job-search-tool

What it does

• companies.yaml queries ATS APIs directly (Greenhouse, Ashby, Workable, Lever, SmartRecruiters) for low noise; aggregators.yaml runs keyword search (Adzuna, Remotive) for reach — both file-driven, retarget by editing the file; filters.yaml holds title, location and stack rules with no code changes
• profile.yaml is gitignored by design and never committed; profile.sample.yaml shows the level of quantified evidence each bullet should carry
• AI step is optional and per-candidate: python -m app.ai_evaluate --dry-run queues nothing and costs nothing, --limit 20 sanity-checks quality and cost; scored results land in data/scored_candidates.csv sorted best-match-first
• make web starts the Next.js board reading and writing data/seen_jobs.sqlite3 directly — no export/import step; Remotive needs no auth but covers remote roles only, Adzuna needs a free developer key

When to use it

Engineers running a targeted, weeks-long search who want genuinely new matching roles surfaced automatically each run without job-board noise — and a written fit rationale before spending an application.

Getting started

git clone https://github.com/fedorchenko-juli/job_search_tool.git
cd job_search_tool
pip install -r requirements.txt
cp .env.example .env                    # fill in the keys you need
cp profile.example.yaml profile.yaml    # then edit in your own background
python -m app.main                      # fetch + filter + dedup
python -m app.ai_evaluate --dry-run     # preview the AI step, no cost
make web                                # review board on localhost:3000

Things to know

The repo ships pre-configured for the author's own search (Alberta/Canada, Python/JS stack) — edit companies.yaml, aggregators.yaml, filters.yaml and profile.yaml before the first run or you will fetch their search. The AI evaluation step costs money per candidate. Company ATS endpoints and aggregator APIs can change without notice. MIT.