Multi-Agent Business Intelligence System

AI-powered business intelligence platform with collaborative multi-agent architecture for enterprise analytics and market research.

Overview

Dual AI Agents that work independently or collaboratively:

• Enterprise Intelligence: Store performance, inventory, business policies
• Market Intelligence: Demographics, market research, competitive analysis

Interfaces: Streamlit web app + CLI with real-time agent execution tracking

Architecture

Agents & Tools

Enterprise Agent

• Store performance analysis (Databricks Genie)
• Product inventory tracking (Databricks Genie)
• Business policy lookup (UC function)

Market Agent

• Geographic demographics (Census API)
• Web research (Perplexity AI)

Collaboration: Agents can call each other as tools for complex multi-step analysis.

Tech Stack

• Python 3.10+, OpenAI Agents, Streamlit
• Databricks (ML platform), MLflow (tracing)
• Census API (demographics), Perplexity AI (research)

Installation

1. Setup

git clone <repository-url>
uv venv --python 3.12 && source .venv/bin/activate
uv pip install -r requirements.txt

2. Environment - Create .env:

# Databricks
DATABRICKS_HOST=https://your-workspace.cloud.databricks.com
DATABRICKS_TOKEN=your_personal_access_token
DATABRICKS_MODEL=your_model_name
GENIE_SPACE_ID=your_store_space_id
GENIE_SPACE_PRODUCT_INV_ID=your_inventory_space_id

# APIs
CENSUS_API_KEY=your_census_api_key
PERPLEXITY_API_KEY=your_perplexity_api_key

Usage

Web Interface

streamlit run app.py

CLI

# Single query
python multi_agent_cli.py --query "What are the demographics around store 110?"

# Interactive mode
python multi_agent_cli.py --interactive

Example Queries

Enterprise: "Performance of store 110 vs region", "Inventory levels for product XYZ" Market: "Demographics around Chicago stores", "Retail technology competitive landscape" Combined: "Store 110's location demographics for product mix optimization"

Project Structure

├── app.py                    # Streamlit web app
├── multi_agent_cli.py        # CLI interface
├── toolkit.py               # Custom tools
├── prompts/                 # Agent instructions
└── requirements.txt         # Dependencies

Prerequisites

• Databricks workspace with Genie spaces and model serving endpoint
• Census API key for demographic data
• Perplexity API key for web research

Troubleshooting

• Auth errors: Verify Databricks token/permissions
• Timeouts: Check workspace connectivity
• Missing deps: Ensure virtual environment setup
• Config: Verify all environment variables set