The Snowflake AI Toolkit is a Streamlit-based native application designed to accelerate AI development within Snowflake. It provides a playground for exploring Snowflake Cortex functions, building AI-powered data pipelines, and implementing features like Retrieval-Augmented Generation (RAG) and hybrid search, targeting data scientists and developers looking to leverage AI capabilities directly within their Snowflake data environment.

How It Works

This toolkit acts as an accelerator and playground, integrating directly with Snowflake's Cortex and AI Functions. It offers distinct modes for Playground (interactive experimentation with Cortex functions and prompt testing), Build (data pipeline construction), RAG (question answering with custom data), and Search (hybrid vector and keyword search). The application orchestrates these capabilities, enabling users to create and manage AI solutions, including Cortex Agents that interact with structured and unstructured data.

Quick Start & Requirements

• Installation: Clone the repository (git clone https://github.com/sgsshankar/Snowflake-AI-Toolkit.git) and install dependencies locally with pip install -r requirements.txt.
• Prerequisites: Active Snowflake account in a supported region, Private Preview access for specific models (contact Snowflake account team), ACCOUNTADMIN or equivalent role, Python 3.7+, Streamlit, and snowflake-cli.
• Model Availability: claude-4-sonnet and claude-4-opus are available only in AWS US (cross-region).
• Authentication: Supports JWT-based authentication for Cortex Agents via RSA key pairs.
• Running Locally: streamlit run streamlit_app.py
• Running Natively on Snowflake: snow streamlit deploy --account "<your_account>" --user "<your_username>" --password "<your_password>" --role "<your_role>" --warehouse "<your_warehouse>" --database "<your_database>" --replace
• Documentation: Snowflake AI Toolkit

Highlighted Details

• Enables text completion and generation using Snowflake Cortex's COMPLETE function.
• Facilitates Retrieval-Augmented Generation (RAG) for question answering on user data.
• Provides a hybrid search engine for text data, supporting RAG applications and chat agents.
• Allows creation and management of Cortex Agents for orchestrating data interactions.

Maintenance & Community

This is an open-source project not officially endorsed by Snowflake. Contributions are welcome via pull requests. For assistance, open an issue on the GitHub repository.

Licensing & Compatibility

Licensed under the Apache License, Version 2.0. This license permits commercial use and linking with closed-source projects, provided compliance with the license terms.

Limitations & Caveats

The application currently only supports JWT-based authentication for Cortex Agents, not OAuth. Private preview models require explicit access requests to Snowflake. The project is explicitly stated as not an official Snowflake offering.