Automation ·2025 · Live
Job-hunting bot with fit scoring
Collects job posts from several boards, de-duplicates against SQLite, scores how well each one fits the profile and pings Telegram only with what is worth reading.
The problem was never finding openings, it was re-reading the same hundred. The bot queries several boards in parallel with per-role queries and stores everything seen in SQLite, so each run only surfaces what is new.
Before notifying it applies a local 0-100 fit filter that looks at the role in the title, the technologies, the seniority (penalising senior, lead and manager) and geographic eligibility for remote work. Anything below the threshold is dropped for free.
Only the candidates that clear the filter optionally go through the Claude API for a commented score and a draft application.
Highlights
- Persistent de-duplication in SQLite: never shows the same post twice
- Free local scoring before spending any AI tokens
- A --solo-buscar dry-run mode to calibrate thresholds with no side effects
- Notifications over Telegram, email or console
- Cover letter and email draft generation for the best matches
Stack
- Python
- SQLite
- Claude API
- Telegram Bot API
- YAML
- requests
Private repository — Code available on request