This is scraping a specific university's LMS through a real browser, which means it works exactly as well as that site's markup allows and will need attention when the site changes. The parts worth borrowing regardless are the session handling — persisted, encrypted, with an expiry — and the two-tier CAPTCHA approach that falls back to OCR when no vision key is configured. Your student credentials sit in a local file, so treat the machine accordingly.
A browser-automation server for one university's learning management system. It logs in on your behalf, solving the CAPTCHA on the way, keeps the session, and exposes the student pages — attendance, marks, timetable and applications — as tools.
- `login_to_lms` signs in with your stored credentials and handles the CAPTCHA; `logout_lms` ends the session
- `check_attendance` returns attendance records broken down by subject
- `get_marks` returns exam scores and assessments
- `get_timetable` returns the class schedule
- `create_application` submits a university application and `view_application_status` checks where it got to
- `solve_captcha` for the case where the automatic pass needs help
- Sessions persist and restore between runs, stored encrypted, with an expiry you set
Python 3.11 or higher, Git and a Chrome installation, since the work is done by driving a real browser. Install with `uv sync` or `pip install -e .`, then put your portal credentials in a `.env` as `LMS_USERNAME` and `LMS_PASSWORD`. `GEMINI_API_KEY` is optional and improves CAPTCHA accuracy through vision; without it the fallback is OCR. `BROWSER_HEADLESS` controls whether you watch it work, `SESSION_EXPIRY_HOURS` defaults to 24 and `ENCRYPT_SESSIONS` keeps the stored session encrypted.