Job Agent — Setup

Fill in the blanks, in order. Takes about 15 minutes.

1. Push to GitHub

cd job-agent
git init
git add .
git commit -m "Initial job agent"
gh repo create ______________ --public --source=. --push

2. Get a free Adzuna API key

Sign up: https://developer.adzuna.com/

app_id:  ______________
app_key: ______________

3. Fill in your real profile (kept private)

Copy this, fill in your values, keep it somewhere for step 4 — do not commit it to the repo:

{
  "name": "______________",
  "salary_floor": ______________,
  "excluded_companies": [______________],
  "preferred_locations": [______________]
}

Any key from config/profile.py can go in here (skills, target_roles, priority_topics, etc.) — only include what you want to override.

4. Add repo secrets

Settings → Secrets and variables → Actions → New repository secret:

ADZUNA_APP_ID        = ______________  (from step 2)
ADZUNA_APP_KEY        = ______________  (from step 2)
PROFILE_OVERRIDE_JSON = ______________  (the JSON from step 3, if you want to keep things incog)

5. Enable GitHub Pages

Settings → Pages → Source: Deploy from a branch → Branch: main, folder: /docs → Save.

Dashboard URL: https://<user_handle>.github.io/<repo_name>/

6. Run it once

Actions tab → Daily job scan → Run workflow. Wait ~30 sec, then refresh the dashboard URL — it should have real data.

---

That's it — it reruns on its own daily. To change the time, edit the cron line in .github/workflows/daily-job-scan.yml.

Optional: AI-judged scoring instead of just keyword matching

By default, ranking is pure keyword matching (free, no setup). To have the top 100 matches also get judged by Claude for actual fit against your resume:

1. Get a key: https://console.anthropic.com  (enable billing — cost at
   this volume is cents to low dollars/month using Haiku, the default)
2. Add resume_text to the JSON from step 3 above:
     "resume_text": "______________"   (paste your real resume text)
3. Add one more repo secret:
     ANTHROPIC_API_KEY = ______________

No other change needed — src/main.py picks it up automatically next run. Jobs that get an AI score show it as the primary number with an "AI" tag, plus a one-line reason; the keyword score stays visible underneath. Everything past the top 100 stays keyword-scored either way.