diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
new file mode 100644
index 00000000..8e2f8abd
--- /dev/null
+++ b/.github/workflows/test.yml
@@ -0,0 +1,60 @@
+name: Tests
+
+# Runs the Django test suite on every push to master (our test-server deploy
+# trigger) and on every pull request. A failing run is a red ✗ + email — it
+# reports status only and does not block the push or the deploy.
+on:
+  push:
+    branches: [master]
+  pull_request:
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+
+    # Postgres service mirrors the local-dev / server `db` container
+    # (postgres:16, makeability / admin / password).
+    services:
+      postgres:
+        image: postgres:16
+        env:
+          POSTGRES_DB: makeability
+          POSTGRES_USER: admin
+          POSTGRES_PASSWORD: password
+        ports:
+          - 5432:5432
+        options: >-
+          --health-cmd "pg_isready -U admin -d makeability"
+          --health-interval 10s
+          --health-timeout 5s
+          --health-retries 5
+
+    env:
+      # settings_test.py reads these to reach the Postgres service above.
+      DATABASE_HOST: localhost
+      DATABASE_PORT: 5432
+      DJANGO_ENV: DEBUG
+
+    steps:
+      - uses: actions/checkout@v4
+
+      # Mirror the Dockerfile's system deps: ImageMagick + Ghostscript power the
+      # PDF→thumbnail path that Artifact.save() runs (exercised by the Talk
+      # fixtures); libpq-dev is needed to build psycopg2 from source.
+      - name: Install system dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends imagemagick ghostscript libpq-dev
+          sudo cp imagemagick-policy.xml /etc/ImageMagick-6/policy.xml
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.13"
+          cache: pip
+
+      - name: Install Python dependencies
+        run: pip install -r requirements.txt
+
+      - name: Run tests
+        run: python manage.py test website --settings=makeabilitylab.settings_test --verbosity=2
diff --git a/CLAUDE.md b/CLAUDE.md
index 67f2b6d1..4d68a492 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -37,11 +37,12 @@ A superuser is required to use `/admin` and add content; create one with `python
 
 ## Tests and accessibility checks
 
-- Tests: `python manage.py test website` (inside container). The suite has two styles, both in `website/tests.py`:
+- Tests: `python manage.py test website --settings=makeabilitylab.settings_test` (inside container). The tests live in the `website/tests/` package (one `test_*.py` per concern; Django auto-discovers them) with shared DB fixtures in `website/tests/base.py`. The suite has two styles:
   - **Unit** — `SimpleTestCase` + `MagicMock` for pure logic (formatters, BibTeX generation, etc.); no DB, runs in ms.
-  - **Integration** — `DatabaseTestCase` (subclass of Django's `TestCase`) for view / queryset / template regressions; each test runs in a transaction and rolls back. Has fixture helpers `make_person` / `make_publication` / `make_news_item`.
+  - **Integration** — `DatabaseTestCase` (subclass of Django's `TestCase`, in `tests/base.py`) for view / queryset / template regressions; each test runs in a transaction and rolls back. Has fixture helpers `make_person` / `make_publication` / `make_talk` / `make_news_item`.
   - When fixing a bug reachable through a real queryset, URL, or view, add a regression test in the matching style before applying the fix (matches the tests-first workflow).
-  - **Gotcha:** `website/migrations/` is gitignored, so each env has its own history. If `manage.py test` fails at DB creation with `column "..." already exists`, drop the stale test DB with `docker exec makeabilitylabwebsite-db-1 psql -U admin -d postgres -c "DROP DATABASE IF EXISTS test_makeability;"` and re-run. See #1267 for the durable fix.
+  - **Always use the `--settings=makeabilitylab.settings_test` shim.** It sets `MIGRATION_MODULES = {'website': None}` so the test DB is built directly from the current models, sidestepping the gitignored, per-environment `website/migrations/` history. This is the durable fix for #1267 — without it, a fresh test DB can fail at creation with `column "..." already exists` (old workaround: `docker exec makeabilitylabwebsite-db-1 psql -U admin -d postgres -c "DROP DATABASE IF EXISTS test_makeability;"`).
+  - **CI:** `.github/workflows/test.yml` runs this same command on every push to `master` and every PR (free/unlimited for this public repo). It reports a green ✓ / red ✗ — it does not block pushes or the deploy. See the testing roadmap in #1278.
 - Accessibility (Pa11y CI + Axe, WCAG 2.0 AA): start the site, then `docker-compose -f docker-compose-local-dev.yml --profile testing run --rm a11y`. URLs to scan are configured in `.pa11yci.json`. Run this before submitting UI changes.
 
 ## Deployment
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index bafd6eaf..1b9c7f67 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -249,20 +249,22 @@ A superuser account is required to access the Django admin interface and add con
 
 ## Running the Test Suite
 
-The Python test suite lives in `website/tests.py` and runs inside the website container:
+The Python test suite lives in the `website/tests/` package (one `test_*.py` module per concern — Django auto-discovers them) and runs inside the website container:
 
 ```bash
-docker exec makeabilitylabwebsite-website-1 python manage.py test website
+docker exec makeabilitylabwebsite-website-1 python manage.py test website --settings=makeabilitylab.settings_test
 ```
 
+**Always pass `--settings=makeabilitylab.settings_test`** (see [Troubleshooting tests](#troubleshooting-tests) for why). The same command runs automatically in CI on every push to `master` and every PR — see [Continuous integration](#continuous-integration).
+
 The suite has two complementary styles:
 
 | Style | Base class | What it's for |
 |---|---|---|
 | **Unit** | `SimpleTestCase` + `MagicMock` | Pure logic — formatters, BibTeX generation, single-method behavior. No DB; runs in milliseconds. |
-| **Integration** | `DatabaseTestCase` (subclass of Django's `TestCase`) | View, queryset, template, and URL-routing regressions. Each test runs inside a transaction that is rolled back, so tests stay isolated. |
+| **Integration** | `DatabaseTestCase` (subclass of Django's `TestCase`, in `website/tests/base.py`) | View, queryset, template, and URL-routing regressions. Each test runs inside a transaction that is rolled back, so tests stay isolated. |
 
-The `DatabaseTestCase` base provides `make_person`, `make_publication`, and `make_news_item` helpers built on plain `Model.objects.create()` — use those rather than hand-rolling fixtures.
+The `DatabaseTestCase` base provides `make_person`, `make_publication`, `make_talk`, and `make_news_item` helpers built on plain `Model.objects.create()` — use those rather than hand-rolling fixtures.
 
 ### When to add a test
 
@@ -276,13 +278,19 @@ If a fix is genuinely not unit-testable (FD leaks, `super().save()`-dependent pa
 
 ### Troubleshooting tests
 
-`website/migrations/` is **gitignored** — each environment (your laptop, test, production) maintains its own migration history on disk. This sometimes drifts. If `manage.py test` fails at test-DB creation, the symptoms and fixes are:
+`website/migrations/` is **gitignored** — each environment (your laptop, test, production) maintains its own migration history on disk, which can drift. The `--settings=makeabilitylab.settings_test` shim is the durable fix (#1267): it sets `MIGRATION_MODULES = {'website': None}`, so the test runner builds the `website` schema directly from the current models instead of replaying that local history. **Use the shim and these symptoms shouldn't appear at all.**
+
+If you forget the shim and the legacy `manage.py test website` fails at test-DB creation:
 
-- **`database "test_makeability" already exists`** — a prior failed run left it half-built. Drop and retry:
+- **`database "test_makeability" already exists`** — a prior failed run left it half-built. Drop and retry (or just switch to the shim):
   ```bash
   docker exec makeabilitylabwebsite-db-1 psql -U admin -d postgres -c "DROP DATABASE IF EXISTS test_makeability;"
   ```
-- **`column "..." of relation "..." already exists`** — a local migration file duplicates a field that a later `0001_initial` regeneration already includes. Same fix (drop the test DB) usually clears it; if it persists, the offending migration is a local stale artifact that needs to be edited or removed. See [#1267](https://github.com/makeabilitylab/makeabilitylabwebsite/issues/1267) for the durable fix (test-only settings shim using `MIGRATION_MODULES`).
+- **`column "..." of relation "..." already exists`** — a local migration file duplicates a field that a later `0001_initial` regeneration already includes. The shim sidesteps this entirely.
+
+### Continuous integration
+
+`.github/workflows/test.yml` runs the suite (with the test-settings shim, against a Postgres 16 service container) on every push to `master` and every pull request. GitHub Actions is free and unlimited for this public repo. A failing run shows a red ✗ on the commit/PR and emails the author — it **reports** status, it does not block the push or the test-server deploy. The broader testing roadmap (coverage, Pa11y-in-CI, test backfill) is tracked in [#1278](https://github.com/makeabilitylab/makeabilitylabwebsite/issues/1278).
 
 ## Accessibility Testing
 
@@ -317,7 +325,7 @@ Edit `.pa11yci.json` to add or remove URLs to test. The `urls` array lists every
 
 - **One issue per branch**: Keep PRs focused on a single issue for easier review.
 
-- **Run the test suite**: `docker exec makeabilitylabwebsite-website-1 python manage.py test website` should pass before opening a PR. If your fix is reachable through a real queryset, view, or template, add a regression test (see [Running the Test Suite](#running-the-test-suite)).
+- **Run the test suite**: `docker exec makeabilitylabwebsite-website-1 python manage.py test website --settings=makeabilitylab.settings_test` should pass before opening a PR (CI runs the same command). If your fix is reachable through a real queryset, view, or template, add a regression test (see [Running the Test Suite](#running-the-test-suite)).
 
 - **Test locally**: Verify your changes work in the browser before submitting.
 
diff --git a/docs/DEPLOYMENT.md b/docs/DEPLOYMENT.md
index c92e1588..b065333e 100644
--- a/docs/DEPLOYMENT.md
+++ b/docs/DEPLOYMENT.md
@@ -97,7 +97,7 @@ View current and past versions on the [Releases page](https://github.com/makeabi
 2. Confirm the Python test suite passes locally:
 
    ```bash
-   docker exec makeabilitylabwebsite-website-1 python manage.py test website
+   docker exec makeabilitylabwebsite-website-1 python manage.py test website --settings=makeabilitylab.settings_test
    ```
 
    (See [Running the Test Suite](../CONTRIBUTING.md#running-the-test-suite) in `CONTRIBUTING.md` for what the suite covers and how to add to it.)
diff --git a/makeabilitylab/settings_test.py b/makeabilitylab/settings_test.py
new file mode 100644
index 00000000..fede8304
--- /dev/null
+++ b/makeabilitylab/settings_test.py
@@ -0,0 +1,45 @@
+"""
+Test-only Django settings.
+
+Run with:
+    python manage.py test website --settings=makeabilitylab.settings_test
+
+Why this exists
+---------------
+``website/migrations/`` is gitignored, so every environment (laptop, CI, the
+servers) carries its own migration history. That drift intermittently breaks a
+fresh test-DB build with ``column "..." already exists`` (see #1267 and
+CLAUDE.md). Setting ``MIGRATION_MODULES = {'website': None}`` tells Django to
+ignore the website app's migration history entirely and build its tables
+directly from the current models during test-DB setup (run_syncdb), which is
+both reproducible across environments and the durable fix for that flakiness.
+
+Only the *website* app is affected; third-party apps (admin, auth, ckeditor,
+sortedm2m, easy_thumbnails, image_cropping, ...) keep their shipped migrations.
+"""
+import os
+
+from makeabilitylab.settings import *  # noqa: F401,F403
+
+# Build website tables from models instead of replaying gitignored migrations.
+MIGRATION_MODULES = {"website": None}
+
+# Let CI point the database at its Postgres service container. Locally (inside
+# the website container) these env vars are unset, so we inherit HOST='db' from
+# the base settings fallback; CI sets DATABASE_HOST=localhost.
+DATABASES["default"]["HOST"] = os.environ.get(  # noqa: F405
+    "DATABASE_HOST", DATABASES["default"]["HOST"]  # noqa: F405
+)
+DATABASES["default"]["PORT"] = os.environ.get(  # noqa: F405
+    "DATABASE_PORT", DATABASES["default"].get("PORT", "5432")  # noqa: F405
+)
+
+# The base settings wire a RotatingFileHandler to /code/media/debug.log (a
+# container path). Django evaluates LOGGING at startup, so on any host without
+# that directory — a CI runner, a fresh checkout — django.setup() crashes
+# before a single test runs. Swap just the 'file' handler for a no-op; this
+# keeps every logger's handler reference valid while never touching disk.
+LOGGING["handlers"]["file"] = {"class": "logging.NullHandler"}  # noqa: F405
+
+# Speed up the auth tests (Data Health suite creates real superuser rows).
+PASSWORD_HASHERS = ["django.contrib.auth.hashers.MD5PasswordHasher"]
diff --git a/website/tests.py b/website/tests.py
deleted file mode 100644
index 589f2785..00000000
--- a/website/tests.py
+++ /dev/null
@@ -1,1578 +0,0 @@
-"""
-Tests for the website app.
-
-Currently covers website.utils.bio_utils.auto_generate_bio. Tests use mocks
-rather than real Person/Position records so they don't have to navigate
-Person.save() side-effects (random image assignment, file I/O); the unit
-under test is pure presentation logic.
-"""
-
-from datetime import date, timedelta
-from unittest.mock import MagicMock, patch
-
-from django.test import SimpleTestCase
-
-from website.utils.bio_utils import (
-    _join_with_oxford_comma,
-    auto_generate_bio,
-    humanize_duration,
-)
-
-
-# --- Mock fixtures ---------------------------------------------------------
-
-
-class FakeQuerySet:
-    """Minimal stand-in for a Django QuerySet with the methods auto-bio uses."""
-
-    def __init__(self, items):
-        self._items = list(items)
-
-    def count(self):
-        return len(self._items)
-
-    def __iter__(self):
-        return iter(self._items)
-
-    def __getitem__(self, key):
-        return self._items[key]
-
-
-def _make_person(
-    *,
-    first_name="Jon",
-    last_name="Doe",
-    url_name="jondoe",
-    has_started=True,
-    is_current_member=False,
-    is_alumni_member=False,
-    is_current_collaborator=False,
-    is_past_collaborator=False,
-    is_active=False,
-    current_title="PhD Student",
-    latest_position=None,
-    publications=0,
-    projects=None,
-    mentors=None,
-    mentees=None,
-    total_time_as_member=None,
-):
-    """Build a mock Person exposing only the attributes auto_generate_bio reads."""
-    # If the caller indicates the person has started, default to a non-None
-    # latest_position so the "no positions at all" early-return doesn't
-    # short-circuit the role-branch tests. Callers exercising the no-position
-    # case pass has_started=False and leave latest_position as None.
-    if latest_position is None and has_started:
-        latest_position = _make_position(
-            start_date=date(2020, 1, 1), title=current_title
-        )
-
-    person = MagicMock(name=f"Person({first_name} {last_name})")
-    person.first_name = first_name
-    person.last_name = last_name
-    person.get_full_name.return_value = f"{first_name} {last_name}"
-    person.get_url_name.return_value = url_name
-    person.has_started = has_started
-    person.is_current_member = is_current_member
-    person.is_alumni_member = is_alumni_member
-    person.is_current_collaborator = is_current_collaborator
-    person.is_past_collaborator = is_past_collaborator
-    person.is_active = is_active
-    person.get_current_title = current_title
-    person.get_latest_position = latest_position
-    person.get_total_time_as_member = total_time_as_member
-    person.get_projects = projects or []
-
-    publication_set = MagicMock()
-    publication_set.exists.return_value = publications > 0
-    publication_set.count.return_value = publications
-    person.publication_set = publication_set
-
-    person.get_grad_mentors.return_value = FakeQuerySet(mentors or [])
-    person.get_mentees.return_value = FakeQuerySet(mentees or [])
-    return person
-
-
-def _make_link_person(first_name, last_name, url_name):
-    p = MagicMock()
-    p.get_full_name.return_value = f"{first_name} {last_name}"
-    p.get_url_name.return_value = url_name
-    return p
-
-
-def _make_project(name, short_name):
-    p = MagicMock()
-    p.name = name
-    p.short_name = short_name
-    return p
-
-
-def _make_position(start_date=None, end_date=None, title="PhD Student"):
-    pos = MagicMock()
-    pos.start_date = start_date
-    pos.end_date = end_date
-    pos.title = title
-    return pos
-
-
-# --- Role-sentence reachability matrix -------------------------------------
-
-
-class RoleSentenceTests(SimpleTestCase):
-    """One test per branch of _role_sentence."""
-
-    def test_no_position_no_publication_returns_empty(self):
-        """
-        Regression for #1258: a Person with no Position records AND no
-        publications must produce an empty bio, not "will be joining…".
-        """
-        person = _make_person(has_started=False, publications=0)
-        self.assertEqual(auto_generate_bio(person), "")
-
-    def test_no_position_with_publication_says_has_published(self):
-        person = _make_person(has_started=False, publications=3)
-        bio = auto_generate_bio(person)
-        self.assertIn("Jon Doe has published with the Makeability Lab.", bio)
-        # Contributions sentence follows.
-        self.assertIn("3 publications", bio)
-
-    def test_future_member_says_will_be_joining_with_pretty_date(self):
-        future_pos = _make_position(start_date=date(2026, 9, 15))
-        person = _make_person(has_started=False, latest_position=future_pos)
-        bio = auto_generate_bio(person)
-        self.assertEqual(
-            bio,
-            "Jon Doe will be joining the Makeability Lab on Sep 2026.",
-        )
-
-    def test_current_member_with_duration(self):
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            total_time_as_member=timedelta(days=int(365 * 3 + 180)),
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn(
-            "Jon Doe is currently a PhD Student in the Makeability Lab.", bio
-        )
-        self.assertIn("Jon has been in the lab for", bio)
-        self.assertIn("years.", bio)
-
-    def test_current_member_without_duration_omits_duration_sentence(self):
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            total_time_as_member=None,
-        )
-        bio = auto_generate_bio(person)
-        self.assertEqual(
-            bio, "Jon Doe is currently a PhD Student in the Makeability Lab."
-        )
-
-    def test_current_member_ms_uses_an_article(self):
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="MS Student",
-            total_time_as_member=None,
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("is currently an MS Student", bio)
-
-    @patch("website.utils.bio_utils._get_earliest_member_position")
-    @patch("website.utils.bio_utils._get_latest_member_position")
-    def test_alumni_member_branch(self, mock_latest, mock_earliest):
-        pos = _make_position(
-            start_date=date(2018, 9, 1),
-            end_date=date(2024, 3, 15),
-            title="PhD Student",
-        )
-        mock_latest.return_value = pos
-        mock_earliest.return_value = pos
-        person = _make_person(
-            is_alumni_member=True,
-            current_title="PhD Student",
-            total_time_as_member=timedelta(days=int(365 * 5.5)),
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("Jon Doe was a PhD Student in the Makeability Lab", bio)
-        self.assertIn("(Sep 2018 to Mar 2024).", bio)
-        # Duration appears between the title and the date range.
-        self.assertIn("years (Sep 2018 to Mar 2024).", bio)
-
-    @patch("website.utils.bio_utils._get_earliest_member_position")
-    @patch("website.utils.bio_utils._get_latest_member_position")
-    def test_alumni_member_now_collaborator_gets_two_sentences(
-        self, mock_latest, mock_earliest
-    ):
-        """
-        Pre-fix bug: a former member who later became a current collaborator
-        was described as "was a Collaborator … (… to present)". Now the
-        first sentence anchors on the latest MEMBER position, and the
-        current collaborator status is a separate trailing sentence.
-        """
-        pos = _make_position(
-            start_date=date(2018, 9, 1),
-            end_date=date(2024, 3, 15),
-            title="PhD Student",
-        )
-        mock_latest.return_value = pos
-        mock_earliest.return_value = pos
-        person = _make_person(
-            is_alumni_member=True,
-            is_current_collaborator=True,
-            current_title="Collaborator",
-            total_time_as_member=timedelta(days=int(365 * 5.5)),
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn(
-            "Jon Doe was a PhD Student in the Makeability Lab", bio
-        )
-        self.assertIn("(Sep 2018 to Mar 2024).", bio)
-        self.assertIn(
-            "Jon is currently a collaborator with the Makeability Lab.", bio
-        )
-        # The first sentence must NOT say "Collaborator" (the title attribute
-        # of get_latest_position would say that — bug we're guarding against).
-        self.assertNotIn("was a Collaborator", bio)
-        self.assertNotIn("to present", bio)
-
-    def test_current_collaborator(self):
-        person = _make_person(
-            is_current_collaborator=True, is_active=True
-        )
-        bio = auto_generate_bio(person)
-        self.assertEqual(
-            bio, "Jon Doe is a collaborator with the Makeability Lab."
-        )
-
-    def test_past_collaborator(self):
-        person = _make_person(is_past_collaborator=True)
-        bio = auto_generate_bio(person)
-        self.assertEqual(
-            bio, "Jon Doe was a collaborator with the Makeability Lab."
-        )
-
-
-# --- Contributions sentence ------------------------------------------------
-
-
-class ContributionsSentenceTests(SimpleTestCase):
-    """Shape of the projects/publications sentence."""
-
-    def test_pubs_only_singular(self):
-        person = _make_person(has_started=False, publications=1)
-        bio = auto_generate_bio(person)
-        self.assertIn("They contributed to 1 publication.", bio)
-
-    def test_pubs_only_plural(self):
-        person = _make_person(has_started=False, publications=5)
-        bio = auto_generate_bio(person)
-        self.assertIn("They contributed to 5 publications.", bio)
-
-    def test_one_project_no_pubs(self):
-        proj = _make_project("Sound Watch", "soundwatch")
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            projects=[proj],
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("They contributed to a project called <a href=", bio)
-        self.assertIn(">Sound Watch</a>.", bio)
-
-    def test_two_projects_uses_and(self):
-        projs = [
-            _make_project("Alpha", "alpha"),
-            _make_project("Beta", "beta"),
-        ]
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            projects=projs,
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("They contributed to 2 projects:", bio)
-        self.assertIn(">Alpha</a> and <a", bio)
-
-    def test_three_projects_uses_oxford_comma(self):
-        projs = [
-            _make_project("Alpha", "alpha"),
-            _make_project("Beta", "beta"),
-            _make_project("Gamma", "gamma"),
-        ]
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            projects=projs,
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("3 projects:", bio)
-        # Oxford comma + "and" before the last item.
-        self.assertIn(">Beta</a>, and <a", bio)
-
-    def test_four_or_more_projects_uses_including(self):
-        projs = [_make_project(f"P{c}", f"p{c}") for c in "abcdef"]
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            projects=projs,
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn(f"{len(projs)} projects, including", bio)
-
-    def test_project_and_pubs_joined_with_as_well_as(self):
-        proj = _make_project("Alpha", "alpha")
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            projects=[proj],
-            publications=4,
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("as well as 4 publications.", bio)
-
-
-# --- Mentor / mentee sentences --------------------------------------------
-
-
-class MentorMenteeSentenceTests(SimpleTestCase):
-    def test_no_mentors_no_mentees_no_sentence(self):
-        person = _make_person(
-            is_current_member=True, is_active=True, current_title="PhD Student"
-        )
-        bio = auto_generate_bio(person)
-        self.assertNotIn("mentored", bio)
-
-    def test_mentor_sentence_active_uses_is(self):
-        m = _make_link_person("Alice", "Smith", "alicesmith")
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            mentors=[m],
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("Jon is mentored by", bio)
-        self.assertIn(">Alice Smith</a>.", bio)
-
-    @patch("website.utils.bio_utils._get_earliest_member_position")
-    @patch("website.utils.bio_utils._get_latest_member_position")
-    def test_mentor_sentence_inactive_uses_was(self, mock_latest, mock_earliest):
-        pos = _make_position(
-            start_date=date(2018, 9, 1),
-            end_date=date(2024, 3, 15),
-            title="PhD Student",
-        )
-        mock_latest.return_value = pos
-        mock_earliest.return_value = pos
-        m = _make_link_person("Alice", "Smith", "alicesmith")
-        person = _make_person(
-            is_alumni_member=True,
-            is_active=False,
-            current_title="PhD Student",
-            mentors=[m],
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("Jon was mentored by", bio)
-
-    def test_mentor_sentence_multiple_uses_oxford_comma(self):
-        mentors = [
-            _make_link_person("Alice", "Smith", "alicesmith"),
-            _make_link_person("Bob", "Jones", "bobjones"),
-            _make_link_person("Carol", "Lee", "carollee"),
-        ]
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            mentors=mentors,
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("Jon is mentored by", bio)
-        self.assertIn(", and ", bio)
-
-    def test_mentee_member_uses_during_their_time_intro(self):
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            mentees=[_make_link_person("Mia", "X", "miax")],
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn(
-            "During their time in the lab, Jon mentored 1 Makeability Lab student",
-            bio,
-        )
-
-    def test_mentee_collaborator_uses_has_mentored_intro(self):
-        person = _make_person(
-            is_current_collaborator=True,
-            is_active=True,
-            mentees=[_make_link_person("Mia", "X", "miax")],
-        )
-        bio = auto_generate_bio(person)
-        self.assertNotIn("During their time in the lab", bio)
-        self.assertIn("Jon has mentored 1 Makeability Lab student", bio)
-
-    def test_mentee_count_three_uses_colon(self):
-        mentees = [
-            _make_link_person("Mia", "X", "miax"),
-            _make_link_person("Noah", "Y", "noahy"),
-            _make_link_person("Olivia", "Z", "oliviaz"),
-        ]
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            mentees=mentees,
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("mentored 3 Makeability Lab students:", bio)
-
-    def test_mentee_count_more_than_three_uses_including(self):
-        mentees = [
-            _make_link_person(f"M{i}", "X", f"m{i}x") for i in range(10)
-        ]
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            mentees=mentees,
-        )
-        bio = auto_generate_bio(person)
-        self.assertIn("mentored 10 Makeability Lab students, including", bio)
-
-
-# --- HTML escaping ---------------------------------------------------------
-
-
-class HtmlEscapingTests(SimpleTestCase):
-    """Defensive: free-text spliced into anchor tags must be escaped."""
-
-    def test_person_name_with_angle_brackets_is_escaped(self):
-        person = _make_person(
-            first_name="<script>",
-            last_name="Doe",
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-        )
-        bio = auto_generate_bio(person)
-        self.assertNotIn("<script>", bio)
-        self.assertIn("&lt;script&gt;", bio)
-
-    def test_project_name_with_html_is_escaped(self):
-        person = _make_person(
-            is_current_member=True,
-            is_active=True,
-            current_title="PhD Student",
-            projects=[_make_project("Bad <Project>", "badproject")],
-        )
-        bio = auto_generate_bio(person)
-        self.assertNotIn("<Project>", bio)
-        self.assertIn("Bad &lt;Project&gt;", bio)
-
-
-# --- Helpers ---------------------------------------------------------------
-
-
-class JoinWithOxfordCommaTests(SimpleTestCase):
-    def test_empty(self):
-        self.assertEqual(_join_with_oxford_comma([]), "")
-
-    def test_one(self):
-        self.assertEqual(_join_with_oxford_comma(["A"]), "A")
-
-    def test_two(self):
-        self.assertEqual(_join_with_oxford_comma(["A", "B"]), "A and B")
-
-    def test_three_uses_oxford_comma(self):
-        self.assertEqual(
-            _join_with_oxford_comma(["A", "B", "C"]), "A, B, and C"
-        )
-
-    def test_four_uses_oxford_comma(self):
-        self.assertEqual(
-            _join_with_oxford_comma(["A", "B", "C", "D"]), "A, B, C, and D"
-        )
-
-
-class HumanizeDurationTests(SimpleTestCase):
-    def test_under_a_month(self):
-        self.assertEqual(
-            humanize_duration(timedelta(days=10)), "less than a month"
-        )
-
-    def test_one_month(self):
-        self.assertEqual(humanize_duration(timedelta(days=30)), "1 month")
-
-    def test_six_months(self):
-        self.assertEqual(humanize_duration(timedelta(days=180)), "6 months")
-
-    def test_one_and_a_half_years(self):
-        self.assertEqual(
-            humanize_duration(timedelta(days=int(365 * 1.5))), "1.5 years"
-        )
-
-    def test_returns_years_string_for_long_durations(self):
-        # Exact float value depends on 30-day-month approximation; we only
-        # assert the format here.
-        result = humanize_duration(timedelta(days=365 * 5))
-        self.assertTrue(
-            result.endswith(" years"),
-            msg=f"Expected a years string, got {result!r}",
-        )
-
-
-# --- BibTeX citation regression -------------------------------------------
-
-
-class BibtexCitationTests(SimpleTestCase):
-    """
-    Regression tests for Publication.get_citation_as_bibtex.
-
-    The method previously compared self.pub_venue_type with ``is`` against
-    PubType.JOURNAL / PubType.ARTICLE. Because TextChoices values are strings
-    loaded from the DB, the identity check was always False and every journal
-    or article paper was emitted as @inproceedings{ instead of @article{.
-    These tests pin the corrected behavior.
-    """
-
-    def _make_publication(self, pub_venue_type):
-        """Mock Publication exposing only what get_citation_as_bibtex reads."""
-        pub = MagicMock()
-        pub.pub_venue_type = pub_venue_type
-        pub.get_bibtex_id.return_value = "Doe2020FooCHI20,"
-        pub.authors.all.return_value = []
-        pub.title = "A Test Title"
-        pub.book_title = "Proceedings of Test"
-        pub.get_formatted_forum_name.return_value = "CHI"
-        pub.date.year = 2020
-        # Falsy values below match the "if self.X" guards in the method:
-        # when falsy the optional bibtex fields are skipped without
-        # exercising any further attribute access.
-        pub.series = ""
-        pub.isbn = ""
-        pub.location = ""
-        pub.page_num_start = None
-        pub.page_num_end = None
-        pub.num_pages = None
-        pub.doi = ""
-        pub.official_url = ""
-        pub.acmid = ""
-        pub.publisher = ""
-        return pub
-
-    def test_journal_uses_article_entry(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.JOURNAL)
-        bibtex = Publication.get_citation_as_bibtex(pub)
-        self.assertTrue(
-            bibtex.startswith("@article{"),
-            msg=f"Expected JOURNAL to emit @article{{, got {bibtex[:60]!r}",
-        )
-
-    def test_article_uses_article_entry(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.ARTICLE)
-        bibtex = Publication.get_citation_as_bibtex(pub)
-        self.assertTrue(
-            bibtex.startswith("@article{"),
-            msg=f"Expected ARTICLE to emit @article{{, got {bibtex[:60]!r}",
-        )
-
-    def test_conference_uses_inproceedings_entry(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.CONFERENCE)
-        bibtex = Publication.get_citation_as_bibtex(pub)
-        self.assertTrue(
-            bibtex.startswith("@inproceedings{"),
-            msg=f"Expected CONFERENCE to emit @inproceedings{{, got {bibtex[:60]!r}",
-        )
-
-
-# --- Formatted forum name regression (#988) -------------------------------
-
-
-class FormattedForumNameTests(SimpleTestCase):
-    """
-    Regression tests for Publication.get_formatted_forum_name (#988).
-
-    The previous implementation bundled posters, demos, work-in-progress,
-    and doctoral consortium papers all under one generic
-    "Extended Abstract Proceedings of …" label, and didn't prefix workshop
-    papers at all. Per the issue, each short-form category should get a
-    specific label ("Poster Proceedings of …", "Demo Proceedings of …",
-    "Workshop Proceedings of …", "Work-in-Progress Proceedings of …",
-    "Doctoral Consortium Proceedings of …"). The bare `extended_abstract`
-    boolean field remains the catch-all fallback when the venue type
-    doesn't match a known short-form category.
-    """
-
-    def _make_publication(self, pub_venue_type, *, forum_name="CHI", year=2024,
-                          extended_abstract=False):
-        pub = MagicMock()
-        pub.forum_name = forum_name
-        pub.pub_venue_type = pub_venue_type
-        pub.extended_abstract = extended_abstract
-        pub.date.year = year
-        return pub
-
-    def test_conference_uses_proceedings_of(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.CONFERENCE)
-        self.assertEqual(
-            Publication.get_formatted_forum_name(pub),
-            "Proceedings of CHI 2024",
-        )
-
-    def test_poster_uses_poster_proceedings_of(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.POSTER)
-        self.assertEqual(
-            Publication.get_formatted_forum_name(pub),
-            "Poster Proceedings of CHI 2024",
-        )
-
-    def test_demo_uses_demo_proceedings_of(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.DEMO)
-        self.assertEqual(
-            Publication.get_formatted_forum_name(pub),
-            "Demo Proceedings of CHI 2024",
-        )
-
-    def test_workshop_uses_workshop_proceedings_of(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.WORKSHOP)
-        self.assertEqual(
-            Publication.get_formatted_forum_name(pub),
-            "Workshop Proceedings of CHI 2024",
-        )
-
-    def test_wip_uses_work_in_progress_proceedings_of(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.WIP)
-        self.assertEqual(
-            Publication.get_formatted_forum_name(pub),
-            "Work-in-Progress Proceedings of CHI 2024",
-        )
-
-    def test_doctoral_consortium_uses_doctoral_consortium_proceedings_of(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.DOCTORAL_CONSORTIUM)
-        self.assertEqual(
-            Publication.get_formatted_forum_name(pub),
-            "Doctoral Consortium Proceedings of CHI 2024",
-        )
-
-    def test_journal_has_no_prefix(self):
-        """Journals are not 'Proceedings of …'; the forum name stands alone."""
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.JOURNAL, forum_name="TOCHI")
-        self.assertEqual(
-            Publication.get_formatted_forum_name(pub),
-            "TOCHI 2024",
-        )
-
-    def test_article_has_no_prefix(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.ARTICLE, forum_name="ArXiv")
-        self.assertEqual(
-            Publication.get_formatted_forum_name(pub),
-            "ArXiv 2024",
-        )
-
-    def test_extended_abstract_flag_is_fallback(self):
-        """
-        The `extended_abstract` BooleanField stays as the catch-all when a
-        pub doesn't fit a specific short-form category (e.g. a panel paper
-        marked as a short submission).
-        """
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.PANEL, extended_abstract=True)
-        self.assertEqual(
-            Publication.get_formatted_forum_name(pub),
-            "Extended Abstract Proceedings of CHI 2024",
-        )
-
-    def test_empty_forum_name_returns_empty_string(self):
-        from website.models.publication import Publication, PubType
-        pub = self._make_publication(PubType.CONFERENCE, forum_name="")
-        self.assertEqual(Publication.get_formatted_forum_name(pub), "")
-
-
-# --- Project member count regression --------------------------------------
-
-
-class ProjectCurrentMemberCountTests(SimpleTestCase):
-    """
-    Regression for Project.get_current_member_count: the method built a
-    queryset but never returned it, so the admin column always displayed
-    None. These tests pin both the return value and the filter semantics
-    so a future caller can't quietly re-introduce the bug.
-    """
-
-    def _make_project_with_count(self, count):
-        project = MagicMock()
-        project.projectrole_set.filter.return_value\
-            .values.return_value\
-            .distinct.return_value\
-            .count.return_value = count
-        return project
-
-    def test_returns_count_not_none(self):
-        from website.models.project import Project
-        project = self._make_project_with_count(7)
-        self.assertEqual(Project.get_current_member_count(project), 7)
-
-    def test_filters_on_open_ended_project_role(self):
-        from website.models.project import Project
-        project = self._make_project_with_count(0)
-        Project.get_current_member_count(project)
-        project.projectrole_set.filter.assert_called_with(end_date__isnull=True)
-
-
-# --- serve_pdf regression -------------------------------------------------
-
-
-class ServePdfTests(SimpleTestCase):
-    """
-    Regression tests for website.views.serve_pdf.
-
-    The previous implementation used
-        Publication.objects.get(pdf_file__icontains=filename)
-    which had two failure modes:
-
-    1. Only ``ObjectDoesNotExist`` was caught. When a substring matched
-       multiple publications' stored paths, ``MultipleObjectsReturned``
-       was raised and propagated as a 500 — e.g. requesting "Speech"
-       when both "Froehlich2010Speech.pdf" and
-       "Froehlich2010Speechalytics.pdf" exist.
-    2. ``__icontains`` is a substring match (SQL ``LIKE %x%``), so a
-       probe for ".pdf" matched every PDF in the database. Combined with
-       the fuzzy difflib fallback that redirects on miss, this enabled
-       enumeration of stored filenames.
-
-    The fix uses ``filter(pdf_file__iendswith=filename).first()`` for
-    the exact branch: ``.first()`` can't raise ``MultipleObjectsReturned``,
-    and ``__iendswith`` only matches paths ending in the requested name.
-    The fuzzy difflib fallback is retained for academic-link integrity
-    and is only invoked on actual miss.
-    """
-
-    def test_exact_match_returns_pdf_response(self):
-        """Happy path: an exact filename match returns the PDF inline."""
-        from website.views.serve_pdf import serve_pdf
-        fake_pub = MagicMock()
-        fake_pub.pdf_file.read.return_value = b"%PDF-1.4 fake"
-        fake_pub.pdf_file.name = "publications/Froehlich2018Speech.pdf"
-        with patch("website.views.serve_pdf.Publication") as MockPub:
-            MockPub.objects.filter.return_value.first.return_value = fake_pub
-            response = serve_pdf(MagicMock(), "Froehlich2018Speech.pdf")
-        self.assertEqual(response.status_code, 200)
-        self.assertEqual(response["Content-Type"], "application/pdf")
-
-    def test_uses_iendswith_not_icontains(self):
-        """
-        Regression for the substring-probe enumeration bug. The exact
-        branch must use ``__iendswith``; a request for ".pdf" must
-        therefore miss (since no pub's path equals ".pdf") rather than
-        matching every pub via substring.
-        """
-        from django.http import Http404
-        from website.views.serve_pdf import serve_pdf
-        with patch("website.views.serve_pdf.Publication") as MockPub, patch(
-            "website.views.serve_pdf.get_closest_filename_from_database",
-            return_value=None,
-        ):
-            MockPub.objects.filter.return_value.first.return_value = None
-            with self.assertRaises(Http404):
-                serve_pdf(MagicMock(), ".pdf")
-            MockPub.objects.filter.assert_called_with(pdf_file__iendswith=".pdf")
-
-    def test_no_exact_match_uses_fuzzy_redirect(self):
-        """
-        Stale-external-link preservation. When the exact branch misses,
-        the difflib fallback should redirect to the closest filename it
-        finds. This is the academic-link-integrity feature.
-        """
-        from website.views.serve_pdf import serve_pdf
-        with patch("website.views.serve_pdf.Publication") as MockPub, patch(
-            "website.views.serve_pdf.get_closest_filename_from_database",
-            return_value="publications/Froehlich2018Updated.pdf",
-        ):
-            MockPub.objects.filter.return_value.first.return_value = None
-            response = serve_pdf(MagicMock(), "Froehlich2018Old.pdf")
-        self.assertEqual(response.status_code, 302)
-        self.assertEqual(response.url, "/media/publications/Froehlich2018Updated.pdf")
-
-    def test_no_exact_no_fuzzy_returns_404(self):
-        """If neither the exact branch nor the fuzzy fallback finds anything, 404."""
-        from django.http import Http404
-        from website.views.serve_pdf import serve_pdf
-        with patch("website.views.serve_pdf.Publication") as MockPub, patch(
-            "website.views.serve_pdf.get_closest_filename_from_database",
-            return_value=None,
-        ):
-            MockPub.objects.filter.return_value.first.return_value = None
-            with self.assertRaises(Http404):
-                serve_pdf(MagicMock(), "NoSuchPaper.pdf")
-
-
-# --- Artifact filename check regression -----------------------------------
-
-
-class ArtifactFilenameUpdateCheckTests(SimpleTestCase):
-    """
-    Regression tests for Artifact.do_filenames_need_updating.
-
-    The raw_file and thumbnail branches each compared against
-    ``artifact.pdf_file.name`` (copy-pasted from the pdf_file branch)
-    instead of ``artifact.raw_file.name`` / ``artifact.thumbnail.name``.
-    The bug masked filename drift in those fields: when pdf_file matched
-    but raw_file or thumbnail had a stale name, the function returned
-    False instead of True. These tests pin the per-branch lookup.
-    """
-
-    def _patch_generate(self, value):
-        return patch(
-            "website.models.artifact.Artifact.generate_filename",
-            return_value=value,
-        )
-
-    def test_all_matching_returns_false(self):
-        from website.models.artifact import Artifact
-        with self._patch_generate("Doe2020Title"):
-            artifact = MagicMock()
-            artifact.pdf_file = MagicMock()
-            artifact.pdf_file.name = "publications/Doe2020Title.pdf"
-            artifact.raw_file = MagicMock()
-            artifact.raw_file.name = "publications/Doe2020Title.zip"
-            artifact.thumbnail = MagicMock()
-            artifact.thumbnail.name = "thumbnails/Doe2020Title.jpg"
-            self.assertFalse(Artifact.do_filenames_need_updating(artifact))
-
-    def test_raw_file_mismatch_when_pdf_file_matches(self):
-        """
-        Under the bug the raw_file branch looked at pdf_file.name (which
-        matches) and returned False; the fix makes it look at
-        raw_file.name and correctly report the mismatch.
-        """
-        from website.models.artifact import Artifact
-        with self._patch_generate("Doe2020Title"):
-            artifact = MagicMock()
-            artifact.pdf_file = MagicMock()
-            artifact.pdf_file.name = "publications/Doe2020Title.pdf"
-            artifact.raw_file = MagicMock()
-            artifact.raw_file.name = "publications/StaleName.zip"
-            artifact.thumbnail = None
-            self.assertTrue(Artifact.do_filenames_need_updating(artifact))
-
-    def test_thumbnail_mismatch_when_pdf_file_matches(self):
-        """Same shape as the raw_file regression, for the thumbnail branch."""
-        from website.models.artifact import Artifact
-        with self._patch_generate("Doe2020Title"):
-            artifact = MagicMock()
-            artifact.pdf_file = MagicMock()
-            artifact.pdf_file.name = "publications/Doe2020Title.pdf"
-            artifact.raw_file = None
-            artifact.thumbnail = MagicMock()
-            artifact.thumbnail.name = "thumbnails/StaleName.jpg"
-            self.assertTrue(Artifact.do_filenames_need_updating(artifact))
-
-
-class ArtifactRawFileLabelTests(SimpleTestCase):
-    """
-    Regression tests for Artifact.raw_file_label (issue #1152).
-
-    The talk snippet previously hardcoded "PPTX" next to the raw_file
-    download link, mislabeling .key (Keynote) and any other format. The
-    label is derived from the file extension.
-    """
-
-    def _artifact_with_raw_file(self, name):
-        from website.models.artifact import Artifact
-        artifact = MagicMock(spec=Artifact)
-        artifact.raw_file = MagicMock() if name else None
-        if name:
-            artifact.raw_file.name = name
-        artifact.RAW_FILE_LABELS = Artifact.RAW_FILE_LABELS
-        return artifact
-
-    def _label(self, name):
-        from website.models.artifact import Artifact
-        return Artifact.raw_file_label.fget(self._artifact_with_raw_file(name))
-
-    def test_pptx_label(self):
-        self.assertEqual(self._label("talks/Doe2020Title.pptx"), "PPTX")
-
-    def test_keynote_label(self):
-        self.assertEqual(self._label("talks/Doe2020Title.key"), "Keynote")
-
-    def test_ai_label(self):
-        self.assertEqual(self._label("posters/Doe2020Title.ai"), "AI")
-
-    def test_figma_label(self):
-        self.assertEqual(self._label("talks/Doe2020Title.fig"), "Figma")
-
-    def test_extension_case_insensitive(self):
-        self.assertEqual(self._label("talks/Doe2020Title.PPTX"), "PPTX")
-        self.assertEqual(self._label("talks/Doe2020Title.Key"), "Keynote")
-
-    def test_unknown_extension_falls_back_to_uppercased_ext(self):
-        self.assertEqual(self._label("talks/Doe2020Title.odp"), "ODP")
-
-    def test_no_raw_file_returns_none(self):
-        self.assertIsNone(self._label(None))
-
-    def test_no_extension_returns_none(self):
-        self.assertIsNone(self._label("talks/Doe2020Title"))
-
-
-class VideoAgeTests(SimpleTestCase):
-    """
-    Regression tests for Video.get_age_in_ms (relates to #1091).
-
-    The method computed `datetime.now().date() - self.date`, which raised
-    TypeError when the nullable `date` field was None. It now returns None
-    for a dateless video so video-age.js falls back to the static date.
-    Pure unit tests: Video instances are constructed but never saved.
-    """
-
-    def test_none_date_returns_none(self):
-        from website.models import Video
-        self.assertIsNone(Video(date=None).get_age_in_ms())
-
-    def test_past_date_returns_positive_int(self):
-        from website.models import Video
-        age = Video(date=date.today() - timedelta(days=7)).get_age_in_ms()
-        self.assertIsInstance(age, int)
-        # 7 days in ms, allowing slack for the now() call crossing midnight.
-        self.assertGreaterEqual(age, 6 * 24 * 60 * 60 * 1000)
-
-    def test_future_date_returns_negative_int(self):
-        from website.models import Video
-        age = Video(date=date.today() + timedelta(days=7)).get_age_in_ms()
-        self.assertIsInstance(age, int)
-        self.assertLess(age, 0)
-
-
-# ===========================================================================
-# Database-backed test infrastructure (#1267)
-# ===========================================================================
-#
-# Everything above this line uses SimpleTestCase + MagicMock — no DB, fast.
-# Tests below this line use Django's TestCase, which wraps each test in a
-# transaction that's rolled back at the end. They exercise real model code,
-# real querysets, and the URL/view layer.
-#
-# Why this exists: several 2.3.4 fixes (publications prefetch_related,
-# news null-author guards, delete_unused_files .path guards) shipped
-# without regression tests because the bugs were only reachable through a
-# real queryset or view. The classes below establish the foundation for
-# backfilling that coverage incrementally. See #1267 for the broader plan.
-
-from django.db import connection
-from django.test import TestCase
-from django.test.utils import CaptureQueriesContext
-from django.core.files.uploadedfile import SimpleUploadedFile
-from django.urls import reverse
-
-
-# Minimal 1x1 GIF used to satisfy Person.image / Person.easter_egg without
-# touching the filesystem. Person.save() picks a random Star Wars image when
-# either field is empty, opening a real file from media/. Pre-setting both
-# with this SimpleUploadedFile skips the fallback branch entirely.
-_GIF_1PX = (
-    b"GIF89a\x01\x00\x01\x00\x80\x00\x00\xff\xff\xff\x00\x00\x00"
-    b"!\xf9\x04\x01\x00\x00\x00\x00,\x00\x00\x00\x00\x01\x00\x01"
-    b"\x00\x00\x02\x02D\x01\x00;"
-)
-
-
-def _make_image_upload(name):
-    """Return a SimpleUploadedFile that satisfies an ImageField."""
-    return SimpleUploadedFile(name, _GIF_1PX, content_type="image/gif")
-
-
-class DatabaseTestCase(TestCase):
-    """
-    Shared base for tests that touch the database. Provides small fixture
-    helpers (make_person / make_publication / make_news_item) built on
-    plain Model.objects.create() — no third-party fixture library. Each
-    test runs inside a transaction and is rolled back, so tests stay
-    isolated without manual cleanup.
-
-    Why a base class instead of module-level helpers: subclasses can
-    override the defaults in setUp() and the helpers can grow without
-    cluttering the module namespace.
-    """
-
-    def make_person(self, first_name="Jane", last_name="Doe", **kwargs):
-        """
-        Create and return a Person. Image fields are pre-populated to
-        skip Person.save()'s Star Wars fallback (which reads a real file
-        from media/). Override by passing image=... explicitly.
-        """
-        from website.models import Person
-        kwargs.setdefault(
-            "image", _make_image_upload(f"{first_name}_{last_name}.gif")
-        )
-        kwargs.setdefault(
-            "easter_egg",
-            _make_image_upload(f"{first_name}_{last_name}_egg.gif"),
-        )
-        return Person.objects.create(
-            first_name=first_name, last_name=last_name, **kwargs
-        )
-
-    def make_publication(self, title="A Test Paper", year=2024, **kwargs):
-        """
-        Create and return a Publication with sensible defaults: post-lab-
-        formation date, conference venue, a forum name, and a dummy PDF
-        (display_pub_snippet.html unconditionally renders pub.pdf_file.url,
-        so tests that go through the publications view need one to render).
-        Override via kwargs.
-        """
-        from datetime import date as _date
-        from website.models import Publication
-        from website.models.publication import PubType
-        kwargs.setdefault("date", _date(year, 1, 1))
-        kwargs.setdefault("forum_name", "CHI")
-        kwargs.setdefault("pub_venue_type", PubType.CONFERENCE)
-        kwargs.setdefault(
-            "pdf_file",
-            SimpleUploadedFile(
-                f"{title.replace(' ', '_')}.pdf",
-                b"%PDF-1.4 test",
-                content_type="application/pdf",
-            ),
-        )
-        return Publication.objects.create(title=title, **kwargs)
-
-    def make_talk(self, title="A Test Talk", year=2024, **kwargs):
-        """
-        Create and return a Talk. Artifact.save() generates a thumbnail
-        from pdf_file (via ImageMagick) on every save, so we provide a
-        small valid PDF and let it run; tests that don't care about the
-        thumbnail just ignore it.
-        """
-        from datetime import date as _date
-        from website.models import Talk
-        from website.models.talk import TalkType
-        kwargs.setdefault("date", _date(year, 1, 1))
-        kwargs.setdefault("forum_name", "CHI")
-        kwargs.setdefault("talk_type", TalkType.CONFERENCE_TALK)
-        kwargs.setdefault(
-            "pdf_file",
-            SimpleUploadedFile(
-                f"{title.replace(' ', '_')}.pdf",
-                b"%PDF-1.4 test",
-                content_type="application/pdf",
-            ),
-        )
-        return Talk.objects.create(title=title, **kwargs)
-
-    def make_news_item(self, title="Test News", author=None, **kwargs):
-        """
-        Create and return a News item. `author` is intentionally optional
-        (the FK is nullable with on_delete=SET_NULL) so tests can exercise
-        the authorless code path that caused the original /news/158/ bug.
-        """
-        from datetime import date as _date
-        from website.models import News
-        kwargs.setdefault("date", _date(2024, 1, 1))
-        kwargs.setdefault("content", "Test news body.")
-        return News.objects.create(title=title, author=author, **kwargs)
-
-
-# --- View-level: null author on /news/<id>/ (regression for #1013) --------
-
-
-class NewsItemNullAuthorViewTests(DatabaseTestCase):
-    """
-    Regression for #1013 — a News item with author=None used to crash the
-    news_item view with AttributeError on cur_news_item.author.authored_news.
-    Fixed in 1c0d6c0 by guarding the access. This test pins the behavior
-    so it can't regress silently.
-    """
-
-    def test_news_item_with_null_author_renders_200(self):
-        item = self.make_news_item(title="Authorless News", author=None)
-        response = self.client.get(
-            reverse("website:news_item_by_id", kwargs={"id": item.id})
-        )
-        self.assertEqual(response.status_code, 200)
-        self.assertContains(response, "Authorless News")
-
-    def test_news_item_with_author_still_renders_200(self):
-        """Sanity check: the non-null path also works."""
-        author = self.make_person(first_name="Ada", last_name="Lovelace")
-        item = self.make_news_item(
-            title="Authored News", author=author
-        )
-        response = self.client.get(
-            reverse("website:news_item_by_id", kwargs={"id": item.id})
-        )
-        self.assertEqual(response.status_code, 200)
-        self.assertContains(response, "Authored News")
-
-
-# --- Talk snippet: external_slides_url rendering (#1273) -----------------
-
-
-class TalkExternalSlidesUrlTests(DatabaseTestCase):
-    """
-    Pins the rendering of Talk.external_slides_url in
-    snippets/display_talk_snippet.html — the "Source" link should appear
-    only when the URL is set. Also pins that Poster persists the field
-    (no display surface for posters yet, but the column must exist).
-    """
-
-    def _render_talk_snippet(self, talk):
-        from django.template.loader import render_to_string
-        return render_to_string(
-            "snippets/display_talk_snippet.html",
-            {"talk": talk, "MEDIA_URL": "/media/"},
-        )
-
-    def test_source_link_renders_when_external_slides_url_set(self):
-        talk = self.make_talk(
-            title="Figma Talk",
-            external_slides_url="https://www.figma.com/file/abc123/slides",
-        )
-        html = self._render_talk_snippet(talk)
-        self.assertIn("https://www.figma.com/file/abc123/slides", html)
-        self.assertIn("fa-up-right-from-square", html)
-        # opens-in-new-tab affordance must be present for accessibility
-        self.assertIn('target="_blank"', html)
-        self.assertIn('rel="noopener"', html)
-
-    def test_source_link_absent_when_external_slides_url_blank(self):
-        talk = self.make_talk(title="No-Source Talk")
-        html = self._render_talk_snippet(talk)
-        self.assertNotIn("fa-up-right-from-square", html)
-
-    def test_poster_external_slides_url_round_trips(self):
-        """Schema pin: Poster.external_slides_url must persist to the DB."""
-        from website.models import Poster
-        poster = Poster.objects.create(
-            title="A Test Poster",
-            external_slides_url="https://www.figma.com/file/xyz/poster",
-            pdf_file=SimpleUploadedFile(
-                "test_poster.pdf",
-                b"%PDF-1.4 test",
-                content_type="application/pdf",
-            ),
-        )
-        reloaded = Poster.objects.get(pk=poster.pk)
-        self.assertEqual(
-            reloaded.external_slides_url,
-            "https://www.figma.com/file/xyz/poster",
-        )
-
-
-# --- Query-count: /publications/ prefetch_related (regression for d4f6d65) -
-
-
-class PublicationsViewQueryCountTests(DatabaseTestCase):
-    """
-    Pins the prefetch_related batch on /publications/ (d4f6d65).
-
-    Before the fix, the snippet template iterated pub.authors.all and
-    pub.projects.all per publication, producing 617 queries on prod with
-    ~250 pubs. The fix added .prefetch_related('authors', 'projects',
-    'keywords') to the view's queryset, dropping that to 60.
-
-    The key correctness property is that the query count is bounded by a
-    constant — it must NOT grow with the number of publications. This
-    test creates publications with M2M relations and asserts the count
-    stays under a generous ceiling for two different data sizes. If a
-    future contributor removes the prefetches, the count climbs linearly
-    and the larger-N test fails.
-    """
-
-    # Generous ceiling well above the steady-state count we measured
-    # locally (~15 queries for the publications view). The ceiling exists
-    # to catch order-of-magnitude regressions, not to pin an exact count.
-    QUERY_CEILING = 30
-
-    def _seed_publications(self, count):
-        author = self.make_person(first_name="Ada", last_name="Lovelace")
-        for i in range(count):
-            pub = self.make_publication(
-                title=f"Paper {i}", year=2024
-            )
-            pub.authors.add(author)
-
-    def _capture_publications_query_count(self):
-        with CaptureQueriesContext(connection) as ctx:
-            response = self.client.get(reverse("website:publications"))
-        return response, len(ctx.captured_queries)
-
-    def test_query_count_is_bounded_with_few_pubs(self):
-        self._seed_publications(2)
-        response, count = self._capture_publications_query_count()
-        self.assertEqual(response.status_code, 200)
-        self.assertLessEqual(
-            count,
-            self.QUERY_CEILING,
-            msg=f"Query count {count} exceeded ceiling {self.QUERY_CEILING}",
-        )
-
-    def test_query_count_does_not_grow_with_pub_count(self):
-        """
-        The real regression guard: query count must be roughly the same
-        whether we render 2 pubs or 20. If prefetches are removed, the
-        count grows by a multiple of N and this test will fail.
-        """
-        self._seed_publications(20)
-        response, count = self._capture_publications_query_count()
-        self.assertEqual(response.status_code, 200)
-        self.assertLessEqual(
-            count,
-            self.QUERY_CEILING,
-            msg=(
-                f"Query count {count} exceeded ceiling {self.QUERY_CEILING} "
-                "with 20 publications — prefetch_related likely regressed"
-            ),
-        )
-
-
-# --- Data Health admin dashboard + checks (issue #1276) -------------------
-
-import os
-import shutil
-import tempfile
-
-from django.contrib.auth.models import User
-from django.test import override_settings
-
-from website.admin.data_health.registry import get_check
-from website.utils.name_utils import normalize_person_name, is_default_person_image
-
-
-class NameUtilsTests(SimpleTestCase):
-    """Unit tests for the shared name/image helpers (no DB)."""
-
-    def test_normalize_folds_accents_and_strips_nonalpha(self):
-        self.assertEqual(normalize_person_name("Jon", "Froehlich"), "jonfroehlich")
-        self.assertEqual(normalize_person_name("Renée", "O'Brien"), "reneeobrien")
-        self.assertEqual(normalize_person_name("Jon-Paul", "Smith Jr."), "jonpaulsmithjr")
-
-    def test_same_name_different_case_same_key(self):
-        self.assertEqual(
-            normalize_person_name("Jane", "Doe"),
-            normalize_person_name("jane", "doe"),
-        )
-
-    def test_default_image_detection(self):
-        self.assertTrue(is_default_person_image(None))
-
-        class _Empty:
-            name = ""
-
-        class _StarWars:
-            name = "person/jane_doe_easteregg_starwars_rebel.png"
-
-        class _Real:
-            name = "person/jane_doe.gif"
-
-        self.assertTrue(is_default_person_image(_Empty()))
-        self.assertTrue(is_default_person_image(_StarWars()))
-        self.assertFalse(is_default_person_image(_Real()))
-
-
-class DataHealthAuthTests(DatabaseTestCase):
-    """Superuser gating on the dashboard, detail, and export views."""
-
-    def setUp(self):
-        self.superuser = User.objects.create_superuser("root", "r@example.com", "pw")
-        self.staff = User.objects.create_user(
-            "staffer", "s@example.com", "pw", is_staff=True
-        )
-
-    def test_superuser_can_view_dashboard(self):
-        self.client.force_login(self.superuser)
-        resp = self.client.get(reverse("admin:data_health_dashboard"))
-        self.assertEqual(resp.status_code, 200)
-
-    def test_staff_nonsuperuser_forbidden(self):
-        self.client.force_login(self.staff)
-        for name, args in [
-            ("admin:data_health_dashboard", []),
-            ("admin:data_health_detail", ["duplicate-people"]),
-            ("admin:data_health_export", ["duplicate-people"]),
-        ]:
-            resp = self.client.get(reverse(name, args=args))
-            self.assertEqual(resp.status_code, 403, msg=name)
-
-    def test_anonymous_redirected_to_login(self):
-        resp = self.client.get(reverse("admin:data_health_dashboard"))
-        self.assertEqual(resp.status_code, 302)
-        self.assertIn("/admin/login/", resp["Location"])
-
-    def test_unknown_check_returns_404(self):
-        self.client.force_login(self.superuser)
-        resp = self.client.get(reverse("admin:data_health_detail", args=["nope"]))
-        self.assertEqual(resp.status_code, 404)
-
-
-class DataHealthCsvTests(DatabaseTestCase):
-    """The CSV export streams text/csv as an attachment with the right rows."""
-
-    def setUp(self):
-        self.superuser = User.objects.create_superuser("root", "r@example.com", "pw")
-        self.client.force_login(self.superuser)
-
-    def test_csv_headers_and_content(self):
-        self.make_person(first_name="Jane", last_name="Doe", email="jane@example.com")
-        self.make_person(first_name="Jane", last_name="Doe", email="jane2@example.com")
-        resp = self.client.get(
-            reverse("admin:data_health_export", args=["duplicate-people"])
-        )
-        self.assertEqual(resp.status_code, 200)
-        self.assertEqual(resp["Content-Type"], "text/csv")
-        self.assertIn(
-            'attachment; filename="duplicate-people-', resp["Content-Disposition"]
-        )
-        body = resp.content.decode()
-        self.assertIn("cluster_key", body)  # header row
-        self.assertIn("jane@example.com", body)
-        self.assertIn("jane2@example.com", body)
-
-
-class DuplicatePeopleCheckTests(DatabaseTestCase):
-    def test_clusters_only_multi_person_same_name(self):
-        self.make_person(first_name="Jane", last_name="Doe")
-        self.make_person(first_name="Jane", last_name="Doe")
-        self.make_person(first_name="John", last_name="Smith")  # unique
-        rows = get_check("duplicate-people").get_rows()
-        self.assertEqual({r["cluster_key"] for r in rows}, {"janedoe"})
-        self.assertEqual(len(rows), 2)
-
-    def test_total_refs_counts_relations(self):
-        p1 = self.make_person(first_name="Jane", last_name="Doe")
-        p2 = self.make_person(first_name="Jane", last_name="Doe")
-        pub = self.make_publication(title="A Counted Paper")
-        pub.authors.add(p1)
-        rows = {r["id"]: r for r in get_check("duplicate-people").get_rows()}
-        self.assertEqual(rows[p1.pk]["pub_count"], 1)
-        self.assertEqual(rows[p1.pk]["total_refs"], 1)
-        self.assertEqual(rows[p2.pk]["total_refs"], 0)  # safe-to-delete shell
-
-    def test_clean_data_no_rows(self):
-        self.make_person(first_name="Solo", last_name="Person")
-        self.assertEqual(get_check("duplicate-people").get_rows(), [])
-
-
-class UrlNameCollisionsCheckTests(DatabaseTestCase):
-    def test_detects_forced_collision_and_placeholder(self):
-        from website.models import Person
-
-        self.make_person(first_name="Jane", last_name="Doe")
-        p2 = self.make_person(first_name="Jane", last_name="Doe")  # -> janedoe2
-        p3 = self.make_person(first_name="Solo", last_name="One")
-        # save() auto-dedupes url_name, so force a historical collision directly.
-        Person.objects.filter(pk=p2.pk).update(url_name="janedoe")
-        Person.objects.filter(pk=p3.pk).update(url_name="placeholder")
-
-        rows = {r["url_name"]: r for r in get_check("url-name-collisions").get_rows()}
-        self.assertIn("janedoe", rows)
-        self.assertEqual(rows["janedoe"]["count"], 2)
-        self.assertIn("placeholder", rows)
-
-
-class PublicationQualityCheckTests(DatabaseTestCase):
-    def test_duplicate_titles_and_missing_venue(self):
-        self.make_publication(title="Dup Paper")
-        self.make_publication(title="Dup Paper")  # duplicate normalized title
-        self.make_publication(title="No Venue Paper", forum_name=None)
-        rows = {r["id"]: r for r in get_check("publication-quality").get_rows()}
-        self.assertEqual(len([r for r in rows.values() if r["dup_title"]]), 2)
-        self.assertTrue(
-            any("forum_name" in r["missing_fields"] for r in rows.values())
-        )
-
-
-class ProjectHealthCheckTests(DatabaseTestCase):
-    def test_flags_incomplete_project(self):
-        from website.models import Project
-
-        Project.objects.create(name="Lonely Project", short_name="lonely")
-        rows = {r["name"]: r for r in get_check("project-health").get_rows()}
-        self.assertIn("Lonely Project", rows)
-        self.assertIn("no thumbnail", rows["Lonely Project"]["issues"])
-        self.assertIn("no publication", rows["Lonely Project"]["issues"])
-
-
-class PositionIntegrityCheckTests(DatabaseTestCase):
-    def test_no_position_and_self_advisor(self):
-        from datetime import date as _date
-
-        from website.models import Position
-        from website.models.position import Title
-
-        p_nopos = self.make_person(first_name="No", last_name="Position")
-        p_self = self.make_person(first_name="Self", last_name="Advisor")
-        Position.objects.create(
-            person=p_self,
-            start_date=_date(2020, 1, 1),
-            title=Title.PHD_STUDENT,
-            advisor=p_self,
-        )
-        issues = {
-            (r["person_id"], r["issue"])
-            for r in get_check("position-integrity").get_rows()
-        }
-        self.assertIn((p_nopos.pk, "no position"), issues)
-        self.assertIn((p_self.pk, "self-advisor"), issues)
-
-
-class NewsHealthCheckTests(DatabaseTestCase):
-    def test_flags_missing_slug_and_author(self):
-        from website.models import News
-
-        author = self.make_person(first_name="News", last_name="Writer")
-        n_noauthor = self.make_news_item(title="Orphan News", author=None)
-        n_noslug = self.make_news_item(title="Slugless News", author=author)
-        News.objects.filter(pk=n_noslug.pk).update(slug="")  # save() auto-slugs
-
-        rows = {r["id"]: r for r in get_check("news-health").get_rows()}
-        self.assertIn(n_noauthor.pk, rows)
-        self.assertFalse(rows[n_noauthor.pk]["has_author"])
-        self.assertIn(n_noslug.pk, rows)
-        self.assertTrue(rows[n_noslug.pk]["missing_slug"])
-
-
-class MediaIntegrityCheckTests(DatabaseTestCase):
-    def setUp(self):
-        # Use a throwaway MEDIA_ROOT so the test never pollutes dev media.
-        self._media_dir = tempfile.mkdtemp(prefix="dh_media_")
-        self._override = override_settings(MEDIA_ROOT=self._media_dir)
-        self._override.enable()
-        self.addCleanup(self._override.disable)
-        self.addCleanup(shutil.rmtree, self._media_dir, ignore_errors=True)
-
-    def test_flags_missing_file(self):
-        pub = self.make_publication(title="Vanishing Paper")
-        path = pub.pdf_file.path
-        if os.path.exists(path):
-            os.remove(path)  # simulate a file that disappeared from disk
-        hits = [
-            r
-            for r in get_check("media-integrity").get_rows()
-            if r["type"] == "Publication"
-            and r["id"] == pub.pk
-            and r["status"] == "missing-file"
-        ]
-        self.assertTrue(hits)
-
-
-class DataHealthReadOnlyTests(DatabaseTestCase):
-    def test_get_rows_does_not_mutate_db(self):
-        from website.models import Person, Publication
-
-        self.make_person(first_name="Jane", last_name="Doe")
-        self.make_person(first_name="Jane", last_name="Doe")
-        before = (Person.objects.count(), Publication.objects.count())
-        for slug in ("duplicate-people", "url-name-collisions", "position-integrity"):
-            get_check(slug).get_rows()
-        after = (Person.objects.count(), Publication.objects.count())
-        self.assertEqual(before, after)
-
-
-# --- Latent correctness fixes (audit batch) -------------------------------
-
-
-class PublicationGetPersonTests(DatabaseTestCase):
-    """
-    Regression for Publication.get_person (models/publication.py).
-
-    The method indexed self.authors.all()[0], raising IndexError on an
-    authorless publication (e.g. a PhD dissertation entered before its
-    author is linked). This path is reached from views/people.py, which
-    iterates dissertations and calls get_person() on each. The fix returns
-    None via .first() so the caller can skip authorless rows.
-    """
-
-    def test_no_authors_returns_none(self):
-        pub = self.make_publication(title="Authorless Paper")
-        self.assertIsNone(pub.get_person())
-
-    def test_returns_first_author_in_sorted_order(self):
-        pub = self.make_publication(title="Authored Paper")
-        first = self.make_person(first_name="First", last_name="Author")
-        second = self.make_person(first_name="Second", last_name="Author")
-        pub.authors.add(first, second)  # SortedManyToManyField preserves order
-        self.assertEqual(pub.get_person(), first)
-
-
-class ProjectGetPeopleTimeWorkedTests(DatabaseTestCase):
-    """
-    Regression for the dead conditional in Project.get_people
-    (models/project.py).
-
-    The original time_worked expression was
-        F('end_date') if F('end_date') is not None else timezone.now()
-    but an F() object is never None, so the else branch was dead: ongoing
-    roles (end_date=None) produced NULL durations and a NULL
-    total_time_worked annotation. The fix coalesces end_date to today in
-    the DB via Coalesce().
-    """
-
-    def _make_project(self, name="Test Project", short_name="testproj"):
-        from website.models import Project
-        return Project.objects.create(name=name, short_name=short_name)
-
-    def _add_role(self, person, project, start, end=None):
-        from website.models import ProjectRole
-        return ProjectRole.objects.create(
-            person=person, project=project, start_date=start, end_date=end
-        )
-
-    def test_ongoing_role_has_non_null_time_worked(self):
-        project = self._make_project()
-        person = self.make_person(first_name="Ada", last_name="Lovelace")
-        self._add_role(
-            person, project, date.today() - timedelta(days=100), end=None
-        )
-
-        result = list(project.get_people())  # default sorted_by="time_on_project"
-        self.assertEqual(len(result), 1)
-        time_worked = result[0].total_time_worked
-        self.assertIsNotNone(time_worked)  # was None before the Coalesce fix
-        self.assertGreaterEqual(time_worked, timedelta(days=99))
-
-    def test_long_completed_role_outranks_short_ongoing_role(self):
-        """
-        Discriminating ordering check. With the bug, the ongoing role's
-        NULL duration sorts NULLS-FIRST under Postgres DESC ordering, so
-        the short ongoing role would wrongly rank first. With the fix the
-        long completed role ranks first by actual duration.
-        """
-        project = self._make_project()
-        long_done = self.make_person(first_name="Long", last_name="Veteran")
-        short_ongoing = self.make_person(first_name="Short", last_name="Newcomer")
-        self._add_role(
-            long_done, project,
-            date.today() - timedelta(days=400),
-            end=date.today() - timedelta(days=35),
-        )  # ~365 days
-        self._add_role(
-            short_ongoing, project,
-            date.today() - timedelta(days=10), end=None,
-        )  # ~10 days
-
-        result = list(project.get_people())
-        self.assertEqual(result[0], long_done)
-        self.assertEqual(result[1], short_ongoing)
diff --git a/website/tests/__init__.py b/website/tests/__init__.py
new file mode 100644
index 00000000..da74058d
--- /dev/null
+++ b/website/tests/__init__.py
@@ -0,0 +1,16 @@
+"""
+Test package for the website app.
+
+Two styles live side by side (see the module docstrings):
+
+- **Unit** — ``SimpleTestCase`` + ``unittest.mock`` for pure logic (formatters,
+  BibTeX generation, etc.); no DB, runs in milliseconds.
+- **Integration** — ``DatabaseTestCase`` (in ``tests/base.py``) for
+  view / queryset / template regressions; each test runs in a transaction that
+  is rolled back.
+
+Django auto-discovers any ``test_*.py`` module here, so adding a new test file
+needs no registration. Run the suite with the test-settings shim::
+
+    python manage.py test website --settings=makeabilitylab.settings_test
+"""
diff --git a/website/tests/base.py b/website/tests/base.py
new file mode 100644
index 00000000..dd88604e
--- /dev/null
+++ b/website/tests/base.py
@@ -0,0 +1,123 @@
+"""
+Shared database-backed test infrastructure (#1267).
+
+Tests built on :class:`DatabaseTestCase` use Django's ``TestCase``, which wraps
+each test in a transaction that's rolled back at the end. They exercise real
+model code, real querysets, and the URL/view/template layer.
+
+Why this exists: several 2.3.4 fixes (publications prefetch_related, news
+null-author guards, delete_unused_files .path guards) shipped without
+regression tests because the bugs were only reachable through a real queryset
+or view. This base class establishes the foundation for backfilling that
+coverage incrementally. See #1267 for the broader plan.
+"""
+
+from django.core.files.uploadedfile import SimpleUploadedFile
+from django.test import TestCase
+
+
+# Minimal 1x1 GIF used to satisfy Person.image / Person.easter_egg without
+# touching the filesystem. Person.save() picks a random Star Wars image when
+# either field is empty, opening a real file from media/. Pre-setting both
+# with this SimpleUploadedFile skips the fallback branch entirely.
+_GIF_1PX = (
+    b"GIF89a\x01\x00\x01\x00\x80\x00\x00\xff\xff\xff\x00\x00\x00"
+    b"!\xf9\x04\x01\x00\x00\x00\x00,\x00\x00\x00\x00\x01\x00\x01"
+    b"\x00\x00\x02\x02D\x01\x00;"
+)
+
+
+def _make_image_upload(name):
+    """Return a SimpleUploadedFile that satisfies an ImageField."""
+    return SimpleUploadedFile(name, _GIF_1PX, content_type="image/gif")
+
+
+class DatabaseTestCase(TestCase):
+    """
+    Shared base for tests that touch the database. Provides small fixture
+    helpers (make_person / make_publication / make_news_item) built on
+    plain Model.objects.create() — no third-party fixture library. Each
+    test runs inside a transaction and is rolled back, so tests stay
+    isolated without manual cleanup.
+
+    Why a base class instead of module-level helpers: subclasses can
+    override the defaults in setUp() and the helpers can grow without
+    cluttering the module namespace.
+    """
+
+    def make_person(self, first_name="Jane", last_name="Doe", **kwargs):
+        """
+        Create and return a Person. Image fields are pre-populated to
+        skip Person.save()'s Star Wars fallback (which reads a real file
+        from media/). Override by passing image=... explicitly.
+        """
+        from website.models import Person
+        kwargs.setdefault(
+            "image", _make_image_upload(f"{first_name}_{last_name}.gif")
+        )
+        kwargs.setdefault(
+            "easter_egg",
+            _make_image_upload(f"{first_name}_{last_name}_egg.gif"),
+        )
+        return Person.objects.create(
+            first_name=first_name, last_name=last_name, **kwargs
+        )
+
+    def make_publication(self, title="A Test Paper", year=2024, **kwargs):
+        """
+        Create and return a Publication with sensible defaults: post-lab-
+        formation date, conference venue, a forum name, and a dummy PDF
+        (display_pub_snippet.html unconditionally renders pub.pdf_file.url,
+        so tests that go through the publications view need one to render).
+        Override via kwargs.
+        """
+        from datetime import date as _date
+        from website.models import Publication
+        from website.models.publication import PubType
+        kwargs.setdefault("date", _date(year, 1, 1))
+        kwargs.setdefault("forum_name", "CHI")
+        kwargs.setdefault("pub_venue_type", PubType.CONFERENCE)
+        kwargs.setdefault(
+            "pdf_file",
+            SimpleUploadedFile(
+                f"{title.replace(' ', '_')}.pdf",
+                b"%PDF-1.4 test",
+                content_type="application/pdf",
+            ),
+        )
+        return Publication.objects.create(title=title, **kwargs)
+
+    def make_talk(self, title="A Test Talk", year=2024, **kwargs):
+        """
+        Create and return a Talk. Artifact.save() generates a thumbnail
+        from pdf_file (via ImageMagick) on every save, so we provide a
+        small valid PDF and let it run; tests that don't care about the
+        thumbnail just ignore it.
+        """
+        from datetime import date as _date
+        from website.models import Talk
+        from website.models.talk import TalkType
+        kwargs.setdefault("date", _date(year, 1, 1))
+        kwargs.setdefault("forum_name", "CHI")
+        kwargs.setdefault("talk_type", TalkType.CONFERENCE_TALK)
+        kwargs.setdefault(
+            "pdf_file",
+            SimpleUploadedFile(
+                f"{title.replace(' ', '_')}.pdf",
+                b"%PDF-1.4 test",
+                content_type="application/pdf",
+            ),
+        )
+        return Talk.objects.create(title=title, **kwargs)
+
+    def make_news_item(self, title="Test News", author=None, **kwargs):
+        """
+        Create and return a News item. `author` is intentionally optional
+        (the FK is nullable with on_delete=SET_NULL) so tests can exercise
+        the authorless code path that caused the original /news/158/ bug.
+        """
+        from datetime import date as _date
+        from website.models import News
+        kwargs.setdefault("date", _date(2024, 1, 1))
+        kwargs.setdefault("content", "Test news body.")
+        return News.objects.create(title=title, author=author, **kwargs)
diff --git a/website/tests/test_artifact.py b/website/tests/test_artifact.py
new file mode 100644
index 00000000..65fab6c2
--- /dev/null
+++ b/website/tests/test_artifact.py
@@ -0,0 +1,115 @@
+"""Tests for Artifact model methods (filename-drift check, raw-file label)."""
+
+from unittest.mock import MagicMock, patch
+
+from django.test import SimpleTestCase
+
+
+# --- Artifact filename check regression -----------------------------------
+
+
+class ArtifactFilenameUpdateCheckTests(SimpleTestCase):
+    """
+    Regression tests for Artifact.do_filenames_need_updating.
+
+    The raw_file and thumbnail branches each compared against
+    ``artifact.pdf_file.name`` (copy-pasted from the pdf_file branch)
+    instead of ``artifact.raw_file.name`` / ``artifact.thumbnail.name``.
+    The bug masked filename drift in those fields: when pdf_file matched
+    but raw_file or thumbnail had a stale name, the function returned
+    False instead of True. These tests pin the per-branch lookup.
+    """
+
+    def _patch_generate(self, value):
+        return patch(
+            "website.models.artifact.Artifact.generate_filename",
+            return_value=value,
+        )
+
+    def test_all_matching_returns_false(self):
+        from website.models.artifact import Artifact
+        with self._patch_generate("Doe2020Title"):
+            artifact = MagicMock()
+            artifact.pdf_file = MagicMock()
+            artifact.pdf_file.name = "publications/Doe2020Title.pdf"
+            artifact.raw_file = MagicMock()
+            artifact.raw_file.name = "publications/Doe2020Title.zip"
+            artifact.thumbnail = MagicMock()
+            artifact.thumbnail.name = "thumbnails/Doe2020Title.jpg"
+            self.assertFalse(Artifact.do_filenames_need_updating(artifact))
+
+    def test_raw_file_mismatch_when_pdf_file_matches(self):
+        """
+        Under the bug the raw_file branch looked at pdf_file.name (which
+        matches) and returned False; the fix makes it look at
+        raw_file.name and correctly report the mismatch.
+        """
+        from website.models.artifact import Artifact
+        with self._patch_generate("Doe2020Title"):
+            artifact = MagicMock()
+            artifact.pdf_file = MagicMock()
+            artifact.pdf_file.name = "publications/Doe2020Title.pdf"
+            artifact.raw_file = MagicMock()
+            artifact.raw_file.name = "publications/StaleName.zip"
+            artifact.thumbnail = None
+            self.assertTrue(Artifact.do_filenames_need_updating(artifact))
+
+    def test_thumbnail_mismatch_when_pdf_file_matches(self):
+        """Same shape as the raw_file regression, for the thumbnail branch."""
+        from website.models.artifact import Artifact
+        with self._patch_generate("Doe2020Title"):
+            artifact = MagicMock()
+            artifact.pdf_file = MagicMock()
+            artifact.pdf_file.name = "publications/Doe2020Title.pdf"
+            artifact.raw_file = None
+            artifact.thumbnail = MagicMock()
+            artifact.thumbnail.name = "thumbnails/StaleName.jpg"
+            self.assertTrue(Artifact.do_filenames_need_updating(artifact))
+
+
+class ArtifactRawFileLabelTests(SimpleTestCase):
+    """
+    Regression tests for Artifact.raw_file_label (issue #1152).
+
+    The talk snippet previously hardcoded "PPTX" next to the raw_file
+    download link, mislabeling .key (Keynote) and any other format. The
+    label is derived from the file extension.
+    """
+
+    def _artifact_with_raw_file(self, name):
+        from website.models.artifact import Artifact
+        artifact = MagicMock(spec=Artifact)
+        artifact.raw_file = MagicMock() if name else None
+        if name:
+            artifact.raw_file.name = name
+        artifact.RAW_FILE_LABELS = Artifact.RAW_FILE_LABELS
+        return artifact
+
+    def _label(self, name):
+        from website.models.artifact import Artifact
+        return Artifact.raw_file_label.fget(self._artifact_with_raw_file(name))
+
+    def test_pptx_label(self):
+        self.assertEqual(self._label("talks/Doe2020Title.pptx"), "PPTX")
+
+    def test_keynote_label(self):
+        self.assertEqual(self._label("talks/Doe2020Title.key"), "Keynote")
+
+    def test_ai_label(self):
+        self.assertEqual(self._label("posters/Doe2020Title.ai"), "AI")
+
+    def test_figma_label(self):
+        self.assertEqual(self._label("talks/Doe2020Title.fig"), "Figma")
+
+    def test_extension_case_insensitive(self):
+        self.assertEqual(self._label("talks/Doe2020Title.PPTX"), "PPTX")
+        self.assertEqual(self._label("talks/Doe2020Title.Key"), "Keynote")
+
+    def test_unknown_extension_falls_back_to_uppercased_ext(self):
+        self.assertEqual(self._label("talks/Doe2020Title.odp"), "ODP")
+
+    def test_no_raw_file_returns_none(self):
+        self.assertIsNone(self._label(None))
+
+    def test_no_extension_returns_none(self):
+        self.assertIsNone(self._label("talks/Doe2020Title"))
diff --git a/website/tests/test_bio_utils.py b/website/tests/test_bio_utils.py
new file mode 100644
index 00000000..13807eff
--- /dev/null
+++ b/website/tests/test_bio_utils.py
@@ -0,0 +1,531 @@
+"""
+Tests for website.utils.bio_utils.auto_generate_bio and its helpers.
+
+Tests use mocks rather than real Person/Position records so they don't have to
+navigate Person.save() side-effects (random image assignment, file I/O); the
+unit under test is pure presentation logic.
+"""
+
+from datetime import date, timedelta
+from unittest.mock import MagicMock, patch
+
+from django.test import SimpleTestCase
+
+from website.utils.bio_utils import (
+    _join_with_oxford_comma,
+    auto_generate_bio,
+    humanize_duration,
+)
+
+
+# --- Mock fixtures ---------------------------------------------------------
+
+
+class FakeQuerySet:
+    """Minimal stand-in for a Django QuerySet with the methods auto-bio uses."""
+
+    def __init__(self, items):
+        self._items = list(items)
+
+    def count(self):
+        return len(self._items)
+
+    def __iter__(self):
+        return iter(self._items)
+
+    def __getitem__(self, key):
+        return self._items[key]
+
+
+def _make_person(
+    *,
+    first_name="Jon",
+    last_name="Doe",
+    url_name="jondoe",
+    has_started=True,
+    is_current_member=False,
+    is_alumni_member=False,
+    is_current_collaborator=False,
+    is_past_collaborator=False,
+    is_active=False,
+    current_title="PhD Student",
+    latest_position=None,
+    publications=0,
+    projects=None,
+    mentors=None,
+    mentees=None,
+    total_time_as_member=None,
+):
+    """Build a mock Person exposing only the attributes auto_generate_bio reads."""
+    # If the caller indicates the person has started, default to a non-None
+    # latest_position so the "no positions at all" early-return doesn't
+    # short-circuit the role-branch tests. Callers exercising the no-position
+    # case pass has_started=False and leave latest_position as None.
+    if latest_position is None and has_started:
+        latest_position = _make_position(
+            start_date=date(2020, 1, 1), title=current_title
+        )
+
+    person = MagicMock(name=f"Person({first_name} {last_name})")
+    person.first_name = first_name
+    person.last_name = last_name
+    person.get_full_name.return_value = f"{first_name} {last_name}"
+    person.get_url_name.return_value = url_name
+    person.has_started = has_started
+    person.is_current_member = is_current_member
+    person.is_alumni_member = is_alumni_member
+    person.is_current_collaborator = is_current_collaborator
+    person.is_past_collaborator = is_past_collaborator
+    person.is_active = is_active
+    person.get_current_title = current_title
+    person.get_latest_position = latest_position
+    person.get_total_time_as_member = total_time_as_member
+    person.get_projects = projects or []
+
+    publication_set = MagicMock()
+    publication_set.exists.return_value = publications > 0
+    publication_set.count.return_value = publications
+    person.publication_set = publication_set
+
+    person.get_grad_mentors.return_value = FakeQuerySet(mentors or [])
+    person.get_mentees.return_value = FakeQuerySet(mentees or [])
+    return person
+
+
+def _make_link_person(first_name, last_name, url_name):
+    p = MagicMock()
+    p.get_full_name.return_value = f"{first_name} {last_name}"
+    p.get_url_name.return_value = url_name
+    return p
+
+
+def _make_project(name, short_name):
+    p = MagicMock()
+    p.name = name
+    p.short_name = short_name
+    return p
+
+
+def _make_position(start_date=None, end_date=None, title="PhD Student"):
+    pos = MagicMock()
+    pos.start_date = start_date
+    pos.end_date = end_date
+    pos.title = title
+    return pos
+
+
+# --- Role-sentence reachability matrix -------------------------------------
+
+
+class RoleSentenceTests(SimpleTestCase):
+    """One test per branch of _role_sentence."""
+
+    def test_no_position_no_publication_returns_empty(self):
+        """
+        Regression for #1258: a Person with no Position records AND no
+        publications must produce an empty bio, not "will be joining…".
+        """
+        person = _make_person(has_started=False, publications=0)
+        self.assertEqual(auto_generate_bio(person), "")
+
+    def test_no_position_with_publication_says_has_published(self):
+        person = _make_person(has_started=False, publications=3)
+        bio = auto_generate_bio(person)
+        self.assertIn("Jon Doe has published with the Makeability Lab.", bio)
+        # Contributions sentence follows.
+        self.assertIn("3 publications", bio)
+
+    def test_future_member_says_will_be_joining_with_pretty_date(self):
+        future_pos = _make_position(start_date=date(2026, 9, 15))
+        person = _make_person(has_started=False, latest_position=future_pos)
+        bio = auto_generate_bio(person)
+        self.assertEqual(
+            bio,
+            "Jon Doe will be joining the Makeability Lab on Sep 2026.",
+        )
+
+    def test_current_member_with_duration(self):
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            total_time_as_member=timedelta(days=int(365 * 3 + 180)),
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn(
+            "Jon Doe is currently a PhD Student in the Makeability Lab.", bio
+        )
+        self.assertIn("Jon has been in the lab for", bio)
+        self.assertIn("years.", bio)
+
+    def test_current_member_without_duration_omits_duration_sentence(self):
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            total_time_as_member=None,
+        )
+        bio = auto_generate_bio(person)
+        self.assertEqual(
+            bio, "Jon Doe is currently a PhD Student in the Makeability Lab."
+        )
+
+    def test_current_member_ms_uses_an_article(self):
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="MS Student",
+            total_time_as_member=None,
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("is currently an MS Student", bio)
+
+    @patch("website.utils.bio_utils._get_earliest_member_position")
+    @patch("website.utils.bio_utils._get_latest_member_position")
+    def test_alumni_member_branch(self, mock_latest, mock_earliest):
+        pos = _make_position(
+            start_date=date(2018, 9, 1),
+            end_date=date(2024, 3, 15),
+            title="PhD Student",
+        )
+        mock_latest.return_value = pos
+        mock_earliest.return_value = pos
+        person = _make_person(
+            is_alumni_member=True,
+            current_title="PhD Student",
+            total_time_as_member=timedelta(days=int(365 * 5.5)),
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("Jon Doe was a PhD Student in the Makeability Lab", bio)
+        self.assertIn("(Sep 2018 to Mar 2024).", bio)
+        # Duration appears between the title and the date range.
+        self.assertIn("years (Sep 2018 to Mar 2024).", bio)
+
+    @patch("website.utils.bio_utils._get_earliest_member_position")
+    @patch("website.utils.bio_utils._get_latest_member_position")
+    def test_alumni_member_now_collaborator_gets_two_sentences(
+        self, mock_latest, mock_earliest
+    ):
+        """
+        Pre-fix bug: a former member who later became a current collaborator
+        was described as "was a Collaborator … (… to present)". Now the
+        first sentence anchors on the latest MEMBER position, and the
+        current collaborator status is a separate trailing sentence.
+        """
+        pos = _make_position(
+            start_date=date(2018, 9, 1),
+            end_date=date(2024, 3, 15),
+            title="PhD Student",
+        )
+        mock_latest.return_value = pos
+        mock_earliest.return_value = pos
+        person = _make_person(
+            is_alumni_member=True,
+            is_current_collaborator=True,
+            current_title="Collaborator",
+            total_time_as_member=timedelta(days=int(365 * 5.5)),
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn(
+            "Jon Doe was a PhD Student in the Makeability Lab", bio
+        )
+        self.assertIn("(Sep 2018 to Mar 2024).", bio)
+        self.assertIn(
+            "Jon is currently a collaborator with the Makeability Lab.", bio
+        )
+        # The first sentence must NOT say "Collaborator" (the title attribute
+        # of get_latest_position would say that — bug we're guarding against).
+        self.assertNotIn("was a Collaborator", bio)
+        self.assertNotIn("to present", bio)
+
+    def test_current_collaborator(self):
+        person = _make_person(
+            is_current_collaborator=True, is_active=True
+        )
+        bio = auto_generate_bio(person)
+        self.assertEqual(
+            bio, "Jon Doe is a collaborator with the Makeability Lab."
+        )
+
+    def test_past_collaborator(self):
+        person = _make_person(is_past_collaborator=True)
+        bio = auto_generate_bio(person)
+        self.assertEqual(
+            bio, "Jon Doe was a collaborator with the Makeability Lab."
+        )
+
+
+# --- Contributions sentence ------------------------------------------------
+
+
+class ContributionsSentenceTests(SimpleTestCase):
+    """Shape of the projects/publications sentence."""
+
+    def test_pubs_only_singular(self):
+        person = _make_person(has_started=False, publications=1)
+        bio = auto_generate_bio(person)
+        self.assertIn("They contributed to 1 publication.", bio)
+
+    def test_pubs_only_plural(self):
+        person = _make_person(has_started=False, publications=5)
+        bio = auto_generate_bio(person)
+        self.assertIn("They contributed to 5 publications.", bio)
+
+    def test_one_project_no_pubs(self):
+        proj = _make_project("Sound Watch", "soundwatch")
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            projects=[proj],
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("They contributed to a project called <a href=", bio)
+        self.assertIn(">Sound Watch</a>.", bio)
+
+    def test_two_projects_uses_and(self):
+        projs = [
+            _make_project("Alpha", "alpha"),
+            _make_project("Beta", "beta"),
+        ]
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            projects=projs,
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("They contributed to 2 projects:", bio)
+        self.assertIn(">Alpha</a> and <a", bio)
+
+    def test_three_projects_uses_oxford_comma(self):
+        projs = [
+            _make_project("Alpha", "alpha"),
+            _make_project("Beta", "beta"),
+            _make_project("Gamma", "gamma"),
+        ]
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            projects=projs,
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("3 projects:", bio)
+        # Oxford comma + "and" before the last item.
+        self.assertIn(">Beta</a>, and <a", bio)
+
+    def test_four_or_more_projects_uses_including(self):
+        projs = [_make_project(f"P{c}", f"p{c}") for c in "abcdef"]
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            projects=projs,
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn(f"{len(projs)} projects, including", bio)
+
+    def test_project_and_pubs_joined_with_as_well_as(self):
+        proj = _make_project("Alpha", "alpha")
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            projects=[proj],
+            publications=4,
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("as well as 4 publications.", bio)
+
+
+# --- Mentor / mentee sentences --------------------------------------------
+
+
+class MentorMenteeSentenceTests(SimpleTestCase):
+    def test_no_mentors_no_mentees_no_sentence(self):
+        person = _make_person(
+            is_current_member=True, is_active=True, current_title="PhD Student"
+        )
+        bio = auto_generate_bio(person)
+        self.assertNotIn("mentored", bio)
+
+    def test_mentor_sentence_active_uses_is(self):
+        m = _make_link_person("Alice", "Smith", "alicesmith")
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            mentors=[m],
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("Jon is mentored by", bio)
+        self.assertIn(">Alice Smith</a>.", bio)
+
+    @patch("website.utils.bio_utils._get_earliest_member_position")
+    @patch("website.utils.bio_utils._get_latest_member_position")
+    def test_mentor_sentence_inactive_uses_was(self, mock_latest, mock_earliest):
+        pos = _make_position(
+            start_date=date(2018, 9, 1),
+            end_date=date(2024, 3, 15),
+            title="PhD Student",
+        )
+        mock_latest.return_value = pos
+        mock_earliest.return_value = pos
+        m = _make_link_person("Alice", "Smith", "alicesmith")
+        person = _make_person(
+            is_alumni_member=True,
+            is_active=False,
+            current_title="PhD Student",
+            mentors=[m],
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("Jon was mentored by", bio)
+
+    def test_mentor_sentence_multiple_uses_oxford_comma(self):
+        mentors = [
+            _make_link_person("Alice", "Smith", "alicesmith"),
+            _make_link_person("Bob", "Jones", "bobjones"),
+            _make_link_person("Carol", "Lee", "carollee"),
+        ]
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            mentors=mentors,
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("Jon is mentored by", bio)
+        self.assertIn(", and ", bio)
+
+    def test_mentee_member_uses_during_their_time_intro(self):
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            mentees=[_make_link_person("Mia", "X", "miax")],
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn(
+            "During their time in the lab, Jon mentored 1 Makeability Lab student",
+            bio,
+        )
+
+    def test_mentee_collaborator_uses_has_mentored_intro(self):
+        person = _make_person(
+            is_current_collaborator=True,
+            is_active=True,
+            mentees=[_make_link_person("Mia", "X", "miax")],
+        )
+        bio = auto_generate_bio(person)
+        self.assertNotIn("During their time in the lab", bio)
+        self.assertIn("Jon has mentored 1 Makeability Lab student", bio)
+
+    def test_mentee_count_three_uses_colon(self):
+        mentees = [
+            _make_link_person("Mia", "X", "miax"),
+            _make_link_person("Noah", "Y", "noahy"),
+            _make_link_person("Olivia", "Z", "oliviaz"),
+        ]
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            mentees=mentees,
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("mentored 3 Makeability Lab students:", bio)
+
+    def test_mentee_count_more_than_three_uses_including(self):
+        mentees = [
+            _make_link_person(f"M{i}", "X", f"m{i}x") for i in range(10)
+        ]
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            mentees=mentees,
+        )
+        bio = auto_generate_bio(person)
+        self.assertIn("mentored 10 Makeability Lab students, including", bio)
+
+
+# --- HTML escaping ---------------------------------------------------------
+
+
+class HtmlEscapingTests(SimpleTestCase):
+    """Defensive: free-text spliced into anchor tags must be escaped."""
+
+    def test_person_name_with_angle_brackets_is_escaped(self):
+        person = _make_person(
+            first_name="<script>",
+            last_name="Doe",
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+        )
+        bio = auto_generate_bio(person)
+        self.assertNotIn("<script>", bio)
+        self.assertIn("&lt;script&gt;", bio)
+
+    def test_project_name_with_html_is_escaped(self):
+        person = _make_person(
+            is_current_member=True,
+            is_active=True,
+            current_title="PhD Student",
+            projects=[_make_project("Bad <Project>", "badproject")],
+        )
+        bio = auto_generate_bio(person)
+        self.assertNotIn("<Project>", bio)
+        self.assertIn("Bad &lt;Project&gt;", bio)
+
+
+# --- Helpers ---------------------------------------------------------------
+
+
+class JoinWithOxfordCommaTests(SimpleTestCase):
+    def test_empty(self):
+        self.assertEqual(_join_with_oxford_comma([]), "")
+
+    def test_one(self):
+        self.assertEqual(_join_with_oxford_comma(["A"]), "A")
+
+    def test_two(self):
+        self.assertEqual(_join_with_oxford_comma(["A", "B"]), "A and B")
+
+    def test_three_uses_oxford_comma(self):
+        self.assertEqual(
+            _join_with_oxford_comma(["A", "B", "C"]), "A, B, and C"
+        )
+
+    def test_four_uses_oxford_comma(self):
+        self.assertEqual(
+            _join_with_oxford_comma(["A", "B", "C", "D"]), "A, B, C, and D"
+        )
+
+
+class HumanizeDurationTests(SimpleTestCase):
+    def test_under_a_month(self):
+        self.assertEqual(
+            humanize_duration(timedelta(days=10)), "less than a month"
+        )
+
+    def test_one_month(self):
+        self.assertEqual(humanize_duration(timedelta(days=30)), "1 month")
+
+    def test_six_months(self):
+        self.assertEqual(humanize_duration(timedelta(days=180)), "6 months")
+
+    def test_one_and_a_half_years(self):
+        self.assertEqual(
+            humanize_duration(timedelta(days=int(365 * 1.5))), "1.5 years"
+        )
+
+    def test_returns_years_string_for_long_durations(self):
+        # Exact float value depends on 30-day-month approximation; we only
+        # assert the format here.
+        result = humanize_duration(timedelta(days=365 * 5))
+        self.assertTrue(
+            result.endswith(" years"),
+            msg=f"Expected a years string, got {result!r}",
+        )
diff --git a/website/tests/test_data_health.py b/website/tests/test_data_health.py
new file mode 100644
index 00000000..afcece2d
--- /dev/null
+++ b/website/tests/test_data_health.py
@@ -0,0 +1,249 @@
+"""
+Tests for the Data Health admin dashboard + checks (issue #1276).
+
+Covers the shared name/image helpers, superuser gating on the dashboard /
+detail / export views, the CSV export, each individual check's row logic, and
+the read-only guarantee (checks must never mutate the DB).
+"""
+
+import os
+import shutil
+import tempfile
+
+from django.contrib.auth.models import User
+from django.test import SimpleTestCase, override_settings
+from django.urls import reverse
+
+from website.admin.data_health.registry import get_check
+from website.tests.base import DatabaseTestCase
+from website.utils.name_utils import normalize_person_name, is_default_person_image
+
+
+class NameUtilsTests(SimpleTestCase):
+    """Unit tests for the shared name/image helpers (no DB)."""
+
+    def test_normalize_folds_accents_and_strips_nonalpha(self):
+        self.assertEqual(normalize_person_name("Jon", "Froehlich"), "jonfroehlich")
+        self.assertEqual(normalize_person_name("Renée", "O'Brien"), "reneeobrien")
+        self.assertEqual(normalize_person_name("Jon-Paul", "Smith Jr."), "jonpaulsmithjr")
+
+    def test_same_name_different_case_same_key(self):
+        self.assertEqual(
+            normalize_person_name("Jane", "Doe"),
+            normalize_person_name("jane", "doe"),
+        )
+
+    def test_default_image_detection(self):
+        self.assertTrue(is_default_person_image(None))
+
+        class _Empty:
+            name = ""
+
+        class _StarWars:
+            name = "person/jane_doe_easteregg_starwars_rebel.png"
+
+        class _Real:
+            name = "person/jane_doe.gif"
+
+        self.assertTrue(is_default_person_image(_Empty()))
+        self.assertTrue(is_default_person_image(_StarWars()))
+        self.assertFalse(is_default_person_image(_Real()))
+
+
+class DataHealthAuthTests(DatabaseTestCase):
+    """Superuser gating on the dashboard, detail, and export views."""
+
+    def setUp(self):
+        self.superuser = User.objects.create_superuser("root", "r@example.com", "pw")
+        self.staff = User.objects.create_user(
+            "staffer", "s@example.com", "pw", is_staff=True
+        )
+
+    def test_superuser_can_view_dashboard(self):
+        self.client.force_login(self.superuser)
+        resp = self.client.get(reverse("admin:data_health_dashboard"))
+        self.assertEqual(resp.status_code, 200)
+
+    def test_staff_nonsuperuser_forbidden(self):
+        self.client.force_login(self.staff)
+        for name, args in [
+            ("admin:data_health_dashboard", []),
+            ("admin:data_health_detail", ["duplicate-people"]),
+            ("admin:data_health_export", ["duplicate-people"]),
+        ]:
+            resp = self.client.get(reverse(name, args=args))
+            self.assertEqual(resp.status_code, 403, msg=name)
+
+    def test_anonymous_redirected_to_login(self):
+        resp = self.client.get(reverse("admin:data_health_dashboard"))
+        self.assertEqual(resp.status_code, 302)
+        self.assertIn("/admin/login/", resp["Location"])
+
+    def test_unknown_check_returns_404(self):
+        self.client.force_login(self.superuser)
+        resp = self.client.get(reverse("admin:data_health_detail", args=["nope"]))
+        self.assertEqual(resp.status_code, 404)
+
+
+class DataHealthCsvTests(DatabaseTestCase):
+    """The CSV export streams text/csv as an attachment with the right rows."""
+
+    def setUp(self):
+        self.superuser = User.objects.create_superuser("root", "r@example.com", "pw")
+        self.client.force_login(self.superuser)
+
+    def test_csv_headers_and_content(self):
+        self.make_person(first_name="Jane", last_name="Doe", email="jane@example.com")
+        self.make_person(first_name="Jane", last_name="Doe", email="jane2@example.com")
+        resp = self.client.get(
+            reverse("admin:data_health_export", args=["duplicate-people"])
+        )
+        self.assertEqual(resp.status_code, 200)
+        self.assertEqual(resp["Content-Type"], "text/csv")
+        self.assertIn(
+            'attachment; filename="duplicate-people-', resp["Content-Disposition"]
+        )
+        body = resp.content.decode()
+        self.assertIn("cluster_key", body)  # header row
+        self.assertIn("jane@example.com", body)
+        self.assertIn("jane2@example.com", body)
+
+
+class DuplicatePeopleCheckTests(DatabaseTestCase):
+    def test_clusters_only_multi_person_same_name(self):
+        self.make_person(first_name="Jane", last_name="Doe")
+        self.make_person(first_name="Jane", last_name="Doe")
+        self.make_person(first_name="John", last_name="Smith")  # unique
+        rows = get_check("duplicate-people").get_rows()
+        self.assertEqual({r["cluster_key"] for r in rows}, {"janedoe"})
+        self.assertEqual(len(rows), 2)
+
+    def test_total_refs_counts_relations(self):
+        p1 = self.make_person(first_name="Jane", last_name="Doe")
+        p2 = self.make_person(first_name="Jane", last_name="Doe")
+        pub = self.make_publication(title="A Counted Paper")
+        pub.authors.add(p1)
+        rows = {r["id"]: r for r in get_check("duplicate-people").get_rows()}
+        self.assertEqual(rows[p1.pk]["pub_count"], 1)
+        self.assertEqual(rows[p1.pk]["total_refs"], 1)
+        self.assertEqual(rows[p2.pk]["total_refs"], 0)  # safe-to-delete shell
+
+    def test_clean_data_no_rows(self):
+        self.make_person(first_name="Solo", last_name="Person")
+        self.assertEqual(get_check("duplicate-people").get_rows(), [])
+
+
+class UrlNameCollisionsCheckTests(DatabaseTestCase):
+    def test_detects_forced_collision_and_placeholder(self):
+        from website.models import Person
+
+        self.make_person(first_name="Jane", last_name="Doe")
+        p2 = self.make_person(first_name="Jane", last_name="Doe")  # -> janedoe2
+        p3 = self.make_person(first_name="Solo", last_name="One")
+        # save() auto-dedupes url_name, so force a historical collision directly.
+        Person.objects.filter(pk=p2.pk).update(url_name="janedoe")
+        Person.objects.filter(pk=p3.pk).update(url_name="placeholder")
+
+        rows = {r["url_name"]: r for r in get_check("url-name-collisions").get_rows()}
+        self.assertIn("janedoe", rows)
+        self.assertEqual(rows["janedoe"]["count"], 2)
+        self.assertIn("placeholder", rows)
+
+
+class PublicationQualityCheckTests(DatabaseTestCase):
+    def test_duplicate_titles_and_missing_venue(self):
+        self.make_publication(title="Dup Paper")
+        self.make_publication(title="Dup Paper")  # duplicate normalized title
+        self.make_publication(title="No Venue Paper", forum_name=None)
+        rows = {r["id"]: r for r in get_check("publication-quality").get_rows()}
+        self.assertEqual(len([r for r in rows.values() if r["dup_title"]]), 2)
+        self.assertTrue(
+            any("forum_name" in r["missing_fields"] for r in rows.values())
+        )
+
+
+class ProjectHealthCheckTests(DatabaseTestCase):
+    def test_flags_incomplete_project(self):
+        from website.models import Project
+
+        Project.objects.create(name="Lonely Project", short_name="lonely")
+        rows = {r["name"]: r for r in get_check("project-health").get_rows()}
+        self.assertIn("Lonely Project", rows)
+        self.assertIn("no thumbnail", rows["Lonely Project"]["issues"])
+        self.assertIn("no publication", rows["Lonely Project"]["issues"])
+
+
+class PositionIntegrityCheckTests(DatabaseTestCase):
+    def test_no_position_and_self_advisor(self):
+        from datetime import date as _date
+
+        from website.models import Position
+        from website.models.position import Title
+
+        p_nopos = self.make_person(first_name="No", last_name="Position")
+        p_self = self.make_person(first_name="Self", last_name="Advisor")
+        Position.objects.create(
+            person=p_self,
+            start_date=_date(2020, 1, 1),
+            title=Title.PHD_STUDENT,
+            advisor=p_self,
+        )
+        issues = {
+            (r["person_id"], r["issue"])
+            for r in get_check("position-integrity").get_rows()
+        }
+        self.assertIn((p_nopos.pk, "no position"), issues)
+        self.assertIn((p_self.pk, "self-advisor"), issues)
+
+
+class NewsHealthCheckTests(DatabaseTestCase):
+    def test_flags_missing_slug_and_author(self):
+        from website.models import News
+
+        author = self.make_person(first_name="News", last_name="Writer")
+        n_noauthor = self.make_news_item(title="Orphan News", author=None)
+        n_noslug = self.make_news_item(title="Slugless News", author=author)
+        News.objects.filter(pk=n_noslug.pk).update(slug="")  # save() auto-slugs
+
+        rows = {r["id"]: r for r in get_check("news-health").get_rows()}
+        self.assertIn(n_noauthor.pk, rows)
+        self.assertFalse(rows[n_noauthor.pk]["has_author"])
+        self.assertIn(n_noslug.pk, rows)
+        self.assertTrue(rows[n_noslug.pk]["missing_slug"])
+
+
+class MediaIntegrityCheckTests(DatabaseTestCase):
+    def setUp(self):
+        # Use a throwaway MEDIA_ROOT so the test never pollutes dev media.
+        self._media_dir = tempfile.mkdtemp(prefix="dh_media_")
+        self._override = override_settings(MEDIA_ROOT=self._media_dir)
+        self._override.enable()
+        self.addCleanup(self._override.disable)
+        self.addCleanup(shutil.rmtree, self._media_dir, ignore_errors=True)
+
+    def test_flags_missing_file(self):
+        pub = self.make_publication(title="Vanishing Paper")
+        path = pub.pdf_file.path
+        if os.path.exists(path):
+            os.remove(path)  # simulate a file that disappeared from disk
+        hits = [
+            r
+            for r in get_check("media-integrity").get_rows()
+            if r["type"] == "Publication"
+            and r["id"] == pub.pk
+            and r["status"] == "missing-file"
+        ]
+        self.assertTrue(hits)
+
+
+class DataHealthReadOnlyTests(DatabaseTestCase):
+    def test_get_rows_does_not_mutate_db(self):
+        from website.models import Person, Publication
+
+        self.make_person(first_name="Jane", last_name="Doe")
+        self.make_person(first_name="Jane", last_name="Doe")
+        before = (Person.objects.count(), Publication.objects.count())
+        for slug in ("duplicate-people", "url-name-collisions", "position-integrity"):
+            get_check(slug).get_rows()
+        after = (Person.objects.count(), Publication.objects.count())
+        self.assertEqual(before, after)
diff --git a/website/tests/test_project.py b/website/tests/test_project.py
new file mode 100644
index 00000000..91e25c95
--- /dev/null
+++ b/website/tests/test_project.py
@@ -0,0 +1,103 @@
+"""Tests for Project model methods (member count, time-worked annotation)."""
+
+from datetime import date, timedelta
+from unittest.mock import MagicMock
+
+from django.test import SimpleTestCase
+
+from website.tests.base import DatabaseTestCase
+
+
+# --- Project member count regression --------------------------------------
+
+
+class ProjectCurrentMemberCountTests(SimpleTestCase):
+    """
+    Regression for Project.get_current_member_count: the method built a
+    queryset but never returned it, so the admin column always displayed
+    None. These tests pin both the return value and the filter semantics
+    so a future caller can't quietly re-introduce the bug.
+    """
+
+    def _make_project_with_count(self, count):
+        project = MagicMock()
+        project.projectrole_set.filter.return_value\
+            .values.return_value\
+            .distinct.return_value\
+            .count.return_value = count
+        return project
+
+    def test_returns_count_not_none(self):
+        from website.models.project import Project
+        project = self._make_project_with_count(7)
+        self.assertEqual(Project.get_current_member_count(project), 7)
+
+    def test_filters_on_open_ended_project_role(self):
+        from website.models.project import Project
+        project = self._make_project_with_count(0)
+        Project.get_current_member_count(project)
+        project.projectrole_set.filter.assert_called_with(end_date__isnull=True)
+
+
+# --- Project.get_people time-worked annotation ----------------------------
+
+
+class ProjectGetPeopleTimeWorkedTests(DatabaseTestCase):
+    """
+    Regression for the dead conditional in Project.get_people
+    (models/project.py).
+
+    The original time_worked expression was
+        F('end_date') if F('end_date') is not None else timezone.now()
+    but an F() object is never None, so the else branch was dead: ongoing
+    roles (end_date=None) produced NULL durations and a NULL
+    total_time_worked annotation. The fix coalesces end_date to today in
+    the DB via Coalesce().
+    """
+
+    def _make_project(self, name="Test Project", short_name="testproj"):
+        from website.models import Project
+        return Project.objects.create(name=name, short_name=short_name)
+
+    def _add_role(self, person, project, start, end=None):
+        from website.models import ProjectRole
+        return ProjectRole.objects.create(
+            person=person, project=project, start_date=start, end_date=end
+        )
+
+    def test_ongoing_role_has_non_null_time_worked(self):
+        project = self._make_project()
+        person = self.make_person(first_name="Ada", last_name="Lovelace")
+        self._add_role(
+            person, project, date.today() - timedelta(days=100), end=None
+        )
+
+        result = list(project.get_people())  # default sorted_by="time_on_project"
+        self.assertEqual(len(result), 1)
+        time_worked = result[0].total_time_worked
+        self.assertIsNotNone(time_worked)  # was None before the Coalesce fix
+        self.assertGreaterEqual(time_worked, timedelta(days=99))
+
+    def test_long_completed_role_outranks_short_ongoing_role(self):
+        """
+        Discriminating ordering check. With the bug, the ongoing role's
+        NULL duration sorts NULLS-FIRST under Postgres DESC ordering, so
+        the short ongoing role would wrongly rank first. With the fix the
+        long completed role ranks first by actual duration.
+        """
+        project = self._make_project()
+        long_done = self.make_person(first_name="Long", last_name="Veteran")
+        short_ongoing = self.make_person(first_name="Short", last_name="Newcomer")
+        self._add_role(
+            long_done, project,
+            date.today() - timedelta(days=400),
+            end=date.today() - timedelta(days=35),
+        )  # ~365 days
+        self._add_role(
+            short_ongoing, project,
+            date.today() - timedelta(days=10), end=None,
+        )  # ~10 days
+
+        result = list(project.get_people())
+        self.assertEqual(result[0], long_done)
+        self.assertEqual(result[1], short_ongoing)
diff --git a/website/tests/test_publication.py b/website/tests/test_publication.py
new file mode 100644
index 00000000..4c473668
--- /dev/null
+++ b/website/tests/test_publication.py
@@ -0,0 +1,211 @@
+"""Tests for Publication model methods (BibTeX, forum name, author lookup)."""
+
+from unittest.mock import MagicMock
+
+from django.test import SimpleTestCase
+
+from website.tests.base import DatabaseTestCase
+
+
+# --- BibTeX citation regression -------------------------------------------
+
+
+class BibtexCitationTests(SimpleTestCase):
+    """
+    Regression tests for Publication.get_citation_as_bibtex.
+
+    The method previously compared self.pub_venue_type with ``is`` against
+    PubType.JOURNAL / PubType.ARTICLE. Because TextChoices values are strings
+    loaded from the DB, the identity check was always False and every journal
+    or article paper was emitted as @inproceedings{ instead of @article{.
+    These tests pin the corrected behavior.
+    """
+
+    def _make_publication(self, pub_venue_type):
+        """Mock Publication exposing only what get_citation_as_bibtex reads."""
+        pub = MagicMock()
+        pub.pub_venue_type = pub_venue_type
+        pub.get_bibtex_id.return_value = "Doe2020FooCHI20,"
+        pub.authors.all.return_value = []
+        pub.title = "A Test Title"
+        pub.book_title = "Proceedings of Test"
+        pub.get_formatted_forum_name.return_value = "CHI"
+        pub.date.year = 2020
+        # Falsy values below match the "if self.X" guards in the method:
+        # when falsy the optional bibtex fields are skipped without
+        # exercising any further attribute access.
+        pub.series = ""
+        pub.isbn = ""
+        pub.location = ""
+        pub.page_num_start = None
+        pub.page_num_end = None
+        pub.num_pages = None
+        pub.doi = ""
+        pub.official_url = ""
+        pub.acmid = ""
+        pub.publisher = ""
+        return pub
+
+    def test_journal_uses_article_entry(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.JOURNAL)
+        bibtex = Publication.get_citation_as_bibtex(pub)
+        self.assertTrue(
+            bibtex.startswith("@article{"),
+            msg=f"Expected JOURNAL to emit @article{{, got {bibtex[:60]!r}",
+        )
+
+    def test_article_uses_article_entry(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.ARTICLE)
+        bibtex = Publication.get_citation_as_bibtex(pub)
+        self.assertTrue(
+            bibtex.startswith("@article{"),
+            msg=f"Expected ARTICLE to emit @article{{, got {bibtex[:60]!r}",
+        )
+
+    def test_conference_uses_inproceedings_entry(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.CONFERENCE)
+        bibtex = Publication.get_citation_as_bibtex(pub)
+        self.assertTrue(
+            bibtex.startswith("@inproceedings{"),
+            msg=f"Expected CONFERENCE to emit @inproceedings{{, got {bibtex[:60]!r}",
+        )
+
+
+# --- Formatted forum name regression (#988) -------------------------------
+
+
+class FormattedForumNameTests(SimpleTestCase):
+    """
+    Regression tests for Publication.get_formatted_forum_name (#988).
+
+    The previous implementation bundled posters, demos, work-in-progress,
+    and doctoral consortium papers all under one generic
+    "Extended Abstract Proceedings of …" label, and didn't prefix workshop
+    papers at all. Per the issue, each short-form category should get a
+    specific label ("Poster Proceedings of …", "Demo Proceedings of …",
+    "Workshop Proceedings of …", "Work-in-Progress Proceedings of …",
+    "Doctoral Consortium Proceedings of …"). The bare `extended_abstract`
+    boolean field remains the catch-all fallback when the venue type
+    doesn't match a known short-form category.
+    """
+
+    def _make_publication(self, pub_venue_type, *, forum_name="CHI", year=2024,
+                          extended_abstract=False):
+        pub = MagicMock()
+        pub.forum_name = forum_name
+        pub.pub_venue_type = pub_venue_type
+        pub.extended_abstract = extended_abstract
+        pub.date.year = year
+        return pub
+
+    def test_conference_uses_proceedings_of(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.CONFERENCE)
+        self.assertEqual(
+            Publication.get_formatted_forum_name(pub),
+            "Proceedings of CHI 2024",
+        )
+
+    def test_poster_uses_poster_proceedings_of(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.POSTER)
+        self.assertEqual(
+            Publication.get_formatted_forum_name(pub),
+            "Poster Proceedings of CHI 2024",
+        )
+
+    def test_demo_uses_demo_proceedings_of(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.DEMO)
+        self.assertEqual(
+            Publication.get_formatted_forum_name(pub),
+            "Demo Proceedings of CHI 2024",
+        )
+
+    def test_workshop_uses_workshop_proceedings_of(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.WORKSHOP)
+        self.assertEqual(
+            Publication.get_formatted_forum_name(pub),
+            "Workshop Proceedings of CHI 2024",
+        )
+
+    def test_wip_uses_work_in_progress_proceedings_of(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.WIP)
+        self.assertEqual(
+            Publication.get_formatted_forum_name(pub),
+            "Work-in-Progress Proceedings of CHI 2024",
+        )
+
+    def test_doctoral_consortium_uses_doctoral_consortium_proceedings_of(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.DOCTORAL_CONSORTIUM)
+        self.assertEqual(
+            Publication.get_formatted_forum_name(pub),
+            "Doctoral Consortium Proceedings of CHI 2024",
+        )
+
+    def test_journal_has_no_prefix(self):
+        """Journals are not 'Proceedings of …'; the forum name stands alone."""
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.JOURNAL, forum_name="TOCHI")
+        self.assertEqual(
+            Publication.get_formatted_forum_name(pub),
+            "TOCHI 2024",
+        )
+
+    def test_article_has_no_prefix(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.ARTICLE, forum_name="ArXiv")
+        self.assertEqual(
+            Publication.get_formatted_forum_name(pub),
+            "ArXiv 2024",
+        )
+
+    def test_extended_abstract_flag_is_fallback(self):
+        """
+        The `extended_abstract` BooleanField stays as the catch-all when a
+        pub doesn't fit a specific short-form category (e.g. a panel paper
+        marked as a short submission).
+        """
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.PANEL, extended_abstract=True)
+        self.assertEqual(
+            Publication.get_formatted_forum_name(pub),
+            "Extended Abstract Proceedings of CHI 2024",
+        )
+
+    def test_empty_forum_name_returns_empty_string(self):
+        from website.models.publication import Publication, PubType
+        pub = self._make_publication(PubType.CONFERENCE, forum_name="")
+        self.assertEqual(Publication.get_formatted_forum_name(pub), "")
+
+
+# --- get_person regression (authorless publications) ----------------------
+
+
+class PublicationGetPersonTests(DatabaseTestCase):
+    """
+    Regression for Publication.get_person (models/publication.py).
+
+    The method indexed self.authors.all()[0], raising IndexError on an
+    authorless publication (e.g. a PhD dissertation entered before its
+    author is linked). This path is reached from views/people.py, which
+    iterates dissertations and calls get_person() on each. The fix returns
+    None via .first() so the caller can skip authorless rows.
+    """
+
+    def test_no_authors_returns_none(self):
+        pub = self.make_publication(title="Authorless Paper")
+        self.assertIsNone(pub.get_person())
+
+    def test_returns_first_author_in_sorted_order(self):
+        pub = self.make_publication(title="Authored Paper")
+        first = self.make_person(first_name="First", last_name="Author")
+        second = self.make_person(first_name="Second", last_name="Author")
+        pub.authors.add(first, second)  # SortedManyToManyField preserves order
+        self.assertEqual(pub.get_person(), first)
diff --git a/website/tests/test_serve_pdf.py b/website/tests/test_serve_pdf.py
new file mode 100644
index 00000000..a7c2f990
--- /dev/null
+++ b/website/tests/test_serve_pdf.py
@@ -0,0 +1,89 @@
+"""Tests for website.views.serve_pdf."""
+
+from unittest.mock import MagicMock, patch
+
+from django.test import SimpleTestCase
+
+
+class ServePdfTests(SimpleTestCase):
+    """
+    Regression tests for website.views.serve_pdf.
+
+    The previous implementation used
+        Publication.objects.get(pdf_file__icontains=filename)
+    which had two failure modes:
+
+    1. Only ``ObjectDoesNotExist`` was caught. When a substring matched
+       multiple publications' stored paths, ``MultipleObjectsReturned``
+       was raised and propagated as a 500 — e.g. requesting "Speech"
+       when both "Froehlich2010Speech.pdf" and
+       "Froehlich2010Speechalytics.pdf" exist.
+    2. ``__icontains`` is a substring match (SQL ``LIKE %x%``), so a
+       probe for ".pdf" matched every PDF in the database. Combined with
+       the fuzzy difflib fallback that redirects on miss, this enabled
+       enumeration of stored filenames.
+
+    The fix uses ``filter(pdf_file__iendswith=filename).first()`` for
+    the exact branch: ``.first()`` can't raise ``MultipleObjectsReturned``,
+    and ``__iendswith`` only matches paths ending in the requested name.
+    The fuzzy difflib fallback is retained for academic-link integrity
+    and is only invoked on actual miss.
+    """
+
+    def test_exact_match_returns_pdf_response(self):
+        """Happy path: an exact filename match returns the PDF inline."""
+        from website.views.serve_pdf import serve_pdf
+        fake_pub = MagicMock()
+        fake_pub.pdf_file.read.return_value = b"%PDF-1.4 fake"
+        fake_pub.pdf_file.name = "publications/Froehlich2018Speech.pdf"
+        with patch("website.views.serve_pdf.Publication") as MockPub:
+            MockPub.objects.filter.return_value.first.return_value = fake_pub
+            response = serve_pdf(MagicMock(), "Froehlich2018Speech.pdf")
+        self.assertEqual(response.status_code, 200)
+        self.assertEqual(response["Content-Type"], "application/pdf")
+
+    def test_uses_iendswith_not_icontains(self):
+        """
+        Regression for the substring-probe enumeration bug. The exact
+        branch must use ``__iendswith``; a request for ".pdf" must
+        therefore miss (since no pub's path equals ".pdf") rather than
+        matching every pub via substring.
+        """
+        from django.http import Http404
+        from website.views.serve_pdf import serve_pdf
+        with patch("website.views.serve_pdf.Publication") as MockPub, patch(
+            "website.views.serve_pdf.get_closest_filename_from_database",
+            return_value=None,
+        ):
+            MockPub.objects.filter.return_value.first.return_value = None
+            with self.assertRaises(Http404):
+                serve_pdf(MagicMock(), ".pdf")
+            MockPub.objects.filter.assert_called_with(pdf_file__iendswith=".pdf")
+
+    def test_no_exact_match_uses_fuzzy_redirect(self):
+        """
+        Stale-external-link preservation. When the exact branch misses,
+        the difflib fallback should redirect to the closest filename it
+        finds. This is the academic-link-integrity feature.
+        """
+        from website.views.serve_pdf import serve_pdf
+        with patch("website.views.serve_pdf.Publication") as MockPub, patch(
+            "website.views.serve_pdf.get_closest_filename_from_database",
+            return_value="publications/Froehlich2018Updated.pdf",
+        ):
+            MockPub.objects.filter.return_value.first.return_value = None
+            response = serve_pdf(MagicMock(), "Froehlich2018Old.pdf")
+        self.assertEqual(response.status_code, 302)
+        self.assertEqual(response.url, "/media/publications/Froehlich2018Updated.pdf")
+
+    def test_no_exact_no_fuzzy_returns_404(self):
+        """If neither the exact branch nor the fuzzy fallback finds anything, 404."""
+        from django.http import Http404
+        from website.views.serve_pdf import serve_pdf
+        with patch("website.views.serve_pdf.Publication") as MockPub, patch(
+            "website.views.serve_pdf.get_closest_filename_from_database",
+            return_value=None,
+        ):
+            MockPub.objects.filter.return_value.first.return_value = None
+            with self.assertRaises(Http404):
+                serve_pdf(MagicMock(), "NoSuchPaper.pdf")
diff --git a/website/tests/test_video.py b/website/tests/test_video.py
new file mode 100644
index 00000000..32b07b9c
--- /dev/null
+++ b/website/tests/test_video.py
@@ -0,0 +1,33 @@
+"""Tests for Video model methods."""
+
+from datetime import date, timedelta
+
+from django.test import SimpleTestCase
+
+
+class VideoAgeTests(SimpleTestCase):
+    """
+    Regression tests for Video.get_age_in_ms (relates to #1091).
+
+    The method computed `datetime.now().date() - self.date`, which raised
+    TypeError when the nullable `date` field was None. It now returns None
+    for a dateless video so video-age.js falls back to the static date.
+    Pure unit tests: Video instances are constructed but never saved.
+    """
+
+    def test_none_date_returns_none(self):
+        from website.models import Video
+        self.assertIsNone(Video(date=None).get_age_in_ms())
+
+    def test_past_date_returns_positive_int(self):
+        from website.models import Video
+        age = Video(date=date.today() - timedelta(days=7)).get_age_in_ms()
+        self.assertIsInstance(age, int)
+        # 7 days in ms, allowing slack for the now() call crossing midnight.
+        self.assertGreaterEqual(age, 6 * 24 * 60 * 60 * 1000)
+
+    def test_future_date_returns_negative_int(self):
+        from website.models import Video
+        age = Video(date=date.today() + timedelta(days=7)).get_age_in_ms()
+        self.assertIsInstance(age, int)
+        self.assertLess(age, 0)
diff --git a/website/tests/test_views.py b/website/tests/test_views.py
new file mode 100644
index 00000000..ec007ba8
--- /dev/null
+++ b/website/tests/test_views.py
@@ -0,0 +1,167 @@
+"""
+View / template / queryset regression tests.
+
+These exercise the URL → view → template stack end to end (or render snippets
+directly), catching the class of bugs that pure unit tests can't — template
+crashes, NoReverseMatch, and N+1 query regressions.
+"""
+
+from django.core.files.uploadedfile import SimpleUploadedFile
+from django.db import connection
+from django.test.utils import CaptureQueriesContext
+from django.urls import reverse
+
+from website.tests.base import DatabaseTestCase
+
+
+# --- View-level: null author on /news/<id>/ (regression for #1013) --------
+
+
+class NewsItemNullAuthorViewTests(DatabaseTestCase):
+    """
+    Regression for #1013 — a News item with author=None used to crash the
+    news_item view with AttributeError on cur_news_item.author.authored_news.
+    Fixed in 1c0d6c0 by guarding the access. This test pins the behavior
+    so it can't regress silently.
+    """
+
+    def test_news_item_with_null_author_renders_200(self):
+        item = self.make_news_item(title="Authorless News", author=None)
+        response = self.client.get(
+            reverse("website:news_item_by_id", kwargs={"id": item.id})
+        )
+        self.assertEqual(response.status_code, 200)
+        self.assertContains(response, "Authorless News")
+
+    def test_news_item_with_author_still_renders_200(self):
+        """Sanity check: the non-null path also works."""
+        author = self.make_person(first_name="Ada", last_name="Lovelace")
+        item = self.make_news_item(
+            title="Authored News", author=author
+        )
+        response = self.client.get(
+            reverse("website:news_item_by_id", kwargs={"id": item.id})
+        )
+        self.assertEqual(response.status_code, 200)
+        self.assertContains(response, "Authored News")
+
+
+# --- Talk snippet: external_slides_url rendering (#1273) -----------------
+
+
+class TalkExternalSlidesUrlTests(DatabaseTestCase):
+    """
+    Pins the rendering of Talk.external_slides_url in
+    snippets/display_talk_snippet.html — the "Source" link should appear
+    only when the URL is set. Also pins that Poster persists the field
+    (no display surface for posters yet, but the column must exist).
+    """
+
+    def _render_talk_snippet(self, talk):
+        from django.template.loader import render_to_string
+        return render_to_string(
+            "snippets/display_talk_snippet.html",
+            {"talk": talk, "MEDIA_URL": "/media/"},
+        )
+
+    def test_source_link_renders_when_external_slides_url_set(self):
+        talk = self.make_talk(
+            title="Figma Talk",
+            external_slides_url="https://www.figma.com/file/abc123/slides",
+        )
+        html = self._render_talk_snippet(talk)
+        self.assertIn("https://www.figma.com/file/abc123/slides", html)
+        self.assertIn("fa-up-right-from-square", html)
+        # opens-in-new-tab affordance must be present for accessibility
+        self.assertIn('target="_blank"', html)
+        self.assertIn('rel="noopener"', html)
+
+    def test_source_link_absent_when_external_slides_url_blank(self):
+        talk = self.make_talk(title="No-Source Talk")
+        html = self._render_talk_snippet(talk)
+        self.assertNotIn("fa-up-right-from-square", html)
+
+    def test_poster_external_slides_url_round_trips(self):
+        """Schema pin: Poster.external_slides_url must persist to the DB."""
+        from website.models import Poster
+        poster = Poster.objects.create(
+            title="A Test Poster",
+            external_slides_url="https://www.figma.com/file/xyz/poster",
+            pdf_file=SimpleUploadedFile(
+                "test_poster.pdf",
+                b"%PDF-1.4 test",
+                content_type="application/pdf",
+            ),
+        )
+        reloaded = Poster.objects.get(pk=poster.pk)
+        self.assertEqual(
+            reloaded.external_slides_url,
+            "https://www.figma.com/file/xyz/poster",
+        )
+
+
+# --- Query-count: /publications/ prefetch_related (regression for d4f6d65) -
+
+
+class PublicationsViewQueryCountTests(DatabaseTestCase):
+    """
+    Pins the prefetch_related batch on /publications/ (d4f6d65).
+
+    Before the fix, the snippet template iterated pub.authors.all and
+    pub.projects.all per publication, producing 617 queries on prod with
+    ~250 pubs. The fix added .prefetch_related('authors', 'projects',
+    'keywords') to the view's queryset, dropping that to 60.
+
+    The key correctness property is that the query count is bounded by a
+    constant — it must NOT grow with the number of publications. This
+    test creates publications with M2M relations and asserts the count
+    stays under a generous ceiling for two different data sizes. If a
+    future contributor removes the prefetches, the count climbs linearly
+    and the larger-N test fails.
+    """
+
+    # Generous ceiling well above the steady-state count we measured
+    # locally (~15 queries for the publications view). The ceiling exists
+    # to catch order-of-magnitude regressions, not to pin an exact count.
+    QUERY_CEILING = 30
+
+    def _seed_publications(self, count):
+        author = self.make_person(first_name="Ada", last_name="Lovelace")
+        for i in range(count):
+            pub = self.make_publication(
+                title=f"Paper {i}", year=2024
+            )
+            pub.authors.add(author)
+
+    def _capture_publications_query_count(self):
+        with CaptureQueriesContext(connection) as ctx:
+            response = self.client.get(reverse("website:publications"))
+        return response, len(ctx.captured_queries)
+
+    def test_query_count_is_bounded_with_few_pubs(self):
+        self._seed_publications(2)
+        response, count = self._capture_publications_query_count()
+        self.assertEqual(response.status_code, 200)
+        self.assertLessEqual(
+            count,
+            self.QUERY_CEILING,
+            msg=f"Query count {count} exceeded ceiling {self.QUERY_CEILING}",
+        )
+
+    def test_query_count_does_not_grow_with_pub_count(self):
+        """
+        The real regression guard: query count must be roughly the same
+        whether we render 2 pubs or 20. If prefetches are removed, the
+        count grows by a multiple of N and this test will fail.
+        """
+        self._seed_publications(20)
+        response, count = self._capture_publications_query_count()
+        self.assertEqual(response.status_code, 200)
+        self.assertLessEqual(
+            count,
+            self.QUERY_CEILING,
+            msg=(
+                f"Query count {count} exceeded ceiling {self.QUERY_CEILING} "
+                "with 20 publications — prefetch_related likely regressed"
+            ),
+        )