browser-use — personal workspace

This repo is a working copy of browser-use with a small custom script that drives a logged-in Chrome profile to explore a web app and produce structured product documentation.

What’s here

• my_agent.py — Prompts an agent to open searchfit.ai, sign in with email/password from environment variables, explore the product, and output markdown product documentation only (no step-by-step narration).
• browser_use/ — The library source (same layout as upstream).

Requirements

• Python 3.11+
• Google Chrome (system install) — the script uses Browser.from_system_chrome and lets you pick a Chrome profile directory.
• An OpenAI API key (default model in script: gpt-4o-mini).
• A .env file in the project root (never commit it).

Setup

cd /path/to/this/repo
python -m venv .venv
# Windows: .venv\Scripts\activate
pip install -e .

Create .env:

OPENAI_API_KEY=sk-...
SEARCHFIT_EMAIL=your@email.com
SEARCHFIT_PASSWORD=your_password

Run

python my_agent.py

You will be prompted to choose a Chrome profile. Use one that can reach the site if you rely on saved session cookies; otherwise the script will use the credentials from .env.

Security

• Keep .env out of git (it should be in .gitignore).
• The agent marks login secrets as sensitive data in code so they are not echoed back by the model in normal operation; still treat the repo and logs as sensitive if they contain credentials.

Upstream

Library docs and community resources live on the official browser-use repository and docs.browser-use.com.