Job Hunter

A personal automation pipeline that scrapes job vacancies, filters and scores them with AI, and delivers the best matches via Telegram.

How it works

Vacancies move through three layers, each with a clear responsibility:

Bronze  →  raw scraped vacancies
Silver  →  rule-filtered, profile-matched, ready for AI
Gold    →  AI-scored, ranked, ready for notification

The candidate profile lives in profiles.json and drives both the Silver filtering rules and the AI scoring prompt. The current implementation covers the full Bronze layer and profile loading — Silver promotion rules and AI scoring are next.

Project structure

File	Responsibility
main.py	Pipeline entrypoint
cathoScraper.py	Catho vacancy scraper (Playwright)
database.py	MongoDB connection, collections, indexes
user_profile.py	Profile loading and prompt generation
profiles.json	Candidate profile config
ai_scoring.py	(reserved — AI scoring)
telegram.py	(reserved — Telegram delivery)
notification.py	(reserved — notification tracking)

Setup

Requirements: Python 3.10+, a MongoDB Atlas cluster, a Playwright-compatible environment.

Environment variables (.env at project root):

MONGODB_URL=your_mongodb_connection_string
MONGODB_DB=JobHunter

Install:

pip install pymongo python-dotenv playwright
playwright install

Run:

python3 main.py

This scrapes Catho vacancies, builds vacancy documents, and inserts them into MongoDB with duplicate URL protection.

What's working

• Catho scraping via Playwright
• MongoDB Atlas connection via environment variables
• Vacancy insertion with duplicate URL blocking
• Profile loading from profiles.json
• Profile-to-prompt conversion for the AI layer

Roadmap

• Deterministic Bronze → Silver promotion rules
• AI scoring in ai_scoring.py
• Persist scored vacancies to Gold
• Telegram notifications
• Scheduling and monitoring

Notes

• .env is gitignored
• profiles.json stores candidate configuration, not application logic