AutoShopper AI

AutoShopper AI is an autonomous shopping assistant that converts natural-language requests into real online actions.
Given a user prompt (e.g., “Find me an iced matcha with oat milk under $12 near 45 Fremont St”), the system:

• Interprets the request
• Navigates the web using an LLM-guided browser agent
• Searches nearby cafés and delivery platforms
• Evaluates options based on constraints such as price, distance, and rating
• Builds the cart and prepares checkout
• Returns a structured summary and an actionable checkout link

The project demonstrates a practical, end-to-end autonomous web-navigation agent capable of completing real-world tasks.

---

Features

Autonomous Browser Agent

• Uses Browser-Use for LLM-guided automation
• Interacts with Google Maps, DoorDash, UberEats, and vendor sites

End-to-End Task Execution

• Understands intent
• Locates nearby vendors
• Applies constraints
• Builds order up to checkout (without submitting payment)

Structured JSON Output

The backend returns predictable JSON containing:

• Restaurant name
• Selected item
• Total price
• Estimated delivery time
• Checkout link
• Additional metadata

Modern Frontend Interface

• Fully responsive
• Animated UI
• 3D visuals using Three.js
• Timeline showing agent progress
• Clear result card with checkout link

API-Driven Architecture

• Frontend communicates with a lightweight REST API
• Browser automation and decision logic run in the backend

---

Tech Stack

Frontend

• HTML, CSS, JavaScript
• Three.js
• Glassmorphism & responsive design

Backend

• Python 3.10
• FastAPI
• Playwright (Chromium)
• Browser-Use agent

Optional Integrations

• Sentry
• .env configuration