Multi-Agent Orchestration for Data Analysis & Simulation Scenarios

A sophisticated data analysis and simulation platform using a multi-agent workflow powered by Chainlit and Mistral AI. This application enables users to perform complex data queries and run what-if scenario simulations through natural language interactions.

Current Status

Desired Status

🚀 Features

• Natural Language Querying: Ask data questions in plain English
• Automatic SQL Generation: AI converts questions to SQL queries
• Scenario Simulation: Run what-if analyses on your data
• Visualization: Automatic chart generation for results
• Multi-Agent Architecture: Specialized agents for different tasks
• Interactive UI: User-friendly Chainlit interface

📋 Table of Contents

• Installation
• Usage
• Demos
• Architecture
• Roadmap

🛠️ Installation

Prerequisites

• Python 3.11+
• UV for package management (recommended)
• Mistral API key

Steps

1. Clone the repository:

2. Set up environment variables:

   • Copy .env.example to .env
   • Add your Mistral API key to .env

3. Create and activate virtual environment:

   uv venv
   source .venv/bin/activate  # Linux/macOS
   # .\.venv\Scripts\activate  # Windows

   uv venv
   source .venv/bin/activate  # Linux/macOS
   # .\.venv\Scripts\activate  # Windows

4. Install dependencies:

   uv sync

   uv sync

5. Run the application:

   uv run -- chainlit run app.py

   uv run -- chainlit run app.py

🎛️ Usage

Starting the Application

1. Run the Chainlit app as shown above
2. Open the provided URL in your browser
3. Use the starter questions or ask your own

Example Queries

• "Show total estimated revenue across all accounts in January 2025"
• "List the top 5 account names by available balance"
• "What if we raise deposit rates by 0.5%?"

Workflow Visualization

Type "Show workflow" to see the system architecture diagram.

📊 Demos

Demo Script

Data Analysis

Scenario Simulation

🏛️ Architecture

The system uses a multi-agent approach:

1. Router Agent: Determines query intent
2. Analysis Agent: Handles data queries
3. Simulation Agent: Runs what-if scenarios
4. Report Agent: Generates visualizations

graph TD;
    A[User] --> B[Router Agent];
    B -->|Data Query| C[Analysis Agent];
    B -->|Simulation| D[Simulation Agent];
    C --> E[Database];
    D --> E;
    C --> F[Report Agent];
    D --> F;
    F --> A;

graph TD;
    A[User] --> B[Router Agent];
    B -->|Data Query| C[Analysis Agent];
    B -->|Simulation| D[Simulation Agent];
    C --> E[Database];
    D --> E;
    C --> F[Report Agent];
    D --> F;
    F --> A;

🗺️ Roadmap

• Add voice input support
• Implement more advanced simulation models
• Add user authentication
• Support additional data sources