ai-ecommerce-agent

An AI-powered shopping assistant leveraging FastAPI, AutoGen, and Ollama with a React ChatUI frontend. The agent can search the product catalog using FAISS embeddings and interact with GrandNode to manage carts and orders.

Running Locally

1. Copy .env.template to .env and adjust values if needed.
2. Build and start all services:

docker-compose up --build

• Frontend: http://localhost:3000
• Backend: http://localhost:8000
• GrandNode: http://localhost:5000
• Ollama: http://localhost:11434

Embedding the Product Catalog

Run the FAISS loader inside the backend container to generate an index from the GrandNode MongoDB:

python backend/tools/faiss_loader.py

This stores vectors in faiss.index which the agent uses for product retrieval.

Testing the Chat Flow

Open the frontend and ask questions such as:

• "Compare the Commander and Z-10 models"
• "Find a rake for a 2-acre property with wet leaves"
• "Add Commander to my cart"
• "Place the order with standard shipping"
• "Where's my order CYL45678?"

The backend /agent endpoint handles chat messages and communicates with GrandNode.

Connecting to GrandNode

The backend talks to GrandNode via REST API defined in grandnode_client.py. Update the GRANDNODE_URL environment variable if GrandNode runs elsewhere. The MongoDB connection string is set with MONGO_URI.