Initial deploy setup

tests/benchmark/test_benchmark.py

@@ -0,0 +1,39 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+import pytest
+
+import scripts.benchmark as benchmark
+
+
+def test_load_benchmark_dataset():
+    dataset = benchmark.load_benchmark_dataset(Path("fixtures/benchmark/quality_benchmark.json"))
+    assert dataset["dataset_version"] == "vr116-2026-07-21"
+    assert isinstance(dataset["parser_coverage"], list)
+    assert isinstance(dataset["dedupe"], dict)
+
+
+def test_percentile_helper_is_deterministic():
+    assert benchmark._percentile([1.0, 2.0, 3.0, 4.0], 50) == 2.0
+    assert benchmark._percentile([], 95) == 0.0
+
+
+@pytest.mark.django_db
+def test_run_benchmark_report_smoke(db):
+    report = benchmark.run_benchmark_report(
+        Path("fixtures/benchmark/quality_benchmark.json"),
+        quick=True,
+        top_n=20,
+        jobs=25,
+        skip_performance=True,
+    )
+
+    assert report["status"] in {"passed", "failed"}
+    assert report["jobs_processed"] == 25
+    assert report["top_n"] == 20
+    assert report["performance"]["status"] == "skipped"
+    assert report["nfr_007"]["measured"] is False
+    assert set(report["hardware"].keys()) >= {"platform", "python", "cpu_count"}
+    assert report["parser"]["coverage_ratio"] >= 0
+    assert report["ranking"]["top_n"] == 20

tests/conftest.py

@@ -0,0 +1,97 @@
+from __future__ import annotations
+
+from decimal import Decimal
+
+import pytest
+from django.contrib.auth import get_user_model
+
+from apps.profiles.models import SearchProfile
+from apps.sources.models import Source
+
+
+@pytest.fixture
+def user(db):
+    return get_user_model().objects.create_user(
+        username="jens",
+        email="jens@example.invalid",
+        password="correct-horse-battery-staple",
+    )
+
+
+@pytest.fixture
+def profile(user):
+    return SearchProfile.objects.create(
+        user=user,
+        name="Testprofiel",
+        is_active=True,
+        home_municipality="Hasselt",
+        home_latitude=Decimal("50.930700"),
+        home_longitude=Decimal("5.332500"),
+        max_distance_km=50,
+        learning_enabled=True,
+        desired_titles=["infrastructure engineer", "systeembeheerder"],
+        excluded_titles=["sales", "recruiter"],
+        desired_skills=["Microsoft 365", "VMware"],
+        allowed_employment_types=["full_time", "permanent"],
+        preferred_workplace=["hybrid", "on_site"],
+    )
+
+
+@pytest.fixture
+def source(db):
+    return Source.objects.create(
+        name="Example jobs",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+
+
+@pytest.fixture
+def employer(db):
+    from apps.jobs.models import Employer
+
+    return Employer.objects.create(
+        name="Example IT",
+        normalized_name="example it",
+        domain="jobs.example.org",
+        is_direct_employer=True,
+        is_recruiter=False,
+        confidence=0.9,
+    )
+
+
+@pytest.fixture
+def job(db, employer):
+    from apps.jobs.models import JobPosting
+
+    return JobPosting.objects.create(
+        employer=employer,
+        original_title="Infrastructure Engineer",
+        normalized_title="infrastructure engineer",
+        job_family="infrastructure",
+        canonical_url="https://jobs.example.org/vacatures/infrastructure-engineer",
+        canonical_key="c" * 64,
+        content_hash="d" * 64,
+        description_text=(
+            "Beheer Microsoft 365 en VMware. Hybride werk met beperkte tweedelijnssupport."
+        ),
+        raw_location="Hasselt, Limburg",
+        region="Limburg",
+        municipality="Hasselt",
+        workplace_type=JobPosting.Workplace.HYBRID,
+        employment_types=["full_time", "permanent"],
+        skills_required=["Microsoft 365"],
+        skills_preferred=["VMware"],
+        direct_employer=True,
+        recruiter=False,
+        extraction_confidence=0.9,
+        analysis_features={
+            "support_ratio": 0.1,
+            "public_sector_signal": 0.0,
+            "experience_years_max": 3,
+        },
+        status=JobPosting.Status.ACTIVE,
+    )

tests/e2e/test_vr114_browser_and_accessibility.py

@@ -0,0 +1,251 @@
+from __future__ import annotations
+
+from datetime import timedelta
+from decimal import Decimal
+from pathlib import Path
+
+import pytest
+from bs4 import BeautifulSoup
+from django.urls import reverse
+from django.utils import timezone
+
+from apps.jobs.models import Application, Feedback, ScoreRun
+
+try:
+    from playwright.sync_api import sync_playwright
+except Exception:  # pragma: no cover - optional dependency
+    sync_playwright = None
+
+
+def _seed_dashboard_score(profile, job) -> None:
+    ScoreRun.objects.create(
+        job=job,
+        profile=profile,
+        profile_version=profile.version,
+        score=Decimal("88.2"),
+        confidence=Decimal("0.76"),
+        recommendation=ScoreRun.Recommendation.STRONG,
+        components={"content": 42.0, "skills": 23.0, "location": 23.2},
+        positives=["Relevante locatie", "Ervaring met ops"],
+        concerns=[],
+        hard_exclusions=[],
+        evidence={"tests": ["vr-114"]},
+        model_version="vr-114",
+        prompt_version="1.0.0",
+    )
+
+
+def _assert_basic_html_accessibility(payload: bytes | str, *, page_name: str) -> None:
+    html = payload.decode("utf-8", errors="replace") if isinstance(payload, bytes) else payload
+    soup = BeautifulSoup(html, "html.parser")
+
+    headings = soup.find_all("h1")
+    assert headings, f"{page_name}: ontbreekt h1"
+    assert len(headings) == 1, f"{page_name}: verwacht exact één h1, gevonden {len(headings)}"
+
+    assert soup.html is not None
+    assert soup.html.get("lang"), f"{page_name}: <html> mist lang-attribuut"
+
+    for field in soup.find_all(["input", "textarea", "select"]):
+        if field.get("type") == "hidden" or field.get("name") == "csrfmiddlewaretoken":
+            continue
+
+        field_id = field.get("id")
+        has_label = bool(field.find_parent("label"))
+        has_label = has_label or bool(field.get("aria-label") or field.get("aria-labelledby"))
+        has_label = has_label or (field_id is not None and bool(soup.find("label", attrs={"for": field_id})))
+
+        assert has_label, (
+            f"{page_name}: formulierveld zonder label of aria-voorziening: "
+            f"{field.get('name') or field.get('type')}"
+        )
+
+    for button in soup.find_all("button"):
+        label = button.get_text(" ", strip=True) or button.get("aria-label")
+        assert label, f"{page_name}: knop zonder tekst of aria-label"
+
+
+def _run_html_accessibility_probe(client, user, profile, job) -> None:
+    _seed_dashboard_score(profile=profile, job=job)
+
+    response = client.get(reverse("login"))
+    assert response.status_code == 200
+    _assert_basic_html_accessibility(response.content, page_name="login")
+
+    client.force_login(user)
+
+    response = client.get(reverse("dashboard:today"))
+    assert response.status_code == 200
+    _assert_basic_html_accessibility(response.content, page_name="dashboard")
+
+    response = client.get(reverse("jobs:list"), {"q": job.original_title})
+    assert response.status_code == 200
+    _assert_basic_html_accessibility(response.content, page_name="jobs-list")
+
+    response = client.get(reverse("jobs:detail", kwargs={"pk": job.pk}))
+    assert response.status_code == 200
+    _assert_basic_html_accessibility(response.content, page_name="jobs-detail")
+
+    response = client.get(reverse("profiles:list"))
+    assert response.status_code == 200
+    _assert_basic_html_accessibility(response.content, page_name="profiles-list")
+
+    response = client.get(reverse("profiles:edit", kwargs={"pk": profile.pk}))
+    assert response.status_code == 200
+    _assert_basic_html_accessibility(response.content, page_name="profiles-edit")
+
+    response = client.get(reverse("jobs:applications"))
+    assert response.status_code == 200
+    _assert_basic_html_accessibility(response.content, page_name="applications")
+
+    response = client.get(reverse("sources:list"))
+    assert response.status_code == 200
+    _assert_basic_html_accessibility(response.content, page_name="sources")
+
+
+def _assert_playwright_accessibility(page, *, expected_h1: str) -> None:
+    page.wait_for_load_state("networkidle")
+    assert page.get_by_role("heading", level=1).count() == 1
+    assert expected_h1 in page.get_by_role("heading", level=1).all_text_contents()
+
+    page.keyboard.press("Tab")
+    focused_tag = page.evaluate("document.activeElement.tagName.toLowerCase()")
+    assert focused_tag != "body"
+    outline_style = page.evaluate("getComputedStyle(document.activeElement).outlineStyle")
+    assert outline_style != "none"
+
+
+def _take_failure_screenshot(page, tmp_path: Path, viewport_name: str) -> None:
+    directory = tmp_path / "artifacts"
+    directory.mkdir(parents=True, exist_ok=True)
+    page.screenshot(path=str(directory / f"vr114_e2e_failure_{viewport_name}.png"), full_page=True)
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_vr114_html_a11y_probe(client, user, profile, job):
+    _run_html_accessibility_probe(client, user, profile, job)
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+@pytest.mark.parametrize(
+    "viewport",
+    [
+        {"name": "desktop", "width": 1366, "height": 900},
+        {"name": "mobile-360", "width": 360, "height": 780},
+    ],
+)
+def test_vr114_browser_flow(viewport, live_server, client, tmp_path, user, profile, job):
+    _seed_dashboard_score(profile=profile, job=job)
+
+    if sync_playwright is None:
+        _run_html_accessibility_probe(client, user, profile, job)
+        pytest.skip(
+            "Playwright niet geïnstalleerd. Vult alleen de HTML/a11y-probe in."
+        )
+
+    base = live_server.url
+    dashboard_url = f"{base}{reverse('dashboard:today')}"
+    login_url = f"{base}{reverse('login')}"
+    jobs_url = f"{base}{reverse('jobs:list')}"
+    detail_url = f"{base}{reverse('jobs:detail', kwargs={'pk': job.pk})}"
+    profile_list_url = f"{base}{reverse('profiles:list')}"
+    applications_url = f"{base}{reverse('jobs:applications')}"
+    sources_url = f"{base}{reverse('sources:list')}"
+
+    with sync_playwright() as playwright:
+        try:
+            browser = playwright.chromium.launch(headless=True)
+        except Exception as exc:  # pragma: no cover - env-dependent
+            _run_html_accessibility_probe(client, user, profile, job)
+            pytest.skip(f"Playwright-browser niet beschikbaar: {exc}")
+
+        context = browser.new_context(
+            viewport={"width": viewport["width"], "height": viewport["height"]},
+            reduced_motion="reduce",
+        )
+        page = context.new_page()
+
+        try:
+            # Login.
+            page.goto(login_url)
+            _assert_playwright_accessibility(page, expected_h1="Welkom terug")
+            page.get_by_label("Gebruikersnaam").fill(user.username)
+            page.get_by_label("Wachtwoord").fill("correct-horse-battery-staple")
+            page.get_by_role("button", name="Aanmelden").click()
+            page.wait_for_url(dashboard_url)
+            _assert_playwright_accessibility(page, expected_h1="Vandaag")
+
+            # Dashboard -> filter -> detail.
+            page.get_by_role("link", name="Vacatures").click()
+            page.get_by_role("textbox", name="Zoeken").fill(job.original_title)
+            page.get_by_role("button", name="Filter").click()
+            assert page.url.startswith(jobs_url)
+            page.get_by_role("link", name=job.original_title).first.click()
+            page.wait_for_url(detail_url)
+            _assert_playwright_accessibility(page, expected_h1=job.original_title)
+
+            # Vier feedbackacties.
+            page.get_by_role("button", name="Interessant").click()
+            page.wait_for_selector(".message.success")
+            page.get_by_role("button", name="Bewaren").click()
+            page.wait_for_selector(".message.success")
+            page.get_by_label("Reden (optioneel)").select_option("Te ver")
+            page.get_by_role("button", name="Verbergen").click()
+            page.wait_for_selector(".message.success")
+            page.get_by_role("button", name="Gesolliciteerd").click()
+            page.wait_for_selector(".message.success")
+
+            # Profiel aanpassen.
+            page.get_by_role("link", name="Profiel").click()
+            page.wait_for_url(profile_list_url)
+            page.get_by_role("link", name="Bewerken").click()
+            page.get_by_label("Gewenste functietitels").fill(
+                "Infrastructure Engineer\nSysteembeheerder"
+            )
+            page.get_by_role("button", name="Profiel opslaan").click()
+            page.wait_for_url(profile_list_url)
+
+            # Sollicitatie dossier openen en bijwerken.
+            application = Application.objects.filter(user=user, job=job).first()
+            assert application is not None
+            page.get_by_role("link", name="Sollicitaties").click()
+            page.wait_for_url(applications_url)
+            page.goto(f"{base}{reverse('jobs:application-edit', kwargs={'pk': application.pk})}")
+            page.locator("select[name='status']").select_option("interview")
+            page.locator("input[name='contact_name']").fill("Sofie Janssens")
+            page.locator("input[name='contact_email']").fill("sofia@example.invalid")
+            page.locator("input[name='follow_up_date']").fill(
+                (timezone.localdate() + timedelta(days=7)).isoformat()
+            )
+            page.locator("textarea[name='notes']").fill("Eerste follow-up ingepland.")
+            page.get_by_role("button", name="Opslaan").click()
+            page.wait_for_url(applications_url)
+
+            # Handmatige import via plaktekst.
+            page.get_by_role("link", name="Brongezondheid").click()
+            page.wait_for_url(sources_url)
+            page.locator("textarea[name='pasted_text']").fill(
+                "Vacaturetitel: Infrastructure Engineer\n"
+                "Werkgever: Example IT\n"
+                "Locatie: Hasselt"
+            )
+            page.get_by_role("button", name="Import uitvoeren").click()
+            page.wait_for_selector("text=Importresultaat")
+            assert page.get_by_text("Modus: paste").is_visible()
+            assert page.get_by_text("Herkenbare items").is_visible()
+
+            # Keyboardvolgorde en focuseigenschappen controleren.
+            page.keyboard.press("Tab")
+            page.keyboard.press("Tab")
+            page.keyboard.press("Tab")
+            _assert_playwright_accessibility(page, expected_h1="Brongezondheid")
+
+            assert Feedback.objects.filter(user=user, job=job).count() >= 4
+        except Exception:
+            _take_failure_screenshot(page, tmp_path, viewport["name"])
+            raise
+        finally:
+            context.close()
+            browser.close()

tests/integration/test_applications.py

@@ -0,0 +1,191 @@
+from __future__ import annotations
+
+import csv
+import io
+import json
+import zipfile
+from datetime import timedelta
+from decimal import Decimal
+
+import pytest
+from django.contrib.auth import get_user_model
+from django.urls import reverse
+from django.utils import timezone
+
+from apps.jobs.models import Application, ApplicationTimelineEvent, JobSourceAlias, ScoreRun
+from apps.jobs.services.feedback import record_feedback
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_applied_feedback_creates_sanitized_application_snapshot_and_timeline(job, profile, source, user):
+    JobSourceAlias.objects.create(
+        job=job,
+        source=source,
+        url=job.canonical_url,
+        canonical_url=job.canonical_url,
+        external_id="ext-applied-1",
+        source_title="Applied page",
+        source_employer=job.employer_name,
+        is_canonical=True,
+        extraction_method="manual-test",
+    )
+    record_feedback(user=user, job=job, action="applied")
+
+    application = Application.objects.get(user=user, job=job)
+    snapshot = application.snapshot
+    assert snapshot["schema_version"] == "1.0.0"
+    assert snapshot["snapshot_kind"] == "application"
+    assert snapshot["title"] == job.original_title
+    assert snapshot["job"]["id"] == str(job.pk)
+    assert snapshot["job"]["description_html"] == job.description_html_sanitized
+    assert snapshot["sources"][0]["url"] == job.canonical_url
+    assert snapshot["sources"][0]["is_canonical"] is True
+    assert snapshot["scores"] is None
+
+    event_types = set(ApplicationTimelineEvent.objects.filter(application=application).values_list("event_type", flat=True))
+    assert ApplicationTimelineEvent.EventType.CREATED in event_types
+    assert ApplicationTimelineEvent.EventType.SNAPSHOT_CAPTURED in event_types
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_application_update_creates_timeline_events_for_status_notes_and_contact(client, job, profile, source, user):
+    record_feedback(user=user, job=job, action="applied")
+    application = Application.objects.get(user=user, job=job)
+
+    client.force_login(user)
+    response = client.post(
+        reverse("jobs:application-edit", kwargs={"pk": application.pk}),
+        {
+            "status": Application.Status.INTERVIEW,
+            "contact_name": "Sofie Van Hecke",
+            "contact_email": "recruiter@example.invalid",
+            "notes": "Eerste follow-up gepland voor vrijdag.",
+        },
+    )
+    assert response.status_code == 302
+
+    application.refresh_from_db()
+    assert application.status == Application.Status.INTERVIEW
+    assert application.contact_name == "Sofie Van Hecke"
+    assert application.contact_email == "recruiter@example.invalid"
+
+    event_types = set(ApplicationTimelineEvent.objects.filter(application=application).values_list("event_type", flat=True))
+    assert ApplicationTimelineEvent.EventType.STATUS_CHANGED in event_types
+    assert ApplicationTimelineEvent.EventType.NOTES_UPDATED in event_types
+    assert ApplicationTimelineEvent.EventType.CONTACT_UPDATED in event_types
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_application_export_generates_sanitized_zip_payload(client, job, source, user, profile):
+    JobSourceAlias.objects.create(
+        job=job,
+        source=source,
+        url=job.canonical_url,
+        canonical_url=job.canonical_url,
+        external_id="ext-export-1",
+        source_title="Export page",
+        source_employer=job.employer_name,
+        is_canonical=True,
+        extraction_method="manual-test",
+    )
+    ScoreRun.objects.create(
+        job=job,
+        profile=profile,
+        profile_version=1,
+        score=Decimal("84.20"),
+        confidence=Decimal("0.71"),
+        recommendation=ScoreRun.Recommendation.POSSIBLE,
+        components={"content": 0.41},
+        positives=["python"],
+        concerns=["travel"],
+        hard_exclusions=[],
+        evidence={"notes": "Voorbeeldscore"},
+        model_version="testsuite",
+        prompt_version="1.0.0",
+    )
+    record_feedback(user=user, job=job, action="applied")
+    record_feedback(user=user, job=job, action="applied")
+    application = Application.objects.get(user=user, job=job)
+
+    application.notes = "Vaste follow-up notitie."
+    application.save(update_fields=["notes", "updated_at"])
+
+    client.force_login(user)
+    response = client.get(reverse("jobs:application-export", kwargs={"pk": application.pk}))
+    assert response.status_code == 200
+    assert response["Content-Type"] == "application/zip"
+    assert 'filename="' in response["Content-Disposition"]
+
+    filename = response["Content-Disposition"].partition("filename=")[2].strip().strip('"')
+    assert filename.startswith(f"application-{application.pk}-")
+    assert filename.endswith(".zip")
+    assert ".." not in filename and "/" not in filename and "\\" not in filename
+
+    archive = zipfile.ZipFile(io.BytesIO(response.content))
+    members = sorted(archive.namelist())
+    assert members == ["application.json", "application_print.html", "timeline.csv"]
+
+    application_payload = json.loads(archive.read("application.json").decode("utf-8"))
+    assert application_payload["application"]["id"] == str(application.pk)
+    assert application_payload["application"]["job_id"] == str(job.pk)
+    assert application_payload["application"]["status"] == Application.Status.APPLIED
+    assert application_payload["snapshot"]["scores"]["score"] == 84.2
+    assert application_payload["snapshot"]["sources"][0]["url"] == job.canonical_url
+
+    csv_rows = list(csv.DictReader(io.StringIO(archive.read("timeline.csv").decode("utf-8"))))
+    assert {row["type"] for row in csv_rows} >= {ApplicationTimelineEvent.EventType.CREATED, ApplicationTimelineEvent.EventType.SNAPSHOT_CAPTURED}
+
+    html = archive.read("application_print.html").decode("utf-8")
+    assert "<html" in html.lower()
+    assert "<link " not in html.lower()
+    assert "<script" not in html.lower()
+
+    assert "raw body" not in json.dumps(application_payload)
+
+    response_json = client.get(reverse("jobs:application-print", kwargs={"pk": application.pk}))
+    assert response_json.status_code == 200
+    assert response_json["Content-Type"] == "text/html; charset=utf-8"
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_application_export_and_delete_are_scoped_and_idempotent(client, job, user, source):
+    other_user = get_user_model().objects.create_user(
+        username="other",
+        email="other@example.invalid",
+        password="correct-horse-battery-staple",
+    )
+
+    record_feedback(user=user, job=job, action="applied")
+    owner_application = Application.objects.get(user=user, job=job)
+    other_application = Application.objects.create(
+        user=other_user,
+        job=job,
+        status=Application.Status.APPLIED,
+        applied_at=timezone.now() - timedelta(days=1),
+        follow_up_date=timezone.now().date(),
+    )
+
+    client.force_login(user)
+    assert client.get(
+        reverse("jobs:application-export", kwargs={"pk": other_application.pk})
+    ).status_code == 404
+    assert client.post(
+        reverse("jobs:application-delete", kwargs={"pk": other_application.pk})
+    ).status_code == 404
+
+    delete_response = client.post(
+        reverse("jobs:application-delete", kwargs={"pk": owner_application.pk}),
+        follow=True,
+    )
+    assert delete_response.status_code == 200
+    assert not Application.objects.filter(pk=owner_application.pk).exists()
+
+    repeat_delete = client.post(
+        reverse("jobs:application-delete", kwargs={"pk": owner_application.pk}),
+        follow=True,
+    )
+    assert repeat_delete.status_code == 200

tests/integration/test_dedupe_lifecycle_feedback.py

@@ -0,0 +1,147 @@
+from datetime import timedelta
+
+import pytest
+from django.utils import timezone
+
+from apps.jobs.models import Application, Feedback, JobPosting, JobSourceAlias
+from apps.jobs.services.dedupe import candidate_similarity, find_existing_job, text_similarity
+from apps.jobs.services.feedback import record_feedback
+from apps.jobs.services.lifecycle import update_lifecycle
+from apps.jobs.services.normalization import CanonicalJobDraft
+from apps.sources.models import RawDocument
+
+
+def _draft(job, **overrides):
+    values = {
+        "source_url": job.canonical_url,
+        "canonical_url": job.canonical_url,
+        "external_id": "external-1",
+        "title": job.original_title,
+        "normalized_title": job.normalized_title,
+        "job_family": job.job_family,
+        "employer_name": job.employer_name,
+        "employer_domain": job.employer.domain,
+        "location_text": job.raw_location,
+        "region": job.region,
+        "municipality": job.municipality,
+        "postal_code": "3500",
+        "country": "BE",
+        "workplace_type": job.workplace_type,
+        "employment_types": job.employment_types,
+        "language": "nl",
+        "description_html": "",
+        "description_text": job.description_text,
+        "date_posted": None,
+        "valid_through": None,
+        "compensation": {},
+        "skills_required": job.skills_required,
+        "skills_preferred": job.skills_preferred,
+        "content_hash": job.content_hash,
+        "canonical_key": job.canonical_key,
+    }
+    values.update(overrides)
+    return CanonicalJobDraft(**values)
+
+
+@pytest.mark.django_db
+def test_dedupe_exact_fuzzy_and_new(job, source):
+    document = RawDocument.objects.create(
+        source=source,
+        kind=RawDocument.Kind.HTML,
+        content_hash="e" * 64,
+    )
+    JobSourceAlias.objects.create(
+        job=job,
+        source=source,
+        raw_document=document,
+        url=job.canonical_url,
+        canonical_url=job.canonical_url,
+        external_id="external-1",
+    )
+    exact = find_existing_job(_draft(job))
+    assert exact.job == job
+    assert exact.reason == "exact_external_id"
+
+    fuzzy_draft = _draft(
+        job,
+        external_id="",
+        canonical_url="https://jobs.example.org/other-id",
+        canonical_key="f" * 64,
+    )
+    fuzzy = find_existing_job(fuzzy_draft, threshold=0.8)
+    assert fuzzy.job == job
+    assert fuzzy.reason == "fuzzy_strong"
+    assert candidate_similarity(job, fuzzy_draft) >= 0.8
+    assert text_similarity("", "x") == 0
+
+    new = find_existing_job(
+        _draft(
+            job,
+            external_id="",
+            canonical_url="https://new.example.org/job",
+            canonical_key="1" * 64,
+            title="Chef kok",
+            normalized_title="chef kok",
+            employer_name="Restaurant",
+            employer_domain="restaurant.example",
+            description_text="Koken in een restaurant",
+        )
+    )
+    assert new.job is None
+    assert new.reason == "new"
+
+
+@pytest.mark.django_db
+def test_applied_feedback_creates_application_and_follow_up(user, profile, job):
+    feedback = record_feedback(user=user, job=job, action=Feedback.Action.APPLIED, reason="goed")
+    assert feedback.profile == profile
+    application = Application.objects.get(user=user, job=job)
+    assert application.status == Application.Status.APPLIED
+    assert application.snapshot["title"] == job.original_title
+    assert application.follow_up_date == timezone.localdate() + timedelta(days=7)
+
+    application.status = Application.Status.PREPARING
+    application.save(update_fields=["status"])
+    record_feedback(user=user, job=job, action=Feedback.Action.APPLIED)
+    application.refresh_from_db()
+    assert application.status == Application.Status.APPLIED
+
+
+@pytest.mark.django_db
+def test_lifecycle_transitions(employer):
+    now = timezone.now()
+    expired = JobPosting.objects.create(
+        employer=employer,
+        original_title="Expired",
+        normalized_title="expired",
+        canonical_key="2" * 64,
+        content_hash="3" * 64,
+        valid_through=now - timedelta(minutes=1),
+        status=JobPosting.Status.ACTIVE,
+    )
+    uncertain = JobPosting.objects.create(
+        employer=employer,
+        original_title="Uncertain",
+        normalized_title="uncertain",
+        canonical_key="4" * 64,
+        content_hash="5" * 64,
+        last_seen=now - timedelta(days=5),
+        status=JobPosting.Status.ACTIVE,
+    )
+    removed = JobPosting.objects.create(
+        employer=employer,
+        original_title="Removed",
+        normalized_title="removed",
+        canonical_key="6" * 64,
+        content_hash="7" * 64,
+        last_seen=now - timedelta(days=20),
+        status=JobPosting.Status.UNCERTAIN,
+    )
+    result = update_lifecycle()
+    assert result == {"expired": 1, "uncertain": 1, "removed": 1}
+    expired.refresh_from_db()
+    uncertain.refresh_from_db()
+    removed.refresh_from_db()
+    assert expired.status == JobPosting.Status.EXPIRED
+    assert uncertain.status == JobPosting.Status.UNCERTAIN
+    assert removed.status == JobPosting.Status.REMOVED

tests/integration/test_discovery_command.py

@@ -0,0 +1,86 @@
+from io import StringIO
+
+import pytest
+from django.core.management import call_command
+
+from apps.sources.models import Source
+
+
+@pytest.mark.django_db
+def test_discover_sources_command_runs_for_all_modes_and_is_replayable():
+    out = StringIO()
+    call_command(
+        "discover_sources",
+        "html",
+        "--input",
+        "fixtures/discovery/html_candidate_discovery.html",
+        "--base-url",
+        "https://example.org/",
+        stdout=out,
+    )
+    call_command(
+        "discover_sources",
+        "sitemap",
+        "--input",
+        "fixtures/discovery/sitemap_index.xml",
+        "--base-url",
+        "https://jobs.example.org/",
+        stdout=out,
+    )
+    call_command(
+        "discover_sources",
+        "feed",
+        "--input",
+        "fixtures/discovery/feed_jobs.xml",
+        "--base-url",
+        "https://jobs.example.org/feed/jobs",
+        stdout=out,
+    )
+    call_command(
+        "discover_sources",
+        "email",
+        "--input",
+        "fixtures/emails/sample_alert.eml",
+        stdout=out,
+    )
+
+    source_types = {
+        (source.domain, source.source_type)
+        for source in Source.objects.filter(status=Source.Status.CANDIDATE)
+    }
+    assert ("example.org", Source.Type.EMPLOYER) in source_types
+    assert ("jobs.example.org", Source.Type.EMPLOYER) in source_types
+    assert ("example.org", Source.Type.RSS) in source_types
+    assert ("jobs.example.org", Source.Type.SITEMAP) in source_types
+    assert ("jobs.example.org", Source.Type.RSS) in source_types
+    assert "Persist result" in out.getvalue()
+
+    out.seek(0)
+    out.truncate(0)
+    call_command(
+        "discover_sources",
+        "html",
+        "--input",
+        "fixtures/discovery/html_candidate_discovery.html",
+        "--base-url",
+        "https://example.org/",
+        stdout=out,
+    )
+    assert Source.objects.filter(status=Source.Status.CANDIDATE).count() == 5
+
+
+@pytest.mark.django_db
+def test_discover_sources_dry_run_keeps_db_empty():
+    out = StringIO()
+    call_command(
+        "discover_sources",
+        "feed",
+        "--input",
+        "fixtures/discovery/feed_jobs.xml",
+        "--base-url",
+        "https://jobs.example.org/feed/jobs",
+        "--dry-run",
+        stdout=out,
+    )
+    assert Source.objects.count() == 0
+    assert "Dry-run modus" in out.getvalue()

tests/integration/test_email_import_notifications.py

@@ -0,0 +1,72 @@
+from datetime import datetime
+from pathlib import Path
+from zoneinfo import ZoneInfo
+
+import pytest
+from django.core import mail
+from django.test import override_settings
+
+from apps.jobs.models import Feedback, ScoreRun
+from apps.jobs.services.scoring import score_and_save
+from apps.notifications.models import DigestOutbox
+from apps.notifications.services import build_digest_payload, create_daily_outbox, send_digest
+from apps.sources.models import EmailMessageRecord, RawDocument
+from apps.sources.services.email_import import ingest_email, message_identity
+
+
+@pytest.mark.django_db
+def test_email_import_is_idempotent(profile):
+    raw = Path("fixtures/emails/sample_alert.eml").read_bytes()
+    identity = message_identity(raw)
+    first = ingest_email(raw)
+    second = ingest_email(raw)
+    assert first.pk == second.pk
+    assert first.message_id == identity
+    assert EmailMessageRecord.objects.count() == 1
+    assert RawDocument.objects.filter(kind=RawDocument.Kind.EMAIL).count() == 1
+    assert len(first.links) == 2
+
+
+@pytest.mark.django_db
+def test_digest_payload_outbox_deduplication_and_send(profile, job):
+    score = score_and_save(job, profile)
+    score.recommendation = ScoreRun.Recommendation.STRONG
+    score.score = 92
+    score.save(update_fields=["recommendation", "score"])
+
+    local_tz = ZoneInfo(profile.timezone)
+    due = datetime(2026, 7, 20, 7, 35, tzinfo=local_tz)
+    payload = build_digest_payload(profile, local_date=due.date())
+    assert payload["strong_count"] == 1
+    assert payload["scores"][0]["title"] == job.original_title
+
+    with override_settings(DIGEST_RECIPIENT="digest@example.invalid"):
+        first = create_daily_outbox(profile, now=due)
+        second = create_daily_outbox(profile, now=due)
+    assert first is not None
+    assert second.pk == first.pk
+    assert first.status == DigestOutbox.Status.PENDING
+
+    send_digest(first)
+    first.refresh_from_db()
+    assert first.status == DigestOutbox.Status.SENT
+    assert first.sent_at is not None
+    assert len(mail.outbox) == 1
+    assert "Infrastructure Engineer" in mail.outbox[0].body
+
+
+@pytest.mark.django_db
+def test_digest_skips_outside_window_and_hidden_jobs(profile, job):
+    score_and_save(job, profile)
+    Feedback.objects.create(
+        user=profile.user,
+        profile=profile,
+        job=job,
+        action=Feedback.Action.HIDE,
+    )
+    payload = build_digest_payload(profile)
+    assert payload["scores"] == []
+
+    local_tz = ZoneInfo(profile.timezone)
+    too_early = datetime(2026, 7, 20, 6, 0, tzinfo=local_tz)
+    assert create_daily_outbox(profile, now=too_early) is None

tests/integration/test_import_geodata.py

@@ -0,0 +1,107 @@
+from __future__ import annotations
+
+from io import StringIO
+
+from django.core.management import call_command
+from django.core.management.base import CommandError
+import pytest
+
+from apps.jobs.models import ScoreRun
+from apps.jobs.services.scoring import score_and_save
+
+@pytest.mark.django_db
+def test_import_geodata_validate_only_and_import_command(tmp_path):
+    csv_path = tmp_path / "geodata.csv"
+    csv_path.write_text(
+        "postal_code,municipality,region,latitude,longitude\n"
+        "9000,Gent,Vlaams-Brabant,51.05,3.73\n"
+        "1040,Brussel,Brussels Hoofdstedelijk Gewest,50.85,4.35\n"
+    )
+    out = StringIO()
+    call_command(
+        "import_geodata",
+        str(csv_path),
+        "--source-name",
+        "local-test",
+        "--dataset-version",
+        "2026-01-01",
+        "--validate-only",
+        "--license-name",
+        "Test Dataset",
+        "--license-url",
+        "https://example.org/license",
+        stdout=out,
+    )
+    assert "Validatie geslaagd (2 rijen)." in out.getvalue()
+
+    out = StringIO()
+    call_command(
+        "import_geodata",
+        str(csv_path),
+        "--source-name",
+        "local-test",
+        "--dataset-version",
+        "2026-01-01",
+        "--license-name",
+        "Test Dataset",
+        "--license-url",
+        "https://example.org/license",
+        stdout=out,
+    )
+    assert "Geocodebron geimporteerd (4 lookuprijen)" in out.getvalue()
+
+
+@pytest.mark.django_db
+def test_import_geodata_rejects_half_license_metadata(tmp_path):
+    csv_path = tmp_path / "geodata.csv"
+    csv_path.write_text(
+        "postal_code,municipality,region,latitude,longitude\n9000,Gent,Vlaams-Brabant,51.05,3.73\n"
+    )
+    out = StringIO()
+        with pytest.raises(CommandError):
+            call_command(
+                "import_geodata",
+                str(csv_path),
+                "--source-name",
+            "local-test",
+            "--dataset-version",
+            "2026-01-01",
+                "--license-name",
+                "Test Dataset",
+                stdout=out,
+            )
+
+
+@pytest.mark.django_db
+def test_import_geodata_triggers_targeted_rescore(tmp_path, profile, job):
+    job.postal_code = "9000"
+    job.municipality = "Gent"
+    job.save(update_fields=["postal_code", "municipality"])
+
+    score_and_save(job, profile)
+    initial_scores = ScoreRun.objects.filter(profile=profile, job=job).count()
+
+    csv_path = tmp_path / "geodata.csv"
+    csv_path.write_text(
+        "postal_code,municipality,region,latitude,longitude\n9000,Gent,Vlaams-Brabant,51.05,3.73\n"
+    )
+    out = StringIO()
+    call_command(
+        "import_geodata",
+        str(csv_path),
+        "--source-name",
+        "local-test",
+        "--dataset-version",
+        "2026-01-01",
+        "--license-name",
+        "Test Dataset",
+        "--license-url",
+        "https://example.org/license",
+        stdout=out,
+    )
+    assert "scoreregels herberekend voor aangepaste geodata" in out.getvalue()
+    assert ScoreRun.objects.filter(profile=profile, job=job).count() == initial_scores + 1
+
+    latest = ScoreRun.objects.filter(profile=profile, job=job).order_by("-created_at").first()
+    assert latest is not None
+    assert latest.evidence["distance_km"] is not None

tests/integration/test_pipeline.py

@@ -0,0 +1,120 @@
+import hashlib
+from datetime import timedelta
+from pathlib import Path
+
+import pytest
+from django.utils import timezone
+
+from apps.jobs.models import Employer, JobPosting, JobSourceAlias, ScoreRun
+from apps.jobs.services.pipeline import process_raw_document
+from apps.sources.models import RawDocument, Source
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_pipeline_is_idempotent_and_scores(source, profile):
+    content = Path("fixtures/pages/sample_jsonld_job.html").read_text(encoding="utf-8")
+    document = RawDocument.objects.create(
+        source=source,
+        url="https://jobs.example.org/vacatures/infrastructure-engineer",
+        final_url="https://jobs.example.org/vacatures/infrastructure-engineer",
+        kind=RawDocument.Kind.HTML,
+        content_type="text/html",
+        content_hash=hashlib.sha256(content.encode()).hexdigest(),
+        body_text=content,
+        byte_length=len(content.encode()),
+        retain_until=timezone.now() + timedelta(days=7),
+    )
+    first = process_raw_document(document)
+    second = process_raw_document(document)
+    assert first["created"] == 1
+    assert second["created"] == 0
+    assert JobPosting.objects.count() == 1
+    assert JobSourceAlias.objects.count() == 1
+    assert ScoreRun.objects.filter(profile=profile).count() == 2
+
+
+def test_pipeline_resolves_recruiter_alias_to_direct_employer():
+    source = Source.objects.create(
+        name="SmartRecruiters",
+        source_type=Source.Type.ATS,
+        base_url="https://jobs.smartrecruiters.com/example",
+        domain="jobs.smartrecruiters.com",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    employer = Employer.objects.create(
+        name="Example Public IT",
+        normalized_name="example public it",
+        domain="jobs.example.org",
+        is_direct_employer=True,
+        is_recruiter=False,
+        confidence=0.9,
+    )
+    direct_job = JobPosting.objects.create(
+        employer=employer,
+        original_title="Security Engineer",
+        normalized_title="security engineer",
+        canonical_url="https://jobs.example.org/vacatures/security-engineer",
+        canonical_key="8" * 64,
+        content_hash="e" * 64,
+        description_text="Beveiliger, hybride werk met support op locatie.",
+        raw_location="Antwerpen",
+        region="Antwerpen",
+        municipality="Antwerpen",
+        workplace_type=JobPosting.Workplace.HYBRID,
+        employment_types=["full_time"],
+        direct_employer=True,
+    )
+    content = """
+    <html lang="nl">
+      <head>
+        <meta charset="utf-8">
+        <title>Security Engineer | Example Public IT</title>
+        <link rel="canonical" href="https://jobs.smartrecruiters.com/example/security-engineer">
+        <script type="application/ld+json">
+        {
+          "@context": "https://schema.org",
+          "@type": "JobPosting",
+          "identifier": {"@type": "PropertyValue", "name": "Example Public IT", "value": "VR-DEMO-SEC-001"},
+          "title": "Security Engineer",
+          "description": "<p>Beveiliging, monitoring en hybride support.</p>",
+          "datePosted": "2026-07-20",
+          "validThrough": "2026-08-31T23:59:00+02:00",
+          "employmentType": ["FULL_TIME"],
+          "hiringOrganization": {"@type": "Organization", "name": "Example Public IT", "sameAs": "https://www.example.org"},
+          "jobLocation": {"@type": "Place", "address": {"@type": "PostalAddress", "streetAddress": "Voorbeeldstraat 1", "addressLocality": "Antwerpen", "addressRegion": "Antwerpen", "postalCode": "2000", "addressCountry": "BE"}},
+          "jobLocationType": "HYBRID",
+          "url": "/example/security-engineer"
+        }
+        </script>
+      </head>
+      <body><main><h1>Security Engineer</h1></main></body>
+    </html>
+    """.strip()
+    body_hash = hashlib.sha256(content.encode()).hexdigest()
+    document = RawDocument.objects.create(
+        source=source,
+        url="https://jobs.smartrecruiters.com/example/security-engineer",
+        final_url="https://jobs.smartrecruiters.com/example/security-engineer",
+        kind=RawDocument.Kind.HTML,
+        content_type="text/html",
+        content_hash=body_hash,
+        body_text=content,
+        byte_length=len(content.encode()),
+        retain_until=timezone.now() + timedelta(days=7),
+    )
+    result = process_raw_document(document)
+    direct_job.refresh_from_db()
+    assert direct_job.canonical_url == "https://jobs.example.org/vacatures/security-engineer"
+    assert result["created"] == 0
+    assert result["duplicates"] == 1
+    alias = JobSourceAlias.objects.get(job=direct_job, source=source)
+    payload = alias.payload.get("employer_resolution")
+    assert payload is not None
+    assert payload["reason"] == "resolved_direct_match"
+    assert payload["resolved_direct"] is True
+    assert (
+        payload["canonical_url"]
+        == "https://jobs.example.org/vacatures/security-engineer"
+    )

tests/integration/test_reminders.py

@@ -0,0 +1,132 @@
+from datetime import datetime, time
+from zoneinfo import ZoneInfo
+from decimal import Decimal
+
+import pytest
+from django.utils import timezone
+
+from apps.jobs.models import Application, Feedback, ScoreRun
+from apps.jobs.services.scoring import score_and_save
+from apps.notifications.models import ReminderOutbox
+from apps.notifications.services import (
+    create_closing_reminders,
+    create_follow_up_reminders,
+    create_top_match_reminders,
+)
+
+
+@pytest.mark.django_db
+def test_closing_reminder_is_idempotent_and_skips_hidden(profile, job):
+    zone = ZoneInfo(profile.timezone)
+    now = timezone.now().astimezone(zone)
+    job.valid_through = datetime(
+        year=now.year,
+        month=now.month,
+        day=now.day,
+        hour=20,
+        tzinfo=zone,
+    )
+    job.save(update_fields=["valid_through"])
+
+    Feedback.objects.create(
+        user=profile.user,
+        profile=profile,
+        job=job,
+        action=Feedback.Action.HIDE,
+    )
+    assert create_closing_reminders(profile, now=now) == []
+
+    Feedback.objects.filter(user=profile.user, profile=profile, job=job).delete()
+    first = create_closing_reminders(profile, now=now)[0]
+    second = create_closing_reminders(profile, now=now)[0]
+    assert first.pk == second.pk
+    assert first.status == ReminderOutbox.Status.PENDING
+
+
+@pytest.mark.django_db
+def test_follow_up_reminder_respects_configured_weekdays(profile, job):
+    zone = ZoneInfo(profile.timezone)
+    now = timezone.now().astimezone(zone)
+    follow_date = now.date()
+    Application.objects.create(
+        user=profile.user,
+        job=job,
+        follow_up_date=follow_date,
+        status=Application.Status.APPLIED,
+        notes="",
+    )
+    profile.follow_up_weekdays = [(follow_date.weekday() + 1) % 7]
+    profile.save(update_fields=["follow_up_weekdays"])
+    assert create_follow_up_reminders(profile, now=now) == []
+
+    profile.follow_up_weekdays = [follow_date.weekday()]
+    profile.save(update_fields=["follow_up_weekdays"])
+    reminders = create_follow_up_reminders(profile, now=now)
+    assert len(reminders) == 1
+    assert reminders[0].reminder_type == ReminderOutbox.ReminderType.FOLLOW_UP
+
+
+@pytest.mark.django_db
+def test_top_match_reminder_requires_threshold_and_profile_setting(profile, job):
+    profile.top_match_reminders_enabled = True
+    profile.top_match_threshold = 90
+    profile.save(update_fields=["top_match_reminders_enabled", "top_match_threshold"])
+    score = score_and_save(job, profile)
+    score.recommendation = ScoreRun.Recommendation.STRONG
+    score.score = 85
+    score.confidence = Decimal("0.9")
+    score.save(update_fields=["recommendation", "score", "confidence"])
+
+    assert create_top_match_reminders(profile) == []
+
+    score.score = 95
+    score.confidence = Decimal("0.90")
+    score.save(update_fields=["score", "confidence"])
+    score.confidence = Decimal("0.84")
+    score.save(update_fields=["confidence"])
+    assert create_top_match_reminders(profile) == []
+
+    score.confidence = Decimal("0.86")
+    score.save(update_fields=["confidence"])
+    reminders = create_top_match_reminders(profile)
+    assert len(reminders) == 1
+    assert reminders[0].reminder_type == ReminderOutbox.ReminderType.TOP_MATCH
+
+    profile.top_match_reminders_enabled = False
+    profile.save(update_fields=["top_match_reminders_enabled"])
+    assert create_top_match_reminders(profile) == []
+
+
+@pytest.mark.django_db
+def test_reminder_obeys_quiet_hours_and_empty_recipient(profile, job):
+    zone = ZoneInfo(profile.timezone)
+    now = timezone.now().astimezone(zone).replace(hour=6, minute=30, second=0, microsecond=0)
+    job.valid_through = datetime(
+        year=now.year,
+        month=now.month,
+        day=now.day,
+        hour=18,
+        tzinfo=zone,
+    )
+    job.save(update_fields=["valid_through"])
+
+    profile.user.email = ""
+    profile.user.save(update_fields=["email"])
+    assert create_closing_reminders(profile, now=now)[0].status == ReminderOutbox.Status.SKIPPED
+
+    profile.user.email = "reminder@example.invalid"
+    profile.user.save(update_fields=["email"])
+    reminder = create_closing_reminders(profile, now=now)[0]
+    assert reminder.status == ReminderOutbox.Status.PENDING
+    assert reminder.scheduled_for.timetz().replace(tzinfo=None) == time(7, 0)
+
+
+@pytest.mark.django_db
+def test_reminder_service_handles_dst_dates(profile, job):
+    zone = ZoneInfo(profile.timezone)
+    now = datetime(2026, 10, 25, 1, 30, tzinfo=zone)
+    job.valid_through = now.replace(hour=20)
+    job.save(update_fields=["valid_through"])
+    reminder = create_closing_reminders(profile, now=now)[0]
+    assert reminder.scheduled_for.tzinfo is not None
+    assert reminder.scheduled_for.tzinfo.key == "Europe/Brussels"

tests/integration/test_source_health.py

@@ -0,0 +1,170 @@
+from datetime import timedelta
+
+from django.utils import timezone
+
+import pytest
+
+from apps.sources.models import RawDocument, Source, SourceRun
+from apps.sources.services.health import collect_source_health
+from apps.sources.tasks import cleanup_raw_documents, enforce_source_health, recover_source_health
+
+
+def _create_run(
+    source: Source,
+    *,
+    status: str,
+    started_at=None,
+    error_category: str = "",
+    extracted_count: int = 0,
+    created_count: int = 0,
+    updated_count: int = 0,
+    duplicate_count: int = 0,
+    http_status: int = 200,
+    metrics: dict[str, object] | None = None,
+):
+    started_at = started_at or timezone.now()
+    finished_at = started_at + timedelta(seconds=1)
+    return SourceRun.objects.create(
+        source=source,
+        started_at=started_at,
+        finished_at=finished_at,
+        status=status,
+        error_category=error_category,
+        http_status=http_status,
+        extracted_count=extracted_count,
+        created_count=created_count,
+        updated_count=updated_count,
+        duplicate_count=duplicate_count,
+        metrics=metrics or {},
+    )
+
+
+@pytest.mark.django_db
+def test_enforce_source_health_quarantines_on_temporary_failures(source):
+    for index in range(6):
+        _create_run(
+            source,
+            status=SourceRun.Status.FAILED,
+            started_at=timezone.now() - timedelta(minutes=index),
+            error_category="timeout",
+        )
+
+    result = enforce_source_health()
+    assert result["quarantined"] == 1
+    source.refresh_from_db()
+    assert source.status == Source.Status.QUARANTINED
+    assert source.policy == Source.Policy.DENY
+
+
+@pytest.mark.django_db
+def test_enforce_source_health_quarantines_on_policy_category(source):
+    for index in range(3):
+        _create_run(
+            source,
+            status=SourceRun.Status.FAILED,
+            started_at=timezone.now() - timedelta(minutes=index),
+            error_category="policy",
+        )
+    result = enforce_source_health()
+    assert result["quarantined"] == 1
+    source.refresh_from_db()
+    assert source.status == Source.Status.QUARANTINED
+    assert "beleid" in source.policy_reason
+
+
+@pytest.mark.django_db
+def test_enforce_source_health_quarantines_on_parser_drift(source):
+    for index in range(3):
+        _create_run(
+            source,
+            status=SourceRun.Status.SUCCESS,
+            started_at=timezone.now() - timedelta(minutes=index),
+            extracted_count=0,
+            created_count=0,
+            updated_count=0,
+            duplicate_count=0,
+            http_status=200,
+            metrics={"parser": "generic-html", "warnings": ["veld ontbreekt", "datum ontbreekt"]},
+        )
+    assert enforce_source_health()["quarantined"] == 1
+    source.refresh_from_db()
+    assert source.status == Source.Status.QUARANTINED
+    assert source.policy == Source.Policy.DENY
+
+
+@pytest.mark.django_db
+def test_collect_source_health_exposes_parser_and_counts(source):
+    _create_run(
+        source,
+        status=SourceRun.Status.SUCCESS,
+        error_category="",
+        extracted_count=2,
+        created_count=1,
+        updated_count=1,
+        duplicate_count=0,
+        metrics={"parser": "jsonld", "warnings": ["warn"]},
+    )
+    row = collect_source_health()[0]
+    assert row.source_id == source.pk
+    assert row.last_parser == "jsonld"
+    assert row.last_parser_warnings == 1
+    assert row.extracted_count == 2
+    assert row.updated_count == 1
+
+
+@pytest.mark.django_db
+def test_recover_source_health_starts_canary_once(source, monkeypatch):
+    source.status = Source.Status.QUARANTINED
+    source.policy = Source.Policy.DENY
+    source.metadata = {
+        "source_health": {
+            "state": "quarantined",
+            "quarantine_reason": "tijdelijke fout",
+            "quarantined_at": timezone.now().isoformat(),
+            "recovery_due_at": timezone.now().isoformat(),
+            "canary_started": False,
+        }
+    }
+    source.save(update_fields=["status", "policy", "metadata"])
+
+    calls: list[int] = []
+
+    def fake_delay(source_id: int, force: bool = False):
+        calls.append(source_id)
+        return source_id
+
+    monkeypatch.setattr("apps.sources.tasks.fetch_source.delay", fake_delay)
+
+    result = recover_source_health()
+    assert result["canary_started"] == 1
+    assert calls == [source.pk]
+
+    source.refresh_from_db()
+    assert source.status == Source.Status.TRIAL
+    assert source.policy == Source.Policy.REVIEW
+
+    # Canary is eenmalig, niet continu herstartbaar.
+    source.refresh_from_db()
+    assert recover_source_health()["canary_started"] == 0
+
+
+@pytest.mark.django_db
+def test_cleanup_retains_quarantined_raw_documents(source):
+    old = RawDocument.objects.create(
+        source=source,
+        kind=RawDocument.Kind.TEXT,
+        content_hash="a" * 64,
+        retain_until=timezone.now() - timedelta(days=1),
+    )
+    keep = RawDocument.objects.create(
+        source=source,
+        kind=RawDocument.Kind.TEXT,
+        content_hash="b" * 64,
+        retain_until=timezone.now() - timedelta(days=1),
+        quarantined=True,
+    )
+
+    result = cleanup_raw_documents()
+    assert result["deleted"] == 1
+    assert RawDocument.objects.filter(pk=old.pk).exists() is False
+    assert RawDocument.objects.filter(pk=keep.pk).exists() is True

tests/integration/test_sources_manual_import.py

@@ -0,0 +1,111 @@
+from __future__ import annotations
+
+from django.core.cache import cache
+from django.test import override_settings
+from django.urls import reverse
+
+import pytest
+
+from apps.sources.services.fetcher import FetchedDocument
+from apps.sources.services.manual_import import ManualImportError
+
+
+def test_manual_import_view_prefills_source_url(client, user):
+    client.force_login(user)
+    response = client.get(
+        reverse("sources:manual_import"), {"source_url": "https://jobs.example.org/vacature"}
+    )
+    assert response.status_code == 200
+    assert 'value="https://jobs.example.org/vacature"' in response.content.decode("utf-8")
+
+
+@pytest.mark.django_db
+def test_manual_import_view_posts_url_and_shows_result(client, user, monkeypatch):
+    def fake_fetch(*_args, **_kwargs):
+        return FetchedDocument(
+            requested_url="https://jobs.example.org/engineering",
+            final_url="https://jobs.example.org/engineering",
+            status_code=200,
+            headers={"content-type": "text/html; charset=utf-8"},
+            content=b"<html><body>vacature</body></html>",
+        )
+
+    def fake_process(_document):
+        return {
+            "extracted": 1,
+            "created": 1,
+            "updated": 0,
+            "duplicates": 0,
+            "parser": "manual-test",
+            "warnings": [],
+        }
+
+    monkeypatch.setattr("apps.sources.services.manual_import.fetch_url", fake_fetch)
+    monkeypatch.setattr("apps.sources.services.manual_import.process_raw_document", fake_process)
+
+    client.force_login(user)
+    response = client.post(
+        reverse("sources:manual_import"),
+        {"source_url": "https://jobs.example.org/engineering"},
+    )
+    html = response.content.decode("utf-8")
+    assert response.status_code == 200
+    assert "Importresultaat" in html
+    assert "Herkenbare items: 1" in html
+
+
+@pytest.mark.django_db
+def test_manual_import_view_paste_prefers_post_flow_and_does_not_fetch(client, user, monkeypatch):
+    def fail_fetch(*_args, **_kwargs):
+        raise AssertionError("Fetcher should not run for paste imports.")
+
+    def fake_process(_document):
+        return {
+            "extracted": 1,
+            "created": 0,
+            "updated": 0,
+            "duplicates": 1,
+            "parser": "manual-test",
+            "warnings": [],
+        }
+
+    monkeypatch.setattr("apps.sources.services.manual_import.fetch_url", fail_fetch)
+    monkeypatch.setattr("apps.sources.services.manual_import.process_raw_document", fake_process)
+
+    client.force_login(user)
+    response = client.post(
+        reverse("sources:manual_import"),
+        {"pasted_text": "Eerste regel\nTweede regel"},
+    )
+    html = response.content.decode("utf-8")
+    assert response.status_code == 200
+    assert "Importresultaat" in html
+    assert "Modus: paste" in html
+
+
+@pytest.mark.django_db
+@override_settings(
+    MANUAL_IMPORT_RATE_LIMIT_MAX_ATTEMPTS=1,
+    MANUAL_IMPORT_RATE_LIMIT_WINDOW_SECONDS=120,
+    MANUAL_IMPORT_RATE_LIMIT_BLOCK_SECONDS=60,
+)
+def test_manual_import_blocks_after_failed_attempts(client, user, monkeypatch):
+    cache.clear()
+
+    def fail_import(*_args, **_kwargs):
+        raise ManualImportError("Handmatige import faalde in test")
+
+    client.force_login(user)
+    monkeypatch.setattr("apps.sources.views.import_manual_source", fail_import)
+
+    response = client.post(
+        reverse("sources:manual_import"),
+        {"source_url": "https://jobs.example.org/engineering"},
+    )
+    assert "Handmatige import faalde in test" in response.content.decode("utf-8")
+
+    response = client.post(
+        reverse("sources:manual_import"),
+        {"source_url": "https://jobs.example.org/engineering"},
+    )
+    assert "Te veel handmatige importverzoeken" in response.content.decode("utf-8")

tests/integration/test_tasks_and_more_views.py

@@ -0,0 +1,217 @@
+from datetime import timedelta
+
+import pytest
+from django.urls import reverse
+from django.utils import timezone
+
+from apps.jobs.models import Feedback, ScoreRun
+from apps.jobs.tasks import rescore_active_jobs, update_job_lifecycle
+from apps.notifications.tasks import build_due_digests
+from apps.sources.models import RawDocument, Source, SourcePolicyReview, SourceLease, SourceRun
+from apps.sources.services.fetcher import RateLimitedError, FetchedDocument
+from apps.sources.tasks import cleanup_raw_documents, fetch_source, schedule_due_sources
+
+
+@pytest.mark.django_db
+def test_task_helpers(profile, job, source, monkeypatch):
+    result = rescore_active_jobs(profile.pk)
+    assert result["scores_created"] == 1
+    assert ScoreRun.objects.filter(job=job, profile=profile).exists()
+    assert update_job_lifecycle() == {"expired": 0, "uncertain": 0, "removed": 0}
+
+    calls = []
+    monkeypatch.setattr(
+        "apps.sources.tasks.fetch_source.delay", lambda source_id: calls.append(source_id)
+    )
+    source.next_run_at = timezone.now() - timedelta(minutes=1)
+    source.save(update_fields=["next_run_at"])
+    assert schedule_due_sources()["scheduled"] == 1
+    assert calls == [source.pk]
+
+    old = RawDocument.objects.create(
+        source=source,
+        kind=RawDocument.Kind.TEXT,
+        content_hash="8" * 64,
+        retain_until=timezone.now() - timedelta(days=1),
+    )
+    assert cleanup_raw_documents()["deleted"] >= 1
+    assert not RawDocument.objects.filter(pk=old.pk).exists()
+
+    source.policy = Source.Policy.DENY
+    source.policy_reason = "test"
+    source.save(update_fields=["policy", "policy_reason"])
+    assert fetch_source(source.pk)["status"] == "skipped"
+    assert SourceRun.objects.filter(source=source, status=SourceRun.Status.SKIPPED).exists()
+
+
+def test_fetch_source_skips_when_lease_blocked(db, source, monkeypatch):
+    SourceLease.objects.create(
+        source=source,
+        token="active-worker",
+        worker_id="active-worker",
+        expires_at=timezone.now() + timedelta(minutes=2),
+    )
+    calls = {"count": 0}
+
+    def fake_fetch_url(*args, **kwargs):
+        calls["count"] += 1
+        return FetchedDocument(
+            requested_url=source.base_url,
+            final_url=source.base_url,
+            status_code=200,
+            headers={},
+            content=b"Vacature",
+        )
+
+    monkeypatch.setattr("apps.sources.tasks.fetch_url", fake_fetch_url)
+    assert fetch_source(source.pk)["status"] == "skipped"
+    assert calls["count"] == 0
+    assert SourceRun.objects.filter(source=source, status=SourceRun.Status.SKIPPED).exists()
+
+
+def test_fetch_source_handles_rate_limit_backoff(db, source, monkeypatch):
+    called = {"count": 0}
+
+    def fake_fetch_url(*args, **kwargs):
+        called["count"] += 1
+        raise RateLimitedError(30)
+
+    monkeypatch.setattr("apps.sources.tasks.fetch_url", fake_fetch_url)
+    result = fetch_source(source.pk)
+    assert result["status"] == "rate_limited"
+    assert called["count"] == 1
+    source.refresh_from_db()
+    assert source.next_run_at is not None
+    assert SourceRun.objects.filter(source=source, status=SourceRun.Status.FAILED).exists()
+
+
+def test_fetch_source_resets_fail_state_on_304(db, source, monkeypatch):
+    source.failure_count = 4
+    source.save(update_fields=["failure_count"])
+
+    def fake_fetch_url(*args, **kwargs):
+        return FetchedDocument(
+            requested_url=source.base_url,
+            final_url=source.base_url,
+            status_code=304,
+            headers={},
+            content=b"",
+        )
+
+    monkeypatch.setattr("apps.sources.tasks.fetch_url", fake_fetch_url)
+    assert fetch_source(source.pk)["status"] == "not_modified"
+    source.refresh_from_db()
+    assert source.failure_count == 0
+
+
+def test_fetch_source_is_blocked_by_conflicting_review(db):
+    source = Source.objects.create(
+        name="Blocked source",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    SourcePolicyReview.objects.create(
+        source=source,
+        decision=SourcePolicyReview.Decision.DENY,
+        reason="Review blokkeert deze bron",
+    )
+    result = fetch_source(source.pk)
+    assert result["status"] == "skipped"
+    assert SourceRun.objects.filter(source=source, status=SourceRun.Status.SKIPPED).exists()
+
+
+def test_fetch_source_is_blocked_by_expired_review(db):
+    source = Source.objects.create(
+        name="Expired review source",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    SourcePolicyReview.objects.create(
+        source=source,
+        decision=SourcePolicyReview.Decision.ALLOW,
+        reason="Review verlopen",
+        expires_at=timezone.now() - timedelta(days=1),
+    )
+    result = fetch_source(source.pk)
+    assert result["status"] == "skipped"
+    assert SourceRun.objects.filter(source=source, status=SourceRun.Status.SKIPPED).exists()
+
+    # Outside the digest window this is a safe no-op.
+    assert build_due_digests()["sent"] == 0
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_job_profile_source_and_application_views(client, user, profile, job, source):
+    client.force_login(user)
+    assert client.get(reverse("jobs:list"), {"q": "Infrastructure"}).status_code == 200
+    assert client.get(reverse("jobs:detail", kwargs={"pk": job.pk})).status_code == 200
+    response = client.post(
+        reverse("jobs:feedback", kwargs={"pk": job.pk}),
+        {"action": Feedback.Action.INTERESTING},
+    )
+    assert response.status_code == 302
+    assert Feedback.objects.filter(user=user, job=job).exists()
+
+    assert client.get(reverse("profiles:list")).status_code == 200
+    assert client.get(reverse("profiles:edit", kwargs={"pk": profile.pk})).status_code == 200
+    assert client.get(reverse("sources:list")).status_code == 200
+    assert client.get(reverse("dashboard:system")).status_code == 200
+    assert client.get(reverse("jobs:applications")).status_code == 200
+
+    source.policy = Source.Policy.DENY
+    source.save(update_fields=["policy"])
+    response = client.post(reverse("sources:retry", kwargs={"pk": source.pk}))
+    assert response.status_code == 302
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_sources_candidate_bulk_actions_and_evidence_view(client, user):
+    candidate = Source.objects.create(
+        name="Discoverd candidate",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.CANDIDATE,
+        policy=Source.Policy.REVIEW,
+        metadata={"discovery": [{"url": "https://jobs.example.org/careers", "discovered_from": "html", "reason": "career-link", "confidence": 0.9, "label": "Werken bij"}]},
+    )
+
+    other = Source.objects.create(
+        name="Another source",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://other.example.org/",
+        domain="other.example.org",
+        status=Source.Status.CANDIDATE,
+        policy=Source.Policy.REVIEW,
+    )
+
+    client.force_login(user)
+    response = client.get(reverse("sources:list"))
+    assert response.status_code == 200
+    content = response.content.decode("utf-8")
+    assert "jobs.example.org/careers" in content
+
+    response = client.post(
+        reverse("sources:bulk"),
+        {"action": "approve_to_trial", "source_ids": [str(candidate.pk)]},
+    )
+    assert response.status_code == 302
+    candidate.refresh_from_db()
+    assert candidate.status == Source.Status.TRIAL
+
+    response = client.post(
+        reverse("sources:bulk"),
+        {"action": "dismiss", "source_ids": [str(other.pk)]},
+    )
+    assert response.status_code == 302
+    other.refresh_from_db()
+    assert other.status == Source.Status.DISABLED
+    assert other.policy == Source.Policy.DENY

tests/integration/test_views.py

@@ -0,0 +1,51 @@
+import pytest
+from django.core.cache import cache
+from django.test import override_settings
+from django.urls import reverse
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_today_requires_login(client):
+    response = client.get(reverse("dashboard:today"))
+    assert response.status_code == 302
+    assert reverse("login") in response.url
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+def test_today_loads_for_user(client, user, profile):
+    client.force_login(user)
+    response = client.get(reverse("dashboard:today"))
+    assert response.status_code == 200
+    assert "Vandaag" in response.content.decode()
+
+
+@pytest.mark.integration
+@pytest.mark.django_db
+@override_settings(
+    AUTH_LOGIN_RATE_LIMIT_MAX_ATTEMPTS=2,
+    AUTH_LOGIN_RATE_LIMIT_WINDOW_SECONDS=120,
+    AUTH_LOGIN_RATE_LIMIT_BLOCK_SECONDS=300,
+)
+def test_login_view_rate_limits_failed_attempts(client, user):
+    cache.clear()
+
+    response = client.post(
+        reverse("login"),
+        {"username": user.username, "password": "wrong"},
+    )
+    assert response.status_code == 200
+
+    response = client.post(
+        reverse("login"),
+        {"username": user.username, "password": "wrong"},
+    )
+    body = response.content.decode("utf-8")
+    assert "Te veel inlogpogingen" in body
+
+    response = client.post(
+        reverse("login"),
+        {"username": user.username, "password": "wrong"},
+    )
+    assert "Te veel inlogpogingen" in response.content.decode("utf-8")

tests/security/test_policy.py

@@ -0,0 +1,94 @@
+from datetime import timedelta
+
+import pytest
+from django.utils import timezone
+
+from apps.sources.models import Source, SourcePolicyReview
+from apps.sources.services.policy import assess_url, create_policy_review, is_denied_domain
+
+
+def test_platform_domains_are_denied_including_subdomains():
+    assert is_denied_domain("www.linkedin.com")
+    assert is_denied_domain("be.indeed.com")
+    assert not is_denied_domain("jobs.example.org")
+    assert not assess_url("https://www.linkedin.com/jobs/view/123").allowed
+
+
+@pytest.mark.security
+def test_source_policy_without_review_is_blocked(db):
+    source = Source.objects.create(
+        name="Example jobs",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    decision = assess_url(source.base_url, source=source)
+    assert not decision.allowed
+    assert decision.status == Source.Policy.REVIEW
+
+
+@pytest.mark.security
+def test_source_review_expiry_blocks_fetch_decision(db):
+    source = Source.objects.create(
+        name="Example jobs",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    create_policy_review(
+        source,
+        actor=None,
+        decision=SourcePolicyReview.Decision.ALLOW,
+        reason="Aanvankelijk goedgekeurd",
+        scope=SourcePolicyReview.Scope.SOURCE,
+        expires_at=timezone.now() - timedelta(days=1),
+    )
+    decision = assess_url(source.base_url, source=source)
+    assert not decision.allowed
+
+
+@pytest.mark.security
+def test_policy_conflict_denies_even_when_allow_policy_set(db):
+    source = Source.objects.create(
+        name="Example jobs",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    create_policy_review(
+        source,
+        actor=None,
+        decision=SourcePolicyReview.Decision.DENY,
+        reason="Juridische review: geblokkeerd",
+        scope=SourcePolicyReview.Scope.SOURCE,
+    )
+    decision = assess_url(source.base_url, source=source)
+    assert not decision.allowed
+    assert decision.status == Source.Policy.DENY
+
+
+@pytest.mark.security
+def test_trial_review_can_allow_fetch(db):
+    source = Source.objects.create(
+        name="Example jobs",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.TRIAL,
+        policy=Source.Policy.REVIEW,
+    )
+    create_policy_review(
+        source,
+        actor=None,
+        decision=SourcePolicyReview.Decision.TRIAL,
+        reason="Technische proefrun gestart",
+        scope=SourcePolicyReview.Scope.SOURCE,
+    )
+    decision = assess_url(source.base_url, source=source)
+    assert decision.allowed

tests/security/test_robots.py

@@ -0,0 +1,77 @@
+from __future__ import annotations
+
+import httpx
+import pytest
+from django.utils import timezone
+
+from apps.sources.models import Source, SourceRobotsCache
+from apps.sources.services.robots import assess_robots
+
+
+@pytest.mark.security
+def test_robots_uses_user_agent_matching_and_allows_default(db):
+    source = Source.objects.create(
+        name="Example jobs",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    calls = {"count": 0}
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        calls["count"] += 1
+        assert request.url.path == "/robots.txt"
+        body = (
+            "User-agent: *\n"
+            "Disallow: /admin\n"
+            "User-agent: vacatureradar/0.1 (+local-personal-use)\n"
+            "Allow: /admin\n"
+        )
+        return httpx.Response(200, text=body)
+
+    client = httpx.Client(transport=httpx.MockTransport(handler))
+    decision = assess_robots(
+        source.base_url + "admin",
+        source=source,
+        user_agent="VacatureRadar/0.1 (+local-personal-use)",
+        client=client,
+        now=timezone.now(),
+    )
+    client.close()
+    assert decision.allowed
+    assert calls["count"] == 1
+
+
+@pytest.mark.security
+def test_robots_cache_refreshes_after_ttl(db, settings):
+    source = Source.objects.create(
+        name="Example jobs",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://jobs.example.org/vacatures/",
+        domain="jobs.example.org",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    settings.ROBOTS_CACHE_TTL_SECONDS = 0
+
+    calls = {"count": 0}
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        calls["count"] += 1
+        return httpx.Response(200, text="User-agent: *\nAllow: /")
+
+    client = httpx.Client(transport=httpx.MockTransport(handler))
+    assess_robots(source.base_url, source=source, user_agent="test-agent", client=client, now=timezone.now())
+    assert calls["count"] == 1
+    assess_robots(
+        source.base_url,
+        source=source,
+        user_agent="test-agent",
+        client=client,
+        now=timezone.now(),
+    )
+    assert calls["count"] == 2
+    assert SourceRobotsCache.objects.filter(origin__contains="jobs.example.org").exists()
+    client.close()

tests/security/test_sanitize.py

@@ -0,0 +1,18 @@
+import pytest
+
+from apps.jobs.services.sanitize import sanitize_job_html
+
+
+@pytest.mark.security
+def test_sanitize_removes_script_handlers_forms_and_iframes():
+    dirty = (
+        '<p onclick="alert(1)">Hallo</p><script>alert(1)</script>'
+        '<iframe src="x"></iframe><form><input></form>'
+        '<a href="javascript:alert(1)">x</a>'
+    )
+    clean = sanitize_job_html(dirty)
+    assert "script" not in clean.lower()
+    assert "onclick" not in clean.lower()
+    assert "iframe" not in clean.lower()
+    assert "form" not in clean.lower()
+    assert "javascript:" not in clean.lower()

tests/security/test_url_security.py

@@ -0,0 +1,34 @@
+import socket
+
+import pytest
+
+from apps.sources.services.url_security import UnsafeUrlError, validate_public_url
+
+
+def fake_resolver(address: str):
+    def resolver(host, port, type=socket.SOCK_STREAM):
+        return [(socket.AF_INET, type, 6, "", (address, port))]
+
+    return resolver
+
+
+@pytest.mark.security
+def test_blocks_private_ipv4_after_dns_resolution():
+    with pytest.raises(UnsafeUrlError, match="Niet-publiek"):
+        validate_public_url("https://jobs.example.org/test", resolver=fake_resolver("10.0.0.5"))
+
+
+@pytest.mark.security
+def test_blocks_loopback_and_embedded_credentials():
+    with pytest.raises(UnsafeUrlError):
+        validate_public_url("http://127.0.0.1/admin")
+    with pytest.raises(UnsafeUrlError):
+        validate_public_url("https://user:pass@example.org/")
+
+
+@pytest.mark.security
+def test_accepts_global_address():
+    result = validate_public_url(
+        "https://jobs.example.org/test", resolver=fake_resolver("93.184.216.34")
+    )
+    assert result.hostname == "jobs.example.org"

tests/unit/test_ai.py

@@ -0,0 +1,141 @@
+import json
+from pathlib import Path
+
+import httpx
+import pytest
+from django.test import override_settings
+
+from apps.jobs.models import AiAnalysisCache
+from apps.jobs.services import ai as ai_service
+from apps.jobs.services.ai import analyze_job_text
+
+
+def _evaluation_cases() -> list[dict[str, object]]:
+    path = Path(__file__).resolve().parents[1] / "fixtures" / "ai" / "evaluation_set.json"
+    return json.loads(path.read_text(encoding="utf-8"))["cases"]
+
+
+@pytest.mark.parametrize("case", _evaluation_cases(), ids=lambda case: case["id"])
+def test_ai_evaluation_cases(monkeypatch, case):
+    if case["response_type"] == "disabled":
+        with override_settings(OLLAMA_ENABLED=False, OLLAMA_MODEL=""):
+            result = analyze_job_text(
+                title=case["title"],
+                description=case["description"],
+                content_hash=case["content_hash"],
+            )
+        assert result.status == ai_service.AiAnalysisCache.Status.DISABLED
+        assert result.error_category == "ollama_disabled"
+        assert result.cached is False
+        assert result.summary_nl == ""
+        return
+
+    if case["response_type"] == "timeout":
+        class FakeClient:
+            def __init__(self, *args, **kwargs):
+                pass
+
+            def __enter__(self):
+                return self
+
+            def __exit__(self, exc_type, exc, tb):
+                return False
+
+            def post(self, *args, **kwargs):
+                raise httpx.TimeoutException("timeout")
+
+        monkeypatch.setattr(httpx, "Client", FakeClient)
+    else:
+        class FakeClient:
+            def __init__(self, *args, **kwargs):
+                pass
+
+            def __enter__(self):
+                return self
+
+            def __exit__(self, exc_type, exc, tb):
+                return False
+
+            def post(self, *args, **kwargs):
+                return _fake_response(case["response"])
+
+        monkeypatch.setattr(httpx, "Client", FakeClient)
+
+    with override_settings(
+        OLLAMA_ENABLED=True,
+        OLLAMA_MODEL="local-model",
+        OLLAMA_BASE_URL="http://ollama:11434",
+        OLLAMA_TIMEOUT_SECONDS=2,
+    ):
+        result = analyze_job_text(
+            title=case["title"],
+            description=case["description"],
+            content_hash=case["content_hash"],
+        )
+
+    assert result.status == case["expected_status"]
+    assert result.error_category == case["expected_error_category"]
+    assert result.schema_version == "1.0.0"
+    if result.status == ai_service.AiAnalysisCache.Status.OK:
+        assert result.summary_nl == case["expected_summary"]
+        assert round(result.features.get("support_ratio", 0), 3) == case["expected_support_ratio"]
+        assert result.features.get("evidence") == case["expected_evidence"]
+    if result.status != ai_service.AiAnalysisCache.Status.OK:
+        assert result.features == {}
+
+
+def _fake_response(response_text: object):
+    class FakeResponse:
+        def raise_for_status(self):
+            return None
+
+        def json(self):
+            return {"response": response_text}
+
+    return FakeResponse()
+
+
+def test_ai_cache_reuses_result(monkeypatch):
+    calls = 0
+
+    class FakeClient:
+        def __init__(self, *args, **kwargs):
+            pass
+
+        def __enter__(self):
+            return self
+
+        def __exit__(self, exc_type, exc, tb):
+            return False
+
+        def post(self, *args, **kwargs):
+            nonlocal calls
+            calls += 1
+            return _fake_response(
+                '{"summary_nl":"Gevalideerde IT-ops advertentie.","features":{"support_ratio":0.03,"consultancy_ratio":0.01,"travel_ratio":0.02,"seniority":"senior","evidence":["IT"]},"warnings":[]}'
+            )
+
+    monkeypatch.setattr(httpx, "Client", FakeClient)
+
+    with override_settings(
+        OLLAMA_ENABLED=True,
+        OLLAMA_MODEL="local-model",
+        OLLAMA_BASE_URL="http://ollama:11434",
+    ):
+        first = analyze_job_text(
+            "Cloud Engineer",
+            "Cloud Engineer vacature. IT-ops taken incl. Azure en Linux.",
+            content_hash="cache-hit-case",
+        )
+        second = analyze_job_text(
+            "Cloud Engineer",
+            "Cloud Engineer vacature. IT-ops taken incl. Azure en Linux.",
+            content_hash="cache-hit-case",
+        )
+
+    assert first.status == ai_service.AiAnalysisCache.Status.OK
+    assert second.status == ai_service.AiAnalysisCache.Status.OK
+    assert first.cached is False
+    assert second.cached is True
+    assert calls == 1
+    assert AiAnalysisCache.objects.count() == 1

tests/unit/test_ats_adapters.py

@@ -0,0 +1,123 @@
+from pathlib import Path
+
+from apps.sources.adapters.ats import (
+    GreenhouseAdapter,
+    LeverAdapter,
+    RecruiteeAdapter,
+    SmartRecruitersAdapter,
+    WorkableAdapter,
+)
+
+
+def _load_fixture(name: str) -> str:
+    return Path(name).read_text(encoding="utf-8")
+
+
+def _assert_evidence_fields(job, *, expected_keys):
+    evidence_fields = {item.field_name for item in job.evidence}
+    assert expected_keys.issubset(evidence_fields)
+
+
+def test_greenhouse_adapter_listing_extracts_open_jobs_and_skips_closed():
+    adapter = GreenhouseAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/greenhouse-listing.json"), url="https://boards.greenhouse.io/example")
+
+    assert result.parser_key == "ats-greenhouse"
+    assert len(result.jobs) == 2
+    assert {job.external_id for job in result.jobs} == {"gh-open-1", "gh-open-2"}
+    assert all("closed" not in result.warnings for _ in [0])
+    assert result.confidence >= 0.89
+    _assert_evidence_fields(result.jobs[0], expected_keys={"external_id", "url", "location_text", "date_posted", "employment_types"})
+
+
+def test_greenhouse_adapter_detail_supports_changed_markup():
+    adapter = GreenhouseAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/greenhouse-detail.json"), url="https://boards.greenhouse.io/example/jobs/senior-software-engineer")
+
+    assert len(result.jobs) == 1
+    job = result.jobs[0]
+    assert job.title == "Senior Software Engineer"
+    assert job.external_id == "gh-open-1"
+    assert job.url == "https://boards.greenhouse.io/example/jobs/senior-software-engineer"
+    assert job.description_html
+    _assert_evidence_fields(job, expected_keys={"external_id", "url", "location_text", "date_posted", "valid_through", "employment_types"})
+
+
+def test_lever_adapter_listing_extracts_open_jobs_and_skips_closed():
+    adapter = LeverAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/lever-listing.json"), url="https://jobs.lever.co/example")
+
+    assert result.parser_key == "ats-lever"
+    assert len(result.jobs) == 2
+    assert {job.external_id for job in result.jobs} == {"lv-open-1", "lv-open-2"}
+    assert result.confidence >= 0.89
+
+
+def test_lever_adapter_detail_supports_changed_markup():
+    adapter = LeverAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/lever-detail.json"), url="https://jobs.lever.co/example/platform-engineer")
+
+    assert len(result.jobs) == 1
+    job = result.jobs[0]
+    assert job.title == "Platform Engineer"
+    assert job.raw["position"]["workplaceType"] == "remote"
+    assert "work_type" in job.raw
+    assert job.location_text == "Aalst"
+
+
+def test_recruitee_adapter_listing_extracts_open_jobs_and_skips_closed():
+    adapter = RecruiteeAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/recruitee-listing.json"), url="https://acme.recruitee.com/o/" )
+
+    assert result.parser_key == "ats-recruitee"
+    assert len(result.jobs) == 2
+    assert {job.title for job in result.jobs} == {"System Engineer", "Network Engineer"}
+
+
+def test_recruitee_adapter_detail_supports_changed_markup():
+    adapter = RecruiteeAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/recruitee-detail.json"), url="https://acme.recruitee.com/o/system-engineer")
+
+    assert len(result.jobs) == 1
+    job = result.jobs[0]
+    assert job.external_id == "rc-open-1"
+    assert job.title == "System Engineer"
+    assert job.employment_types == ["permanent"]
+
+
+def test_smartrecruiters_adapter_listing_extracts_open_jobs_and_skips_closed():
+    adapter = SmartRecruitersAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/smartrecruiters-listing.json"), url="https://jobs.smartrecruiters.com/example")
+
+    assert result.parser_key == "ats-smartrecruiters"
+    assert len(result.jobs) == 2
+    assert "sr-closed-1" not in {job.external_id for job in result.jobs}
+
+
+def test_smartrecruiters_adapter_detail_supports_changed_markup():
+    adapter = SmartRecruitersAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/smartrecruiters-detail.json"), url="https://jobs.smartrecruiters.com/example/security-engineer")
+
+    assert len(result.jobs) == 1
+    job = result.jobs[0]
+    assert job.title == "Security Engineer"
+    assert job.workplace_type in {"remote", "hybrid", ""}
+
+
+def test_workable_adapter_listing_extracts_open_jobs_and_skips_closed():
+    adapter = WorkableAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/workable-listing.json"), url="https://apply.workable.com/example")
+
+    assert result.parser_key == "ats-workable"
+    assert len(result.jobs) == 2
+    assert result.jobs[0].title == "Data Engineer"
+
+
+def test_workable_adapter_detail_supports_changed_markup():
+    adapter = WorkableAdapter()
+    result = adapter.extract(_load_fixture("fixtures/ats/workable-detail.json"), url="https://apply.workable.com/example/data-engineer/")
+
+    assert len(result.jobs) == 1
+    job = result.jobs[0]
+    assert job.external_id == "wb-open-1"
+    assert job.location_text == "Brussel"

tests/unit/test_canonicalize.py

@@ -0,0 +1,13 @@
+from apps.sources.services.canonicalize import canonicalize_url, domain_matches
+
+
+def test_canonicalize_url_removes_tracking_and_fragment():
+    value = canonicalize_url(
+        "HTTPS://Jobs.Example.org:443/jobs/../jobs/42/?utm_source=mail&b=2&a=1#apply"
+    )
+    assert value == "https://jobs.example.org/jobs/42/?a=1&b=2"
+
+
+def test_domain_matches_subdomains_but_not_suffix_attack():
+    assert domain_matches("www.linkedin.com", "linkedin.com")
+    assert not domain_matches("linkedin.com.evil.example", "linkedin.com")

tests/unit/test_discovery_rss.py

@@ -0,0 +1,107 @@
+from pathlib import Path
+
+import pytest
+
+from apps.sources.models import Source
+from apps.sources.services.discovery import (
+    SourceCandidate,
+    discover_career_links,
+    discover_from_email,
+    discover_from_feed,
+    discover_from_html,
+    discover_from_sitemap,
+    persist_discovery_candidates,
+)
+
+
+def test_discovery_finds_career_links_and_skips_denied_domains():
+    html = """
+    <a href="/werken-bij">Werken bij ons</a>
+    <a href="https://careers.partner.example/jobs">Jobs partner</a>
+    <a href="https://www.linkedin.com/jobs">Jobs op LinkedIn</a>
+    <a href="/about">Over ons</a>
+    """
+    result = discover_career_links(html, base_url="https://example.org/")
+    assert [item.url for item in result] == [
+        "https://example.org/werken-bij",
+        "https://careers.partner.example/jobs",
+    ]
+    assert result[0].confidence > result[1].confidence
+
+
+def test_discovery_from_html_includes_jsonld_and_feed_links():
+    html = Path("fixtures/discovery/html_candidate_discovery.html").read_text(encoding="utf-8")
+    result = discover_from_html(html, base_url="https://example.org/")
+
+    candidate_urls = {candidate.url: candidate.discovered_from for candidate in result}
+
+    assert candidate_urls["https://example.org/werken-bij"] == "html"
+    assert candidate_urls["https://jobs.example.org/vacatures/"] == "html"
+    assert candidate_urls["https://jobs.example.org/vacatures/data-engineer"] == "html-jsonld"
+    assert candidate_urls["https://jobs.example.org/vacatures/data-engineer/solliciteer"] == "html-jsonld"
+    assert candidate_urls["https://example.org/feed/jobs.xml"] == "html"
+    assert candidate_urls["https://jobs.example.org/vacatures/data-engineer"].startswith("html-jsonld")
+    assert any(
+        c.source_type == Source.Type.RSS and c.reason == "feed" for c in result
+    )
+    assert any(
+        c.source_type == Source.Type.EMPLOYER and c.url == "https://example.org/werken-bij"
+        for c in result
+    )
+
+
+def test_discovery_from_feed_detects_internal_links_only():
+    content = Path("fixtures/discovery/feed_jobs.xml").read_text(encoding="utf-8")
+    result = discover_from_feed(content, base_url="https://jobs.example.org/feed/jobs")
+
+    assert len(result) == 1
+    assert result[0].url == "https://jobs.example.org/vacatures/cloud-architect"
+    assert result[0].source_type == Source.Type.RSS
+
+
+def test_discovery_from_sitemap_prefers_candidates_and_filters_private_or_denied():
+    index_content = Path("fixtures/discovery/sitemap_index.xml").read_text(encoding="utf-8")
+    urlset_content = Path("fixtures/discovery/sitemap_urlset.xml").read_text(encoding="utf-8")
+
+    index_candidates = discover_from_sitemap(index_content, base_url="https://jobs.example.org/")
+    urlset_candidates = discover_from_sitemap(urlset_content, base_url="https://jobs.example.org/")
+
+    assert len(index_candidates) == 1
+    assert index_candidates[0].url == "https://jobs.example.org/sitemap-jobs.xml"
+    assert len(urlset_candidates) >= 2
+    assert any(item.url == "https://jobs.example.org/vacatures/data-engineer" for item in urlset_candidates)
+    assert all("linkedin.com" not in item.url for item in index_candidates + urlset_candidates)
+
+
+def test_discovery_from_email_extracts_employer_domain_root():
+    raw = Path("fixtures/emails/sample_alert.eml").read_bytes()
+    result = discover_from_email(raw)
+    assert [item.url for item in result] == ["https://jobs.example.org"]
+    assert result[0].source_type == Source.Type.EMPLOYER
+
+
+@pytest.mark.django_db
+def test_discovery_persistence_is_idempotent():
+    candidates = [
+        SourceCandidate(
+            url="https://jobs.example.org/vacatures/data-engineer",
+            domain="jobs.example.org",
+            source_type=Source.Type.EMPLOYER,
+            label="Testbron",
+            confidence=0.95,
+            reason="test",
+            discovered_from="unit-test",
+        )
+    ]
+
+    created, updated, skipped = persist_discovery_candidates(candidates)
+    assert (created, updated, skipped) == (1, 0, 0)
+
+    created, updated, skipped = persist_discovery_candidates(candidates)
+    assert (created, updated, skipped) == (0, 0, 1)
+
+    source = Source.objects.get(domain="jobs.example.org", source_type=Source.Type.EMPLOYER)
+    assert source.status == Source.Status.CANDIDATE
+    assert source.policy == Source.Policy.REVIEW
+    assert source.name == "Testbron"
+    assert len(source.discovery_evidence) == 1

tests/unit/test_email_adapter.py

@@ -0,0 +1,12 @@
+from pathlib import Path
+
+from apps.sources.adapters.email_alert import EmailAlertAdapter
+
+
+def test_email_adapter_extracts_job_links_and_skips_unsubscribe():
+    raw = Path("fixtures/emails/sample_alert.eml").read_bytes()
+    result = EmailAlertAdapter().extract_message(raw)
+    urls = {job.url for job in result.jobs}
+    assert "https://jobs.example.org/vacatures/infrastructure-engineer" in urls
+    assert "https://www.linkedin.com/jobs/view/123456" in urls
+    assert all("unsubscribe" not in url for url in urls)

tests/unit/test_employer_resolution.py

@@ -0,0 +1,176 @@
+from __future__ import annotations
+
+from datetime import UTC, datetime
+
+import pytest
+
+from apps.jobs.models import Employer, JobPosting
+from apps.jobs.services.employer_resolution import resolve_direct_employer_match
+from apps.jobs.services.normalization import CanonicalJobDraft
+from apps.sources.models import Source
+
+
+def _draft_for_test(job: JobPosting, *, **overrides) -> CanonicalJobDraft:
+    values = {
+        "source_url": "https://jobs.smartrecruiters.com/example/security-engineer",
+        "canonical_url": "https://jobs.smartrecruiters.com/example/security-engineer",
+        "external_id": "sr-security-123",
+        "title": job.original_title,
+        "normalized_title": job.normalized_title,
+        "job_family": job.job_family,
+        "employer_name": job.employer_name,
+        "employer_domain": "jobs.example.org",
+        "location_text": job.raw_location,
+        "region": job.region,
+        "municipality": job.municipality,
+        "postal_code": job.postal_code,
+        "country": job.country,
+        "workplace_type": job.workplace_type,
+        "employment_types": job.employment_types,
+        "language": "nl",
+        "description_html": "<p>Test</p>",
+        "description_text": "test beschrijving",
+        "date_posted": datetime(2026, 1, 10, tzinfo=UTC),
+        "valid_through": datetime(2026, 6, 1, tzinfo=UTC),
+        "compensation": {},
+        "skills_required": [],
+        "skills_preferred": [],
+        "content_hash": job.content_hash,
+        "canonical_key": "x" * 64,
+        "evidence": [],
+        "raw": {},
+    }
+    values.update(overrides)
+    return CanonicalJobDraft(**values)
+
+
+@pytest.mark.django_db
+def test_resolver_merges_recruiter_alias_to_direct_source():
+    employer = Employer.objects.create(
+        name="Example IT",
+        normalized_name="example it",
+        domain="jobs.example.org",
+    )
+    JobPosting.objects.create(
+        employer=employer,
+        original_title="Security Engineer",
+        normalized_title="security engineer",
+        canonical_url="https://jobs.example.org/vacatures/security-engineer",
+        canonical_key="a" * 64,
+        job_family="security",
+        content_hash="a" * 64,
+        description_text="Security engineer voor beveiliging.",
+        raw_location="Antwerpen",
+        region="Antwerpen",
+        municipality="Antwerpen",
+        workplace_type=JobPosting.Workplace.HYBRID,
+        employment_types=["full_time"],
+        direct_employer=True,
+    )
+    source = Source.objects.create(
+        name="SmartRecruiters",
+        source_type=Source.Type.ATS,
+        base_url="https://jobs.smartrecruiters.com/example",
+        domain="jobs.smartrecruiters.com",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    # Use an existing direct employer job as the resolver target.
+    target_job = JobPosting.objects.first()
+    decision = resolve_direct_employer_match(
+        _draft_for_test(target_job),
+        source=source,
+    )
+    assert decision.job == target_job
+    assert decision.reason == "resolved_direct_match"
+    assert decision.canonical_url == "https://jobs.example.org/vacatures/security-engineer"
+
+
+@pytest.mark.django_db
+def test_resolver_flags_conflicting_location_for_review():
+    employer = Employer.objects.create(
+        name="Example IT",
+        normalized_name="example it",
+        domain="jobs.example.org",
+    )
+    JobPosting.objects.create(
+        employer=employer,
+        original_title="Security Engineer",
+        normalized_title="security engineer",
+        canonical_url="https://jobs.example.org/vacatures/security-engineer",
+        canonical_key="b" * 64,
+        job_family="security",
+        content_hash="b" * 64,
+        description_text="Security engineer voor beveiliging.",
+        raw_location="Antwerpen",
+        region="Antwerpen",
+        municipality="Antwerpen",
+        workplace_type=JobPosting.Workplace.HYBRID,
+        employment_types=["full_time"],
+        direct_employer=True,
+    )
+    source = Source.objects.create(
+        name="SmartRecruiters",
+        source_type=Source.Type.ATS,
+        base_url="https://jobs.smartrecruiters.com/example",
+        domain="jobs.smartrecruiters.com",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    decision = resolve_direct_employer_match(
+        _draft_for_test(
+            JobPosting.objects.first(),
+            location_text="Gent",
+            municipality="Gent",
+            region="Oost-Vlaanderen",
+        ),
+        source=source,
+    )
+    assert decision.job is None
+    assert decision.reason == "review_direct_conflict"
+
+
+@pytest.mark.django_db
+def test_resolver_keeps_candidate_unmerged_when_similarity_is_conservative():
+    employer = Employer.objects.create(
+        name="Example IT",
+        normalized_name="example it",
+        domain="jobs.example.org",
+    )
+    JobPosting.objects.create(
+        employer=employer,
+        original_title="Security Engineer",
+        normalized_title="security engineer",
+        canonical_url="https://jobs.example.org/vacatures/security-engineer",
+        canonical_key="c" * 64,
+        job_family="security",
+        content_hash="c" * 64,
+        description_text="Security engineer voor beveiliging.",
+        raw_location="Antwerpen",
+        region="Antwerpen",
+        municipality="Antwerpen",
+        workplace_type=JobPosting.Workplace.HYBRID,
+        employment_types=["full_time"],
+        direct_employer=True,
+    )
+    source = Source.objects.create(
+        name="StaffingBoard",
+        source_type=Source.Type.ATS,
+        base_url="https://jobs.staffingboard.com/example",
+        domain="jobs.staffingboard.com",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    decision = resolve_direct_employer_match(
+        _draft_for_test(
+            JobPosting.objects.first(),
+            title="Data Analyst",
+            normalized_title="data analyst",
+            location_text="Brussel",
+            employer_name="Another Employer",
+            employer_domain="jobs.staffingboard.com",
+        ),
+        source=source,
+    )
+    assert decision.job is None
+    assert decision.reason == "no_direct_resolution"

tests/unit/test_features_distance.py

@@ -0,0 +1,45 @@
+from apps.jobs.services.distance import CommuteEstimate, estimate_commute, haversine_km
+from apps.jobs.services.features import extract_deterministic_features, term_ratio
+
+
+def test_feature_extraction_and_distance_are_deterministic():
+    text = (
+        "Voor een gemeente zoeken we een consultant. Rijbewijs B en 5 jaar ervaring vereist. "
+        "De rol bevat helpdesk en telefonische support bij klanten."
+    )
+    features = extract_deterministic_features("Support Engineer", text)
+    assert features["support_ratio"] > 0
+    assert features["consultancy_ratio"] > 0
+    assert features["travel_ratio"] > 0
+    assert features["public_sector_signal"] > 0
+    assert features["experience_years_max"] == 5
+    assert term_ratio("", ["x"]) == 0
+    assert 60 < haversine_km(50.9307, 5.3325, 50.8503, 4.3517) < 80
+
+
+def test_conservative_commute_estimate_is_marked_as_estimate():
+    result = estimate_commute(20.0)
+    assert isinstance(result, CommuteEstimate)
+    assert result.is_estimate
+    assert result.minutes > 0
+    assert result.source == "road"
+
+
+def test_custom_commute_estimator_is_accepted():
+    class CustomEstimator:
+        name = "custom"
+        version = "test"
+
+        def estimate(self, distance_km: float) -> CommuteEstimate:
+            return CommuteEstimate(
+                minutes=int(distance_km * 2),
+                km=distance_km,
+                source=self.name,
+                source_version=self.version,
+                confidence=0.7,
+                is_estimate=False,
+            )
+
+    result = estimate_commute(12.0, estimator=CustomEstimator())
+    assert result.minutes == 24
+    assert result.source_version == "test"

tests/unit/test_feedback_learning.py

@@ -0,0 +1,33 @@
+from __future__ import annotations
+
+from apps.jobs.models import Feedback
+from apps.jobs.services.feedback import record_feedback
+
+
+def test_learning_signals_wait_for_min_samples(profile, job, user):
+    first = record_feedback(user=user, job=job, action=Feedback.Action.INTERESTING)
+    second = record_feedback(user=user, job=job, action=Feedback.Action.INTERESTING)
+
+    first_metadata = first.metadata.get("learning", {})
+    second_metadata = second.metadata.get("learning", {})
+    assert first_metadata.get("status") == "queued"
+    assert first_metadata.get("reason_code") == "positive_feedback"
+    assert second_metadata.get("status") == "applied"
+    profile.refresh_from_db()
+    assert profile.weights["content"] == 26.0
+
+
+def test_learning_skip_for_implicit_hide(profile, job, user):
+    feedback = record_feedback(user=user, job=job, action=Feedback.Action.HIDE)
+    metadata = feedback.metadata.get("learning", {})
+    assert metadata["status"] == "queued"
+    assert metadata["reason_code"] == "implicit_hide_no_reason"
+    profile.refresh_from_db()
+    assert profile.weights["content"] == 25.0
+
+
+def test_learning_marks_explicit_title_hide_as_nonlearning(profile, job, user):
+    feedback = record_feedback(user=user, job=job, action=Feedback.Action.HIDE, reason="Titel past niet bij mij")
+    metadata = feedback.metadata.get("learning", {})
+    assert metadata["status"] == "queued"
+    assert metadata["reason_code"] == "non_learning_title"

tests/unit/test_fetcher.py

@@ -0,0 +1,148 @@
+import httpx
+import pytest
+from django.test import override_settings
+
+from apps.sources.services.fetcher import (
+    ContentRejectedError,
+    FetchError,
+    FetchTimeoutError,
+    RateLimitedError,
+    PolicyBlockedError,
+    fetch_url,
+)
+from apps.sources.services.url_security import ValidatedUrl
+
+PUBLIC_DNS = [(None, None, None, None, ("93.184.216.34", 443))]
+
+
+def _patch_dns(monkeypatch):
+    monkeypatch.setattr(
+        "apps.sources.services.url_security.socket.getaddrinfo",
+        lambda *args, **kwargs: PUBLIC_DNS,
+    )
+    # validate_public_url captured the default resolver at definition time, so patch call site too.
+    original = __import__(
+        "apps.sources.services.url_security", fromlist=["validate_public_url"]
+    ).validate_public_url
+
+    def validate(url, **kwargs):
+        return original(url, resolver=lambda *a, **k: PUBLIC_DNS, **kwargs)
+
+    monkeypatch.setattr("apps.sources.services.fetcher.validate_public_url", validate)
+
+
+def test_fetcher_success_redirect_304_and_hash(monkeypatch):
+    _patch_dns(monkeypatch)
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        if request.url.path == "/start":
+            return httpx.Response(302, headers={"location": "/job"})
+        return httpx.Response(
+            200,
+            headers={"content-type": "text/html; charset=utf-8"},
+            content=b"Vacature",
+        )
+
+    client = httpx.Client(transport=httpx.MockTransport(handler))
+    document = fetch_url("https://example.org/start", client=client)
+    assert document.final_url == "https://example.org/job"
+    assert document.text == "Vacature"
+    assert len(document.sha256) == 64
+    client.close()
+
+    client = httpx.Client(
+        transport=httpx.MockTransport(lambda request: httpx.Response(304, headers={"etag": "x"}))
+    )
+    unchanged = fetch_url("https://example.org/job", client=client)
+    assert unchanged.status_code == 304
+    assert unchanged.content == b""
+    client.close()
+
+
+def test_fetcher_respects_rate_limit_and_retry_after(monkeypatch):
+    _patch_dns(monkeypatch)
+    client = httpx.Client(
+        transport=httpx.MockTransport(
+            lambda request: httpx.Response(
+                429, headers={"retry-after": "45"}, content=b"too many requests"
+            )
+        )
+    )
+    with pytest.raises(RateLimitedError) as exc:
+        fetch_url("https://example.org/too-fast", client=client)
+    assert exc.value.retry_after_seconds == 45
+    client.close()
+
+
+def test_fetcher_rejects_dns_rebinding(monkeypatch):
+    _patch_dns(monkeypatch)
+
+    def validate(url: str, **kwargs):
+        if "/final" in url:
+            return ValidatedUrl(url=url, hostname="example.org", port=443, addresses=("198.51.100.12",))
+        return ValidatedUrl(url=url, hostname="example.org", port=443, addresses=("93.184.216.34",))
+
+    monkeypatch.setattr("apps.sources.services.fetcher.validate_public_url", validate)
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        if request.url.path == "/start":
+            return httpx.Response(302, headers={"location": "/final"})
+        return httpx.Response(
+            200,
+            headers={"content-type": "text/html; charset=utf-8"},
+            content=b"Vacature",
+        )
+
+    client = httpx.Client(transport=httpx.MockTransport(handler))
+    with pytest.raises(FetchError, match="DNS-rebindcontrole"):
+        fetch_url("https://example.org/start", client=client)
+    client.close()
+
+
+def test_fetcher_rejects_policy_content_size_and_http(monkeypatch):
+    _patch_dns(monkeypatch)
+    with pytest.raises(PolicyBlockedError):
+        fetch_url("https://linkedin.com/jobs", client=httpx.Client())
+
+    client = httpx.Client(
+        transport=httpx.MockTransport(
+            lambda request: httpx.Response(200, headers={"content-type": "image/png"}, content=b"x")
+        )
+    )
+    with pytest.raises(ContentRejectedError):
+        fetch_url("https://example.org/image", client=client)
+    client.close()
+
+    with override_settings(FETCHER_MAX_BYTES=2):
+        client = httpx.Client(
+            transport=httpx.MockTransport(
+                lambda request: httpx.Response(
+                    200, headers={"content-type": "text/plain"}, content=b"long"
+                )
+            )
+        )
+        with pytest.raises(ContentRejectedError):
+            fetch_url("https://example.org/large", client=client)
+        client.close()
+
+    client = httpx.Client(
+        transport=httpx.MockTransport(lambda request: httpx.Response(500, content=b"error"))
+    )
+    with pytest.raises(FetchError, match="HTTP 500"):
+        fetch_url("https://example.org/error", client=client)
+    client.close()
+
+    client = httpx.Client(
+        transport=httpx.MockTransport(lambda request: (_ for _ in ()).throw(httpx.TimeoutException("timeout")))
+    )
+    with pytest.raises(FetchTimeoutError):
+        fetch_url("https://example.org/slow", client=client)
+    client.close()
+
+
+def test_fetcher_rejects_redirect_without_location(monkeypatch):
+    _patch_dns(monkeypatch)
+    client = httpx.Client(transport=httpx.MockTransport(lambda request: httpx.Response(302)))
+    with pytest.raises(FetchError, match="Location"):
+        fetch_url("https://example.org/start", client=client)
+    client.close()

tests/unit/test_generic_html_adapter.py

@@ -0,0 +1,11 @@
+from pathlib import Path
+
+from apps.sources.adapters.generic_html import GenericHtmlAdapter
+
+
+def test_generic_html_fallback_extracts_core_fields():
+    html = Path("fixtures/pages/sample_generic_job.html").read_text(encoding="utf-8")
+    result = GenericHtmlAdapter().extract(html, url="https://careers.example.net/jobs/1")
+    assert result.jobs[0].title == "Workplace Engineer"
+    assert result.jobs[0].employer_name == "Example Health"
+    assert "Microsoft Intune" in result.jobs[0].description_text

tests/unit/test_geocoding.py

@@ -0,0 +1,210 @@
+from __future__ import annotations
+
+from decimal import Decimal
+from pathlib import Path
+
+import pytest
+from django.core.management import CommandError
+
+from apps.jobs.models import GeocodeLocationLookup
+from apps.jobs.services.geocoding import (
+    CsvGeocodeProvider,
+    GeoPoint,
+    LocationMatch,
+    LocationMatchResult,
+    import_csv_geodata,
+    parse_belgian_location_query,
+    resolve_location,
+    validate_csv_geodata,
+)
+
+
+@pytest.mark.django_db
+def test_parse_belgian_location_query():
+    postal, municipality = parse_belgian_location_query("Gent, 9000")
+    assert postal == "9000"
+    assert municipality == "gent"
+
+    postal, municipality = parse_belgian_location_query("1050 Ixelles")
+    assert postal == "1050"
+    assert municipality == "ixelles"
+
+    postal, municipality = parse_belgian_location_query("Brussel")
+    assert postal is None
+    assert municipality == "brussel"
+
+
+@pytest.mark.django_db
+def test_geocode_provider_resolves_postal_and_municipality_exactly():
+    GeocodeLocationLookup.objects.create(
+        source_name="fixture",
+        source_version="2026-01-01",
+        source_license_name="",
+        source_license_url="",
+        source_metadata={},
+        query_kind="postal",
+        query_value="9000",
+        postal_code="9000",
+        municipality="Gent",
+        region="Vlaams-Brabant",
+        latitude=Decimal("51.0500"),
+        longitude=Decimal("3.7300"),
+        confidence=Decimal("0.90"),
+    )
+    GeocodeLocationLookup.objects.create(
+        source_name="fixture",
+        source_version="2026-01-01",
+        source_license_name="",
+        source_license_url="",
+        source_metadata={},
+        query_kind="municipality",
+        query_value="gent",
+        postal_code="9000",
+        municipality="Gent",
+        region="Vlaams-Brabant",
+        latitude=Decimal("51.0500"),
+        longitude=Decimal("3.7300"),
+        confidence=Decimal("0.90"),
+    )
+
+    provider = CsvGeocodeProvider(source_name="fixture", source_version="2026-01-01")
+    result = resolve_location("9000 Gent", provider)
+    assert result.ambiguous is False
+    assert result.location is not None
+    assert result.location.postal_code == "9000"
+    assert result.location.municipality == "Gent"
+    assert result.location.point == GeoPoint(latitude=51.05, longitude=3.73)
+
+
+@pytest.mark.django_db
+def test_geocode_provider_marks_ambiguous_as_unknown():
+    GeocodeLocationLookup.objects.create(
+        source_name="fixture",
+        source_version="2026-01-01",
+        source_license_name="",
+        source_license_url="",
+        source_metadata={},
+        query_kind="municipality",
+        query_value="brussel",
+        postal_code="1000",
+        municipality="Brussel",
+        region="Brussels Hoofdstedelijk Gewest",
+        latitude=Decimal("50.8500"),
+        longitude=Decimal("4.3500"),
+        confidence=Decimal("0.85"),
+    )
+    GeocodeLocationLookup.objects.create(
+        source_name="fixture",
+        source_version="2026-01-01",
+        source_license_name="",
+        source_license_url="",
+        source_metadata={},
+        query_kind="municipality",
+        query_value="brussel",
+        postal_code="1000-200",
+        municipality="Brussel",
+        region="Brussels Hoofdstedelijk Gewest",
+        latitude=Decimal("50.8400"),
+        longitude=Decimal("4.3400"),
+        confidence=Decimal("0.85"),
+    )
+
+    provider = CsvGeocodeProvider(source_name="fixture", source_version="2026-01-01")
+    result = resolve_location("Brussel", provider)
+    assert result.ambiguous is True
+    assert result.location is None
+
+
+@pytest.mark.django_db
+def test_validate_csv_geodata_flags_duplicates_and_invalid_coordinates(tmp_path: Path):
+    source = tmp_path / "geodata.csv"
+    source.write_text(
+        "postal_code,municipality,region,latitude,longitude\n"
+        "9000,Gent,Vlaams-Brabant,51.05,3.73\n"
+        "9000,Gent,Vlaams-Brabant,51.05,3.74\n"
+    )
+    with pytest.raises(CommandError, match="dubbel record"):
+        validate_csv_geodata(source)
+
+    source.write_text(
+        "postal_code,municipality,region,latitude,longitude\n"
+        "9000,Gent,Vlaams-Brabant,99,3.73\n"
+    )
+    with pytest.raises(CommandError, match="latitude buiten bereik"):
+        validate_csv_geodata(source)
+
+
+@pytest.mark.django_db
+def test_import_csv_geodata_is_atomic_without_replace(tmp_path: Path):
+    source = tmp_path / "geodata.csv"
+    source.write_text(
+        "postal_code,municipality,region,latitude,longitude\n"
+        "9000,Gent,Vlaams-Brabant,51.05,3.73\n"
+        "1040,Brussel,Brussels Hoofdstedelijk Gewest,50.85,4.35\n"
+    )
+    GeocodeLocationLookup.objects.create(
+        source_name="fixture",
+        source_version="manual",
+        source_license_name="",
+        source_license_url="",
+        source_metadata={},
+        query_kind="municipality",
+        query_value="gent",
+        postal_code="9000",
+        municipality="Gent",
+        region="Vlaams-Brabant",
+        latitude=Decimal("51.05"),
+        longitude=Decimal("3.73"),
+        confidence=Decimal("1.00"),
+    )
+    GeocodeLocationLookup.objects.create(
+        source_name="fixture",
+        source_version="manual",
+        source_license_name="",
+        source_license_url="",
+        source_metadata={},
+        query_kind="postal",
+        query_value="9000",
+        postal_code="9000",
+        municipality="Gent",
+        region="Vlaams-Brabant",
+        latitude=Decimal("51.05"),
+        longitude=Decimal("3.73"),
+        confidence=Decimal("1.00"),
+    )
+
+    with pytest.raises(CommandError, match="overschrijven zonder --replace"):
+        import_csv_geodata(source, source_name="fixture", source_version="manual")
+    assert GeocodeLocationLookup.objects.count() == 2
+
+
+def test_resolve_location_keeps_ambiguous_unknown() -> None:
+    class StubProvider:
+        def resolve(self, query: str):
+            return [
+                LocationMatch(
+                    postal_code="1000",
+                    municipality="Brussel",
+                    region="Brussels",
+                    point=None,
+                    confidence=0.85,
+                    source="csv",
+                    source_version="manual",
+                    metadata={},
+                ),
+                LocationMatch(
+                    postal_code="1001",
+                    municipality="Brussel",
+                    region="Brussels",
+                    point=None,
+                    confidence=0.85,
+                    source="csv",
+                    source_version="manual",
+                    metadata={},
+                ),
+            ]
+
+    result = resolve_location("Brussel", StubProvider())
+    assert result.ambiguous is True
+    assert isinstance(result, LocationMatchResult)
+    assert result.location is None

tests/unit/test_jsonld_adapter.py

@@ -0,0 +1,15 @@
+import json
+from pathlib import Path
+
+from apps.sources.adapters.jsonld import JsonLdJobPostingAdapter
+
+
+def test_extracts_sample_jobposting():
+    html = Path("fixtures/pages/sample_jsonld_job.html").read_text(encoding="utf-8")
+    expected = json.loads(Path("fixtures/golden/sample_jsonld_job.json").read_text())
+    result = JsonLdJobPostingAdapter().extract(html, url="https://jobs.example.org/source")
+    assert result.confidence >= 0.9
+    assert len(result.jobs) == 1
+    job = result.jobs[0]
+    for key, value in expected.items():
+        assert getattr(job, key) == value

tests/unit/test_manual_import_service.py

@@ -0,0 +1,87 @@
+from __future__ import annotations
+
+import pytest
+
+from apps.sources.models import Source, SourceRun
+from apps.sources.services.fetcher import FetchedDocument
+from apps.sources.services.manual_import import ManualImportError, import_manual_source
+
+
+def test_manual_import_url_uses_fetch_and_succeeds(monkeypatch, user):
+    def fake_fetch(*_args, **_kwargs):
+        return FetchedDocument(
+            requested_url="https://jobs.example.org/jobs",
+            final_url="https://jobs.example.org/jobs",
+            status_code=200,
+            headers={
+                "content-type": "text/html; charset=utf-8",
+                "etag": "abc",
+                "last-modified": "Mon, 01 Jan 2000 00:00:00 GMT",
+            },
+            content=b"<html><body>vacature</body></html>",
+        )
+
+    def fake_process(document):
+        return {
+            "extracted": 2,
+            "created": 1,
+            "updated": 1,
+            "duplicates": 0,
+            "parser": "manual-test",
+            "warnings": [],
+        }
+
+    monkeypatch.setattr("apps.sources.services.manual_import.fetch_url", fake_fetch)
+    monkeypatch.setattr("apps.sources.services.manual_import.process_raw_document", fake_process)
+
+    summary = import_manual_source(
+        actor=user, source_url="https://jobs.example.org/jobs"
+    )
+    source = Source.objects.get(pk=summary.source_id)
+
+    assert summary.mode == "url"
+    assert summary.extracted_count == 2
+    assert summary.created_count == 1
+    assert source.source_type == Source.Type.MANUAL
+    assert source.domain == "jobs.example.org"
+    assert SourceRun.objects.filter(source=source, status=SourceRun.Status.SUCCESS).exists()
+
+
+def test_manual_import_paste_text_sanitizes_and_skips_fetch(monkeypatch, user):
+    def fail_fetch(*_args, **_kwargs):
+        raise AssertionError("Fetcher should not be called for pasted text.")
+
+    def fake_process(document):
+        return {
+            "extracted": 0,
+            "created": 0,
+            "updated": 0,
+            "duplicates": 0,
+            "parser": "manual-test",
+            "warnings": ["noop"],
+        }
+
+    monkeypatch.setattr("apps.sources.services.manual_import.fetch_url", fail_fetch)
+    monkeypatch.setattr("apps.sources.services.manual_import.process_raw_document", fake_process)
+
+    summary = import_manual_source(
+        actor=user, pasted_text="<script>alert('x')</script>\nVacaturetekst"
+    )
+    source = Source.objects.get(pk=summary.source_id)
+
+    assert summary.mode == "paste"
+    assert source.source_type == Source.Type.MANUAL
+    assert source.base_url.startswith("https://manual-")
+
+
+def test_manual_import_rejects_denied_domain(user):
+    with pytest.raises(ManualImportError):
+        import_manual_source(actor=user, source_url="https://linkedin.com/jobs/123")
+
+
+def test_manual_import_rejects_paste_payload_when_too_large(user, settings):
+    settings.MANUAL_IMPORT_PASTE_MAX_BYTES = 12
+    huge_text = "a" * 20
+
+    with pytest.raises(ManualImportError):
+        import_manual_source(actor=user, pasted_text=huge_text)

tests/unit/test_normalization.py

@@ -0,0 +1,15 @@
+from apps.jobs.services.normalization import (
+    infer_language,
+    normalize_employment_types,
+    normalize_title,
+)
+
+
+def test_title_and_employment_normalization():
+    assert normalize_title("Senior System Engineer (m/v/x) - Fulltime") == "senior system engineer"
+    assert normalize_employment_types(["FULL_TIME", "vast"]) == ["full_time", "permanent"]
+
+
+def test_language_detection_is_conservative():
+    assert infer_language("Je beheert de infrastructuur en werkt met een team") == "nl"
+    assert infer_language("You manage the infrastructure and work with a team") == "en"

tests/unit/test_profiles.py

@@ -0,0 +1,71 @@
+from decimal import Decimal
+
+import pytest
+from django.core.exceptions import ValidationError
+
+from apps.profiles.models import ProfileRevision, SearchProfile
+from apps.profiles.services import apply_feedback_delta, save_profile_revision
+
+
+@pytest.mark.django_db
+def test_profile_validation_snapshot_activation_and_revisions(user, profile):
+    profile.full_clean()
+    snapshot = profile.snapshot()
+    assert snapshot["home_municipality"] == "Hasselt"
+    assert snapshot["home_latitude"] == pytest.approx(50.9307)
+
+    other = SearchProfile.objects.create(user=user, name="Tweede", is_active=False)
+    other.activate()
+    profile.refresh_from_db()
+    other.refresh_from_db()
+    assert other.is_active is True
+    assert profile.is_active is False
+
+    revision = save_profile_revision(other, reason="test")
+    assert revision.version == 1
+    assert ProfileRevision.objects.filter(profile=other).count() == 1
+    second = save_profile_revision(other, reason="test-2")
+    assert second.version == 2
+    other.refresh_from_db()
+    assert other.version == 2
+
+
+@pytest.mark.django_db
+def test_feedback_delta_is_bounded_and_can_be_disabled(profile):
+    apply_feedback_delta(profile, "content", 100)
+    profile.refresh_from_db()
+    assert profile.weights["content"] == 40.0
+    apply_feedback_delta(profile, "content", -100)
+    profile.refresh_from_db()
+    assert profile.weights["content"] == 0.0
+
+    profile.learning_enabled = False
+    profile.weights["skills"] = 20.0
+    profile.save()
+    apply_feedback_delta(profile, "skills", 5)
+    profile.refresh_from_db()
+    assert profile.weights["skills"] == 20.0
+
+
+@pytest.mark.parametrize(
+    ("field", "value"),
+    [
+        ("weights", {"content": "not-a-number"}),
+        ("weights", {"content": -1}),
+        ("home_latitude", Decimal("91")),
+        ("home_longitude", Decimal("181")),
+    ],
+)
+@pytest.mark.django_db
+def test_profile_rejects_invalid_values(profile, field, value):
+    setattr(profile, field, value)
+    with pytest.raises(ValidationError):
+        profile.full_clean()
+
+
+@pytest.mark.django_db
+def test_profile_rejects_inverted_thresholds(profile):
+    profile.recommendation_threshold = 91
+    profile.top_match_threshold = 90
+    with pytest.raises(ValidationError):
+        profile.full_clean()

tests/unit/test_rate_limit.py

@@ -0,0 +1,95 @@
+from __future__ import annotations
+
+from types import SimpleNamespace
+
+import pytest
+from django.core.cache import cache
+from django.test import RequestFactory
+
+from apps.core.rate_limit import clear_rate_limit, is_rate_limited, register_rate_limit_failure
+
+
+@pytest.fixture(autouse=True)
+def _clear_rate_limit_cache():
+    cache.clear()
+    yield
+    cache.clear()
+
+
+def _post_request(payload: dict[str, str], *, username: str | None = None):
+    factory = RequestFactory()
+    request = factory.post("/login/", data=payload)
+    request.META["REMOTE_ADDR"] = "203.0.113.10"
+    if username is None:
+        request.user = SimpleNamespace(is_authenticated=False, is_anonymous=True)
+    else:
+        request.user = SimpleNamespace(is_authenticated=True, pk=42)
+        payload["username"] = username
+    request.user.username = username
+    return request
+
+
+def test_rate_limit_blocks_after_max_attempts_and_reports_block_state():
+    request = _post_request({"username": "alice"}, username="alice")
+
+    state = is_rate_limited(
+        request,
+        namespace="manual_import",
+        max_attempts=2,
+        window_seconds=300,
+        block_seconds=120,
+    )
+    assert not state.is_blocked
+    assert state.attempts == 0
+
+    state = register_rate_limit_failure(
+        request,
+        namespace="manual_import",
+        max_attempts=2,
+        window_seconds=300,
+        block_seconds=120,
+    )
+    assert not state.is_blocked
+    assert state.attempts == 1
+
+    state = register_rate_limit_failure(
+        request,
+        namespace="manual_import",
+        max_attempts=2,
+        window_seconds=300,
+        block_seconds=120,
+    )
+    assert state.is_blocked
+    assert state.remaining_seconds > 0
+
+    assert is_rate_limited(
+        request,
+        namespace="manual_import",
+        max_attempts=2,
+        window_seconds=300,
+        block_seconds=120,
+    ).is_blocked
+
+
+def test_rate_limit_clear_resets_state_for_authenticated_identity():
+    request = _post_request({"username": "alice"}, username="alice")
+
+    state = register_rate_limit_failure(
+        request,
+        namespace="login",
+        max_attempts=1,
+        window_seconds=300,
+        block_seconds=10,
+    )
+    assert state.is_blocked
+
+    clear_rate_limit(request, namespace="login")
+    state = is_rate_limited(
+        request,
+        namespace="login",
+        max_attempts=1,
+        window_seconds=300,
+        block_seconds=10,
+    )
+    assert not state.is_blocked
+    assert state.attempts == 0

tests/unit/test_scheduling.py

@@ -0,0 +1,83 @@
+from datetime import timedelta
+
+import pytest
+from django.utils import timezone
+
+from apps.sources.models import Source, SourceLease
+from apps.sources.services.scheduling import (
+    acquire_source_lease,
+    calculate_failure_backoff_seconds,
+    release_source_lease,
+)
+
+
+@pytest.mark.django_db
+def test_source_lease_blocks_concurrent_workers_and_recovers_when_expired(source):
+    now = timezone.now()
+    lease_a = acquire_source_lease(source_id=source.pk, worker_token="worker-a", now=now)
+    assert lease_a is not None
+    assert acquire_source_lease(source_id=source.pk, worker_token="worker-b", now=now) is None
+
+    assert release_source_lease(source_id=source.pk, worker_token="worker-a", now=now)
+    lease_b = acquire_source_lease(source_id=source.pk, worker_token="worker-b", now=timezone.now())
+    assert lease_b is not None
+
+
+@pytest.mark.django_db
+def test_source_lease_reuses_expired_lease(db, source):
+    now = timezone.now()
+    SourceLease.objects.create(
+        source=source,
+        token="expired-worker",
+        worker_id="expired-worker",
+        expires_at=now - timedelta(minutes=1),
+    )
+    lease = acquire_source_lease(source_id=source.pk, worker_token="recovery-worker", now=now)
+    assert lease is not None
+    assert lease.worker_id == "recovery-worker"
+
+
+@pytest.mark.django_db
+def test_source_lease_allows_origin_concurrency(db):
+    shared_domain = "shared.example.org"
+    source_a = Source.objects.create(
+        name="Source A",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://shared.example.org/jobs/",
+        domain=shared_domain,
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+        max_concurrency=2,
+    )
+    source_b = Source.objects.create(
+        name="Source B",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://shared.example.org/careers/",
+        domain=shared_domain,
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+        max_concurrency=2,
+    )
+    now = timezone.now()
+    lease_a = acquire_source_lease(source_id=source_a.pk, worker_token="worker-a", now=now)
+    lease_b = acquire_source_lease(source_id=source_b.pk, worker_token="worker-b", now=now)
+    assert lease_a is not None
+    assert lease_b is not None
+
+
+@pytest.mark.django_db
+def test_failure_backoff_uses_retry_after_and_bounds():
+    source = Source(
+        name="Temp",
+        source_type=Source.Type.EMPLOYER,
+        base_url="https://retry.example.org/",
+        domain="retry.example.org",
+        status=Source.Status.ACTIVE,
+        policy=Source.Policy.ALLOW,
+    )
+    source.save()
+    backoff = calculate_failure_backoff_seconds(
+        source, failure_count=2, retry_after_seconds=120
+    )
+    assert backoff >= 120
+    assert backoff <= 3600 + 45

tests/unit/test_scoring.py

@@ -0,0 +1,194 @@
+from __future__ import annotations
+
+from decimal import Decimal
+
+import pytest
+
+from apps.jobs.models import AiAnalysisCache
+from apps.jobs.services.ai import AiAnalysis
+from apps.jobs.models import Employer, JobPosting, ScoreRun
+from apps.jobs.services import scoring
+from apps.jobs.services.scoring import calculate_score
+
+
+@pytest.fixture
+def matching_job(db):
+    employer = Employer.objects.create(
+        name="Example Public IT",
+        normalized_name="example public it",
+        domain="jobs.example.org",
+        is_direct_employer=True,
+    )
+    return JobPosting.objects.create(
+        employer=employer,
+        original_title="Infrastructure Engineer",
+        normalized_title="infrastructure engineer",
+        canonical_url="https://jobs.example.org/jobs/1",
+        canonical_key="a" * 64,
+        content_hash="b" * 64,
+        description_text=(
+            "Beheer Microsoft 365 en VMware. Hybride werk en beperkte tweedelijnssupport."
+        ),
+        raw_location="Hasselt, Limburg",
+        region="Limburg",
+        municipality="Hasselt",
+        postal_code="3500",
+        latitude=Decimal("50.930700"),
+        longitude=Decimal("5.332500"),
+        workplace_type=JobPosting.Workplace.HYBRID,
+        employment_types=["full_time", "permanent"],
+        skills_required=["Microsoft 365", "VMware"],
+        analysis_features={"support_ratio": 0.1, "public_sector_signal": 0.8},
+        direct_employer=True,
+        recruiter=False,
+        extraction_confidence=Decimal("0.95"),
+        status=JobPosting.Status.ACTIVE,
+    )
+
+
+def test_matching_job_gets_recommendation(matching_job, profile):
+    result = calculate_score(matching_job, profile)
+    assert result.score >= profile.recommendation_threshold
+    assert result.recommendation in {
+        ScoreRun.Recommendation.STRONG,
+        ScoreRun.Recommendation.POSSIBLE,
+    }
+    assert not result.hard_exclusions
+
+
+def test_hard_title_exclusion_wins(matching_job, profile):
+    matching_job.original_title = "IT Sales Infrastructure Engineer"
+    matching_job.save(update_fields=["original_title"])
+    result = calculate_score(matching_job, profile)
+    assert result.recommendation == ScoreRun.Recommendation.HIDDEN
+    assert any("sales" in reason.lower() for reason in result.hard_exclusions)
+
+
+def test_distance_boundary_inclusief(matching_job, profile, monkeypatch):
+    profile.max_distance_km = 50
+    profile.save(update_fields=["max_distance_km"])
+    monkeypatch.setattr(scoring, "haversine_km", lambda *_args, **_kwargs: 50.0)
+    matching_job.refresh_from_db()
+    matching_job.save(update_fields=["latitude", "longitude"])
+    result = calculate_score(matching_job, profile)
+    assert result.evidence["distance_km"] == 50.0
+    assert result.recommendation != ScoreRun.Recommendation.HIDDEN
+    assert not any("boven maximum" in issue for issue in result.hard_exclusions)
+
+
+def test_remote_job_heeft_geen_afstandsexclusie(matching_job, profile):
+    matching_job.workplace_type = JobPosting.Workplace.REMOTE
+    matching_job.postal_code = None
+    matching_job.municipality = None
+    matching_job.raw_location = None
+    matching_job.save(update_fields=["workplace_type", "postal_code", "municipality", "raw_location"])
+    matching_job.latitude = None
+    matching_job.longitude = None
+    result = calculate_score(matching_job, profile)
+    assert result.recommendation != ScoreRun.Recommendation.HIDDEN
+    assert all("boven maximum" not in issue for issue in result.hard_exclusions)
+
+
+def test_onbekende_locatie_leidt_niet_tot_automatische_uitsluiting(matching_job, profile):
+    matching_job.postal_code = None
+    matching_job.municipality = None
+    matching_job.raw_location = "onbekend"
+    matching_job.latitude = None
+    matching_job.longitude = None
+    matching_job.save(update_fields=["postal_code", "municipality", "raw_location", "latitude", "longitude"])
+    result = calculate_score(matching_job, profile)
+    assert result.evidence["distance_km"] is None
+    assert result.recommendation != ScoreRun.Recommendation.HIDDEN
+
+
+def test_score_is_tijdzoneonafhankelijk(profile, user, matching_job):
+    utc_profile = profile
+    utc_profile.timezone = "UTC"
+    utc_profile.save(update_fields=["timezone"])
+    result_utc = calculate_score(matching_job, utc_profile)
+
+    ny_profile = profile.__class__.objects.create(
+        user=user,
+        name="UTC vergelijken",
+        is_active=False,
+        home_municipality=utc_profile.home_municipality,
+        home_latitude=utc_profile.home_latitude,
+        home_longitude=utc_profile.home_longitude,
+        max_distance_km=utc_profile.max_distance_km,
+        desired_titles=utc_profile.desired_titles,
+        excluded_titles=utc_profile.excluded_titles,
+        desired_skills=utc_profile.desired_skills,
+        excluded_skills=utc_profile.excluded_skills,
+        allowed_employment_types=utc_profile.allowed_employment_types,
+        preferred_workplace=utc_profile.preferred_workplace,
+        preferred_regions=utc_profile.preferred_regions,
+        excluded_regions=utc_profile.excluded_regions,
+        recommendation_threshold=utc_profile.recommendation_threshold,
+        top_match_threshold=utc_profile.top_match_threshold,
+        digest_time=utc_profile.digest_time,
+        quiet_hours_start=utc_profile.quiet_hours_start,
+        quiet_hours_end=utc_profile.quiet_hours_end,
+        learning_enabled=utc_profile.learning_enabled,
+        weights=utc_profile.weights,
+        timezone="America/New_York",
+    )
+    result_ny = calculate_score(matching_job, ny_profile)
+    assert result_ny.score == result_utc.score
+
+
+def test_ai_score_influence_is_opt_in_and_bounded(matching_job, profile, monkeypatch):
+    profile.ai_scoring_enabled = True
+    profile.weights = {**profile.weights, "ai": 999}
+    profile.save(update_fields=["ai_scoring_enabled", "weights"])
+
+    monkeypatch.setattr(
+        scoring,
+        "analyze_job_text",
+        lambda *args, **kwargs: AiAnalysis(
+            features={
+                "support_ratio": 0.1,
+                "consultancy_ratio": 0.0,
+                "travel_ratio": 0.0,
+                "seniority": "senior",
+                "evidence": ["microsoft 365"],
+            },
+            summary_nl="Sterke technische rol.",
+            warnings=[],
+            model="local-model",
+            status=AiAnalysisCache.Status.OK,
+            error_category="",
+            cached=False,
+        ),
+    )
+
+    result = calculate_score(matching_job, profile)
+    assert result.recommendation in {ScoreRun.Recommendation.STRONG, ScoreRun.Recommendation.POSSIBLE}
+    assert result.model_version == "local-model"
+    assert result.evidence["ai"]["status"] == AiAnalysisCache.Status.OK
+    assert result.evidence["ai"]["weight_applied"] == 20.0
+    assert "ai" in result.evidence
+
+
+def test_ai_failure_keeps_deterministic_scoring(profile, matching_job, monkeypatch):
+    profile.ai_scoring_enabled = True
+    profile.weights = {**profile.weights, "ai": 10}
+    profile.save(update_fields=["ai_scoring_enabled", "weights"])
+
+    monkeypatch.setattr(
+        scoring,
+        "analyze_job_text",
+        lambda *args, **kwargs: AiAnalysis(
+            features={},
+            summary_nl="",
+            warnings=["service-timeout"],
+            model="local-model",
+            status=AiAnalysisCache.Status.TIMEOUT,
+            error_category="timeout",
+            cached=False,
+        ),
+    )
+
+    result = calculate_score(matching_job, profile)
+    assert result.evidence["ai"]["status"] == AiAnalysisCache.Status.TIMEOUT
+    assert "AI-analyse" in "".join(result.concerns)
+    assert "components" in result.__dict__

tests/unit/test_settings_security.py

@@ -0,0 +1,52 @@
+from __future__ import annotations
+
+import pytest
+from django.core.exceptions import ImproperlyConfigured
+
+from config.settings import _validate_production_security
+
+
+def test_validate_production_security_skips_when_debug_mode_is_enabled():
+    _validate_production_security(
+        debug=True,
+        secret_key="dev-only-change-me",
+        allowed_hosts=[],
+        csrf_trusted_origins=[],
+        session_cookie_secure=False,
+        csrf_cookie_secure=False,
+        secure_ssl_redirect=False,
+    )
+
+
+@pytest.mark.parametrize(
+    "secret_key, allowed_hosts, csrf_trusted_origins, session_cookie_secure, csrf_cookie_secure, secure_ssl_redirect",
+    [
+        ("", ["example.org"], ["https://example.org"], True, True, True),
+        ("change-me", ["example.org"], ["https://example.org"], True, True, True),
+        ("too-short", ["example.org"], ["https://example.org"], True, True, True),
+        ("a" * 64, [], ["https://example.org"], True, True, True),
+        ("a" * 64, ["example.org"], [], True, True, True),
+        ("a" * 64, ["example.org"], ["http://example.org"], True, True, True),
+        ("a" * 64, ["example.org"], ["https://example.org"], False, True, True),
+        ("a" * 64, ["example.org"], ["https://example.org"], True, False, True),
+        ("a" * 64, ["example.org"], ["https://example.org"], True, True, False),
+    ],
+)
+def test_validate_production_security_requires_secure_settings_for_prod(
+    secret_key,
+    allowed_hosts,
+    csrf_trusted_origins,
+    session_cookie_secure,
+    csrf_cookie_secure,
+    secure_ssl_redirect,
+):
+    with pytest.raises(ImproperlyConfigured):
+        _validate_production_security(
+            debug=False,
+            secret_key=secret_key,
+            allowed_hosts=allowed_hosts,
+            csrf_trusted_origins=csrf_trusted_origins,
+            session_cookie_secure=session_cookie_secure,
+            csrf_cookie_secure=csrf_cookie_secure,
+            secure_ssl_redirect=secure_ssl_redirect,
+        )

tests/unit/test_source_models.py

@@ -0,0 +1,42 @@
+from datetime import timedelta
+
+import pytest
+from django.core.exceptions import ValidationError
+from django.utils import timezone
+
+from apps.sources.models import Source, SourceRun
+
+
+@pytest.mark.django_db
+def test_source_validation_and_scheduling(source):
+    now = timezone.now()
+    source.schedule_after_success(now=now)
+    source.refresh_from_db()
+    assert source.failure_count == 0
+    assert source.next_run_at == now + timedelta(minutes=source.crawl_interval_minutes)
+
+    source.schedule_after_failure(now=now, backoff_minutes=30)
+    source.refresh_from_db()
+    assert source.failure_count == 1
+    assert source.next_run_at == now + timedelta(minutes=30)
+
+    source.base_url = "https://other.example.org/jobs"
+    with pytest.raises(ValidationError):
+        source.full_clean()
+
+    source.base_url = "https://jobs.example.org/jobs"
+    source.policy = Source.Policy.DENY
+    source.status = Source.Status.ACTIVE
+    with pytest.raises(ValidationError):
+        source.full_clean()
+
+
+@pytest.mark.django_db
+def test_source_run_finish_updates_metrics(source):
+    run = SourceRun.objects.create(source=source)
+    run.finish(SourceRun.Status.SUCCESS, http_status=200, extracted_count=3)
+    run.refresh_from_db()
+    assert run.status == SourceRun.Status.SUCCESS
+    assert run.finished_at is not None
+    assert run.http_status == 200
+    assert run.extracted_count == 3