schema name	Yes
warehouse	Warehouse name	Yes
role	Role name	No
table_name	Table name for storing vectors	Yes
embedding_function	Function to generate embeddings	Yes
embedding_dimension	Dimension of embedding vectors	Yes

Advanced Usage

Custom Table Creation

# Recreate table with custom settings
vector_store.recreate_table()

Similarity Search with Scores

# Get similarity scores along with documents
results_with_scores = vector_store.similarity_search_with_score("query", k=5)
for doc, score in results_with_scores:
    print(f"Score: {score}, Content: {doc.page_content}")

Adding Documents with Metadata

texts = ["Document 1", "Document 2"]
metadatas = [{"source": "file1.txt"}, {"source": "file2.txt"}]
ids = vector_store.add_texts(texts, metadatas=metadatas)

Batch Operations

# Create from texts (class method)
vector_store = SnowflakeVectorStore.from_texts(
    texts=texts,
    embedding=embeddings,
    account="your-account",
    # ... other parameters
)

Requirements

• Python 3.8+
• Snowflake account with vector support
• LangChain Core
• Snowflake Connector for Python
• SQLAlchemy

Snowflake Setup

Your Snowflake account must support the VECTOR data type and VECTOR_COSINE_SIMILARITY function. These features are available in recent Snowflake versions.

Required Permissions

Ensure your Snowflake role has the following permissions:

• CREATE TABLE on the target schema
• INSERT, SELECT, UPDATE, DELETE on the vector table
• USAGE on the database, schema, and warehouse

Testing

Run the test suite:

# Unit tests
pytest tests/test_vectorstore.py

# Integration tests (requires Snowflake credentials)
export SNOWFLAKE_ACCOUNT=your-account
# ... set other environment variables
pytest tests/test_integration.py -m integration

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Support

For issues and questions:

• GitHub Issues: Create an issue
• LangChain Community: Join the discussion

Changelog

v0.1.0

• Initial release
• Basic vector store functionality
• Snowflake integration with VECTOR data type
• Similarity search using VECTOR_COSINE_SIMILARITY
• Comprehensive test suite