Snowflake Chat (RAG Agent)

A local Streamlit application that allows you to query Snowflake data using natural language with the power of LangGraph agents.

Click for streamlit live demo - Requires DB credentials and API key

Features

• Natural language to SQL translation
• Interactive chat interface
• Query results in tabular format
• CSV and Excel export functionality
• LangGraph agent-based processing

Setup Instructions

Prerequisites

• Python 3.8+
• Snowflake account with appropriate credentials
• OpenAI API key

Installation

1. Clone this repository:

   git clone <repository-url>
   cd snowflake-rag

2. Install dependencies:

   pip install -r requirements.txt

3. Create a .env file in the root directory with your credentials:

   # OpenAI API Key
   OPENAI_API_KEY=your_openai_api_key
   
   # Snowflake Connection (optional, can also be entered in the UI)
   SNOWFLAKE_ACCOUNT=your_account
   SNOWFLAKE_USER=your_username
   SNOWFLAKE_PASSWORD=your_password
   SNOWFLAKE_WAREHOUSE=your_warehouse
   SNOWFLAKE_DATABASE=your_database
   SNOWFLAKE_SCHEMA=your_schema
   

Running the Application

1. Start the Streamlit app:

   streamlit run app/app.py

2. Open your browser and navigate to http://localhost:8501

3. If you didn't set up Snowflake credentials in the .env file, you'll need to enter them in the sidebar.

4. Enter your OpenAI API key in the sidebar if not already provided in the .env file.

5. Select a table from the dropdown and start asking questions!

Usage

1. Connect to your Snowflake database using the connection form in the sidebar
2. Select a table to query from the dropdown menu
3. Enter a natural language question in the chat input, such as:
   • "How many rows are in this table?"
   • "What are the top 5 entries sorted by date?"
   • "Show me records where the value is greater than 100"
4. View the results in the table view
5. Download results as CSV or Excel using the provided buttons

Technology Stack

• Streamlit - UI framework
• LangChain - LLM framework
• LangGraph - Agent orchestration
• Snowflake Snowpark - Database connectivity
• OpenAI GPT - Natural language processing

Examples of Natural Language Queries

• "How many records are in this table?"
• "What are the top 5 customers by order value?"
• "Show me sales data from last month"
• "What is the average transaction amount by product category?"
• "Find all records where the status is 'pending'"

Security Considerations

This application includes several security features:

• Prevention of DROP, DELETE, and UPDATE operations
• Result set size limitations
• Input validation
• Secure credential handling

Troubleshooting

• If you encounter connection issues, ensure your Snowflake credentials are correctly configured
• For SQL generation errors, try rephrasing your question to be more specific
• If export fails, check if your result set is within the size limitations

Development

This application is built with:

• Streamlit
• LangChain
• Snowflake Snowpark Python
• LangGraph
• OpenAI's language models (GPT-4o-mini by default)