Skip to content

Latest commit

 

History

History
112 lines (74 loc) · 4.07 KB

README.md

File metadata and controls

112 lines (74 loc) · 4.07 KB

CorsaiFlow Crew

Welcome to the CorsaiFlow Crew project, powered by crewAI. This project demonstrates a multi-agent AI system that integrates with Snowflake's Model Context Protocol (MCP) to intelligently route and execute data queries across multiple data repositories.

Features

  • Snowflake MCP Integration: Connect to Snowflake's managed MCP server to access Cortex Search, Cortex Analyst, SQL execution, and Cortex Agents
  • Intelligent Query Routing: Automatically determine which data repository is best suited for a user's query
  • Multi-Agent Workflow: Two specialized agents work together to route queries and retrieve data
  • Flow-Based Architecture: Built on CrewAI's flow system for orchestrated task execution

Installation

Ensure you have Python >=3.10 <3.14 installed on your system. This project uses UV for dependency management and package handling, offering a seamless setup and execution experience.

First, if you haven't already, install uv:

pip install uv

Next, navigate to your project directory and install the dependencies:

(Optional) Lock the dependencies and install them by using the CLI command:

crewai install

Environment Setup

Add your API keys and Snowflake configuration to the .env file:

OPENAI_API_KEY=your_openai_api_key

# Snowflake MCP Configuration
SNOWFLAKE_MCP_SERVER_URL=https://your-account.snowflakecomputing.com
SNOWFLAKE_DATABASE=your_database
SNOWFLAKE_SCHEMA=your_schema
SNOWFLAKE_MCP_SERVER_NAME=your_mcp_server
SNOWFLAKE_ACCESS_TOKEN=your_oauth_access_token

See SNOWFLAKE_MCP_SETUP.md for detailed setup instructions.

Customizing

  • Modify src/corsaiflow/crews/data_routing_crew/config/agents.yaml to customize agent behavior
  • Modify src/corsaiflow/crews/data_routing_crew/config/tasks.yaml to customize task definitions
  • Update src/corsaiflow/crews/data_routing_crew/data_routing_crew.py to configure data repositories
  • Modify src/corsaiflow/main.py to add custom flows and inputs

Running the Project

To kickstart your flow and begin execution, run this from the root folder of your project:

crewai run

This command initializes the corsaiFlow Flow as defined in your configuration.

Running the Data Query Flow

To run the Snowflake data query flow:

# Using Python directly
python -m corsaiflow.main query_snowflake_data

# Or with a trigger payload
python -m corsaiflow.main query_snowflake_with_trigger '{"user_query": "What products are in stock?"}'

The default example will route a query about product inventory and retrieve data from the appropriate Snowflake repository.

Understanding Your Crew

Data Routing Crew

The Data Routing Crew consists of two specialized agents:

  1. Query Router Agent: Analyzes user queries and determines which Snowflake data repository (Cortex Search, Cortex Analyst, SQL database, or Cortex Agent) is most appropriate for answering the question.

  2. Data Retriever Agent: Executes queries against the selected Snowflake repository using MCP tools, formats the results, and presents them to the user.

Crew Structure

  • Agents: Defined in src/corsaiflow/crews/data_routing_crew/config/agents.yaml
  • Tasks: Defined in src/corsaiflow/crews/data_routing_crew/config/tasks.yaml
  • Tools:
    • SnowflakeMCPTool: Wraps Snowflake MCP server interactions
    • DataRepositoryTool: Helps select appropriate data repositories

Flow Architecture

The DataQueryFlow orchestrates the query process:

  1. Receives user query
  2. Routes query to appropriate repository
  3. Retrieves data from Snowflake
  4. Formats and returns results

Support

For support, questions, or feedback regarding the {{crew_name}} Crew or crewAI.

Let's create wonders together with the power and simplicity of crewAI.