SlothScout: a local agent that reads my job alerts at 3 a.m. so I don't have to
Описание и применение
Автоматический агент SlothScout обрабатывает письма с вакансиями, извлекает предложения, удаляет дубликаты, оценивает их локальной моделью Gemma 3 4B и выдаёт утром краткое резюме, экономя около 3 часов в день.
Задача
Spending about 4 hours a day reviewing job‑alert emails and finding that many listings are irrelevant or duplicate.
Результат
In the first run, 15 alert emails yielded 49 job postings, 1 scored above the 70 cutoff, and the entire process took 152 seconds. The agent saved about 3 hours of daily manual filtering.
Рабочий процесс
- Schedule a Spring @Scheduled job at 03:00 America/Sao_Paulo
- Read only the Vagas Gmail label via IMAP
- Track processed emails by Message‑ID in H2 DB
- Parse each sender’s email with Jsoup to extract job link and ID
- Deduplicate postings by source:jobId
- Score each posting with Gemma 3 4B via Ollama using a structured JSON schema
- Generate a two‑line summary with a second Gemma call
- Serve report as Markdown and Thymeleaf page
- Provide a run‑now button for immediate execution
Доказательства
I was spending about 4 hours a day on them: reading the alerts, opening each site, reading the requirements, and finding out (again) that "Junior" meant 5 years of experience.
My first real run: Alert emails read 15, Job postings scored 49, Above my cutoff (70) 1, Total time 152 seconds
The lazy payoff: it saves me about 3 hours a day.
- Дата записи: 2026-10-24T23:11:11.592Z