Amazon Automator

Lightweight prototype that lets you describe a goal (e.g. "search RTX 4090 and sort by price high to low") and an LLM (Gemini 2.5 Pro) iteratively drives a real Chromium browser on Amazon using Playwright.

What it does

• Opens Amazon in a visible browser
• Captures reduced or full DOM each step
• Asks the model for the next 1–4 Playwright actions until DONE or step limit
• Shows you the executed action history in the web UI

Requirements

• Python 3.9+
• Conda (optional) or virtualenv
• A valid GEMINI_API_KEY in a .env file

Example .env:

GEMINI_API_KEY=your_key_here

Install

conda create -n amazon_automator_env python=3.9 -y
conda activate amazon_automator_env
pip install -r requirements.txt
playwright install

Run

uvicorn main:app --reload

Open http://127.0.0.1:8000 and enter a goal.

Notes

• The model may hit rate limits; it auto-retries a few times.
• Generated actions are sandboxed; unsafe patterns are filtered.
• Increase MAX_STEPS or adjust MAX_CHARS_PER_CHUNK in main.py if needed.

Disclaimer

Prototype quality; expect occasional selector or quota failures.