langchain-kusto 0.0.1

LangChain vector store implementation for Azure Data Explorer (Kusto) - PRE-ALPHA VERSION

LangChain Kusto Vector Store

⚠️ PRE-ALPHA VERSION - This package is in very early development and not recommended for production use.

A LangChain vector store implementation for Azure Data Explorer (Kusto), Microsoft Fabric Eventhouse, and other Kusto-compatible databases.

Current Status

This is a very initial version with only retrieval capabilities. Document storage functionality is not yet implemented.

Features

• ✅ Retrieve vector embeddings from Azure Data Explorer (Kusto) or Microsoft Fabric Eventhouse
• ✅ Compatible with LangChain's vector store interface
• ✅ Similarity search with cosine similarity metric
• ❌ Document storage (not yet implemented)
• ❌ Batch operations (not yet implemented)

Installation

pip install langchain-kusto

Quick Start

from langchain_kusto import KustoVectorStore
from langchain_openai import AzureOpenAIEmbeddings
from azure.identity import DefaultAzureCredential

# Initialize embeddings
embeddings = AzureOpenAIEmbeddings(
    azure_endpoint="your-openai-endpoint",
    azure_deployment="your-embedding-deployment",
    openai_api_version="2023-05-15"
)

# Initialize the vector store (retrieval only)
vector_store = KustoVectorStore(
    connection="https://your-cluster.kusto.windows.net",  # or KustoConnectionStringBuilder
    database="your_database",
    collection_name="your_table",
    embedding=embeddings,
    embedding_column="embedding_text",  # optional, defaults to "embedding"
    id_column="vector_id",              # optional, defaults to "id"
    content_column="doc_text"           # optional, defaults to "text"
)

# Search for similar documents (this requires pre-existing data in Kusto)
results = vector_store.similarity_search("your query text", k=5)

Complete Example

See demo.py for a complete working example using Azure OpenAI embeddings and a RAG (Retrieval-Augmented Generation) pipeline.

Requirements

• Python >= 3.8
• Azure Data Explorer cluster or Microsoft Fabric Eventhouse with pre-existing vector data
• LangChain Core >= 0.1.0
• Azure authentication (DefaultAzureCredential)

Data Prerequisites

Since this version only supports retrieval, you need to have your vector embeddings already stored in a Kusto table with the following structure:

.create table your_table (
    vector_id: string,
    doc_text: string,
    embedding_text: dynamic  // Array of float values representing the vector
    // ... other metadata columns
)

Development Status

This package is currently in pre-alpha development. APIs may change significantly between versions. The current version only supports reading existing vector data from Kusto - document ingestion and storage capabilities will be added in future releases.

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.