Multi-Agent Retail Inventory Optimization System

📌 Project Overview

This project leverages a multi-agent architecture integrated with local LLMs (Ollama) to optimize retail inventory management. Three autonomous agents—StoreAgent, WarehouseAgent, and SupplierAgent—collaborate to:

• Monitor sales trends and inventory levels
• Forecast demand and suggest pricing strategies
• Automate decision-making for reordering and supplier communication

The agents use structured data and dynamic prompts to interact with a locally hosted LLM (phi via Ollama) for context-aware retail decisions.

---

⚙️ Setup Instructions

1. Clone the Repository

git clone https://github.com/your-username/multi-agent-inventory.git
cd multi-agent-inventory

2. Create a Virtual Environment (Optional but Recommended)

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4. Download Ollama and Run the Model

• Install Ollama
• Pull the Phi model:

ollama run phi

• Keep Ollama running locally at http://localhost:11434

---

🚀 How to Run Agents

Main File

python agents/main.py

Store Agent

python agents/store_agent.py

Warehouse Agent

python agents/warehouse_agent.py

Supplier Agent

python agents/supplier_agent.py

You can simulate a particular day by modifying the simulate_day(day_number) function call.

---

🧠 Ollama Integration

The agents communicate with the local Ollama LLM via REST API. Each agent:

• Extracts daily data from CSV files
• Formats a natural language prompt
• Sends the prompt to the Ollama server (phi model)
• Parses the streaming NDJSON response for decision output

Example API call:

response = requests.post(
    "http://localhost:11434/api/generate",
    json={"model": "phi", "prompt": prompt},
    stream=True
)

---

📊 Example Output

[StoreAgent] Day 1 Decision:
Based on high sales volume and good customer reviews, consider a slight price increase and marketing promotion.

[WarehouseAgent] Day 1 Decision:
Stock levels are low and close to the reorder point. Reorder today to avoid stockouts.

[SupplierAgent] Day 1 Decision:
Prepare a shipment based on historical lead times and warehouse capacity. Ensure timely dispatch.

---

📬 Contact

Dhanashri Patil
📧 patil.dhanashrik@gmail.com
🐙 DhanashriPatil11
🔗 dhanashri-patil24

---

This project was built as part of the Hackathon challenge: Optimizing Retail Inventory with Multi Agents 💡