feat(homeassistant): waiting updates, backup age, and updates installed after a confirmed backup (#6355)

A new plugin reads Home Assistant with a read-only token every 5 minutes and
shows which Core, OS and add-on updates are waiting and how old the last backup
is. Install takes a backup through a local-only webhook, confirms Home
Assistant's last-backup time moved, installs, and follows the update to its end,
through Home Assistant's and Steward's own restarts. One runs at a time. Each
ends as one timeline entry.

Steward holds nothing that can act on the house: the webhook's automation can
only take a backup or install an allowed update (System 231's ruling). A refused
token stops polling, as Home Assistant bans an address after a few refusals.

Milestone 576 steps 1 to 5.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/integration/test_homeassistant.py

@@ -0,0 +1,290 @@
+"""Integration: Steward reads Home Assistant, and installs one update through its
+webhook only after a newer backup, following it to its end. A fake Home Assistant
+answers over httpx's mock transport. Requires STEWARD_DATABASE_URL."""
+from __future__ import annotations
+
+import asyncio
+import importlib
+import json
+import os
+import uuid
+
+import httpx
+import pytest
+from sqlalchemy import delete, select
+
+pytestmark = pytest.mark.integration
+
+ADDON = "update.example_update"
+BEFORE, AFTER = "2026-10-30T03:00:00+00:00", "2026-10-31T03:00:00+00:00"
+
+
+@pytest.fixture
+def app():
+    if not os.environ.get("STEWARD_DATABASE_URL"):
+        pytest.skip("needs Postgres")
+    from steward.app import create_app
+    return create_app(testing=False)
+
+
+def _ha(sub):
+    from steward.core.plugin_manager import get_loaded_plugin_modules
+    loaded = get_loaded_plugin_modules().get("homeassistant")
+    base = loaded.__name__ if loaded is not None else "plugins.homeassistant"
+    return importlib.import_module(f"{base}.{sub}")
+
+
+class FakeHomeAssistant:
+    """Its entities, the webhook calls it received, and what it refuses."""
+
+    def __init__(self):
+        self.calls = []
+        self.tokens = set()
+        self.refuse = False
+        self.down = False
+        self.hook_runs = True
+        self.entities = {
+            ADDON: self._update("on", "1.0"),
+            "sensor.backup_last_successful_automatic_backup": {"state": BEFORE, "attributes": {}},
+            "event.backup_automatic_backup": {"state": BEFORE, "attributes": {"event_type": "completed"}},
+            "automation.steward_maintenance": {"state": "on", "attributes": {"id": "steward_maintenance",
+                                                                             "last_triggered": None}},
+        }
+
+    def _update(self, state, installed):
+        return {"state": state, "attributes": {"title": "Example", "installed_version": installed,
+                                               "latest_version": "1.1", "in_progress": False,
+                                               "entity_picture": "/api/hassio/addons/example/icon"}}
+
+    def backup_lands(self):
+        self.entities["sensor.backup_last_successful_automatic_backup"]["state"] = AFTER
+
+    def update_lands(self):
+        self.entities[ADDON] = self._update("off", "1.1")
+
+    def handler(self, request: httpx.Request) -> httpx.Response:
+        if self.down:
+            raise httpx.ConnectError("connection refused")
+        path = request.url.path
+        if path.startswith("/api/webhook/"):
+            assert "authorization" not in request.headers
+            self.calls.append(json.loads(request.content))
+            if self.hook_runs:
+                self.entities["automation.steward_maintenance"]["attributes"]["last_triggered"] = AFTER
+            return httpx.Response(200)
+        self.tokens.add(request.headers.get("authorization"))
+        if self.refuse:
+            return httpx.Response(401)
+        if path == "/api/states":
+            return httpx.Response(200, json=[{"entity_id": k, **v} for k, v in self.entities.items()])
+        entity = self.entities.get(path.removeprefix("/api/states/"))
+        return httpx.Response(200, json=entity) if entity else httpx.Response(404)
+
+    @property
+    def transport(self):
+        return httpx.MockTransport(self.handler)
+
+
+async def _clean(app, **connection):
+    from steward.core.crypto import encrypt_secret
+    m = _ha("models")
+    async with app.db_sessionmaker() as s:
+        async with s.begin():
+            await s.execute(delete(m.HomeAssistantUpdate))
+            await s.execute(delete(m.HomeAssistantConnection))
+            s.add(m.HomeAssistantConnection(url="http://ha.test:8123", token=encrypt_secret("read-only"),
+                                            webhook_id=encrypt_secret(f"hook{uuid.uuid4().hex}"), **connection))
+
+
+async def _state(app):
+    m = _ha("models")
+    async with app.db_sessionmaker() as s:
+        connection = (await s.execute(select(m.HomeAssistantConnection))).scalars().first()
+        runs = (await s.execute(select(m.HomeAssistantUpdate))).scalars().all()
+    return connection, runs
+
+
+def test_a_poll_stores_what_it_read_and_records_its_metrics(app):
+    from steward.models.metrics import PluginMetricLatest
+    sched = _ha("scheduler")
+    ha = FakeHomeAssistant()
+
+    async def go():
+        await _clean(app)
+        await sched.poll(app, transport=ha.transport)
+        connection, _ = await _state(app)
+        async with app.db_sessionmaker() as s:
+            rows = (await s.execute(select(PluginMetricLatest).where(
+                PluginMetricLatest.source_module == "homeassistant"))).scalars().all()
+        return connection, {r.metric_name: r.value for r in rows}
+
+    connection, metrics = asyncio.run(go())
+    assert ha.tokens == {"Bearer read-only"}
+    assert connection.automation == "on" and connection.last_error is None and not connection.refused
+    assert [u["entity_id"] for u in connection.updates] == [ADDON]
+    assert connection.backup["last_ok"] == BEFORE
+    assert metrics["is_up"] == 1.0 and metrics["updates_waiting"] == 1.0 and metrics["backup_age_hours"] >= 0
+
+
+def test_a_refused_token_is_asked_once_and_not_again(app):
+    sched = _ha("scheduler")
+    ha = FakeHomeAssistant()
+    ha.refuse = True
+    seen = []
+    real = ha.handler
+
+    def counted(request):
+        seen.append(request.url.path)
+        return real(request)
+    transport = httpx.MockTransport(counted)
+
+    async def go():
+        await _clean(app)
+        await sched.poll(app, transport=transport)
+        await sched.poll(app, transport=transport)
+        return (await _state(app))[0]
+
+    connection = asyncio.run(go())
+    assert seen == ["/api/states"]
+    assert connection.refused and "refused the token" in connection.last_error
+
+
+def test_an_update_waits_for_a_newer_backup_then_installs_and_ends_on_the_timeline(app):
+    from steward.models.timeline import TimelineEntry
+    sched, upd = _ha("scheduler"), _ha("updates")
+    ha = FakeHomeAssistant()
+
+    async def go():
+        await _clean(app)
+        await sched.poll(app, transport=ha.transport)
+        run = await upd.start_update(app, ADDON, "someone", transport=ha.transport)
+        assert ha.calls == [{"action": "backup"}]
+        assert run.state == "backing_up" and run.from_version == "1.0" and run.to_version == "1.1"
+
+        # No newer backup yet: nothing is installed, however often it is asked.
+        await upd.advance(app, transport=ha.transport)
+        await upd.advance(app, transport=ha.transport)
+        assert ha.calls == [{"action": "backup"}]
+
+        ha.backup_lands()
+        await upd.advance(app, transport=ha.transport)
+        _, [row] = await _state(app)
+        assert row.state == "installing" and row.backup_at is not None
+        await upd.advance(app, transport=ha.transport)
+        assert ha.calls == [{"action": "backup"}, {"action": "install", "entity_id": ADDON}]
+
+        # Home Assistant restarting is expected, and isn't the end of the update.
+        ha.down = True
+        await upd.advance(app, transport=ha.transport)
+        _, [row] = await _state(app)
+        assert row.state == "installing"
+        ha.down = False
+
+        ha.update_lands()
+        await upd.advance(app, transport=ha.transport)
+        connection, [row] = await _state(app)
+        assert row.state == "done" and row.finished_at is not None
+        assert connection.last_polled_at is None        # the page shows the new version at the next tick
+        await upd.advance(app, transport=ha.transport)
+        assert len(ha.calls) == 2
+        async with app.db_sessionmaker() as s:
+            entries = (await s.execute(select(TimelineEntry).where(
+                TimelineEntry.ref_type == "homeassistant_update", TimelineEntry.ref_id == row.id))).scalars().all()
+        return entries
+
+    [entry] = asyncio.run(go())
+    assert entry.type == "homeassistant.updated" and entry.actor == "someone"
+    assert "1.0 to 1.1" in entry.summary and entry.area == "service"
+
+
+def test_a_failed_backup_installs_nothing(app):
+    sched, upd = _ha("scheduler"), _ha("updates")
+    ha = FakeHomeAssistant()
+
+    async def go():
+        await _clean(app)
+        await sched.poll(app, transport=ha.transport)
+        await upd.start_update(app, ADDON, "someone", transport=ha.transport)
+        ha.entities["event.backup_automatic_backup"] = {
+            "state": AFTER, "attributes": {"event_type": "failed", "failed_reason": "upload_failed"}}
+        await upd.advance(app, transport=ha.transport)
+        return (await _state(app))[1]
+
+    [run] = asyncio.run(go())
+    assert run.state == "failed" and "upload_failed" in run.message and "Nothing was installed" in run.message
+    assert ha.calls == [{"action": "backup"}]
+
+
+def test_a_webhook_that_never_ran_the_automation_says_so(app):
+    from datetime import timedelta
+    from sqlalchemy import update
+    sched, upd, m = _ha("scheduler"), _ha("updates"), _ha("models")
+    ha = FakeHomeAssistant()
+    ha.hook_runs = False
+
+    async def go():
+        await _clean(app)
+        await sched.poll(app, transport=ha.transport)
+        run = await upd.start_update(app, ADDON, "someone", transport=ha.transport)
+        await upd.advance(app, transport=ha.transport)
+        assert (await _state(app))[1][0].state == "backing_up"      # within its grace
+        async with app.db_sessionmaker() as s:
+            async with s.begin():
+                await s.execute(update(m.HomeAssistantUpdate).where(m.HomeAssistantUpdate.id == run.id).values(
+                    started_at=run.started_at - timedelta(seconds=upd.TRIGGER_GRACE_S + 1)))
+        await upd.advance(app, transport=ha.transport)
+        return (await _state(app))[1]
+
+    [run] = asyncio.run(go())
+    assert run.state == "failed" and "webhook id" in run.message and "Nothing was installed" in run.message
+
+
+def test_an_install_past_its_deadline_fails_and_only_one_runs_at_a_time(app):
+    from datetime import timedelta
+    from sqlalchemy import update
+    sched, upd, m = _ha("scheduler"), _ha("updates"), _ha("models")
+    ha = FakeHomeAssistant()
+
+    async def go():
+        await _clean(app)
+        await sched.poll(app, transport=ha.transport)
+        run = await upd.start_update(app, ADDON, "someone", transport=ha.transport)
+        with pytest.raises(upd.CannotStart, match="under way"):
+            await upd.start_update(app, ADDON, "someone else", transport=ha.transport)
+        ha.backup_lands()
+        await upd.advance(app, transport=ha.transport)
+        await upd.advance(app, transport=ha.transport)
+        async with app.db_sessionmaker() as s:
+            async with s.begin():
+                row = await s.get(m.HomeAssistantUpdate, run.id)
+                row.install_sent_at -= timedelta(seconds=upd.INSTALL_DEADLINE_S["addon"] + 1)
+        await upd.advance(app, transport=ha.transport)
+        return (await _state(app))[1]
+
+    [run] = asyncio.run(go())
+    assert run.state == "failed" and "Still not on 1.1" in run.message
+    assert len(ha.calls) == 2
+
+
+@pytest.mark.parametrize("entity,change,says", [
+    ("update.unknown", None, "isn't a Core, OS or add-on"),
+    (ADDON, "no_longer_waiting", "no longer waiting"),
+    (ADDON, "no_automation", "automation isn't on"),
+])
+def test_nothing_starts_unless_the_update_can_run(app, entity, change, says):
+    sched, upd = _ha("scheduler"), _ha("updates")
+    ha = FakeHomeAssistant()
+
+    async def go():
+        await _clean(app)
+        if change == "no_automation":
+            del ha.entities["automation.steward_maintenance"]
+        await sched.poll(app, transport=ha.transport)
+        if change == "no_longer_waiting":
+            ha.update_lands()
+        with pytest.raises(upd.CannotStart, match=says):
+            await upd.start_update(app, entity, "someone", transport=ha.transport)
+        return (await _state(app))[1]
+
+    assert asyncio.run(go()) == []
+    assert ha.calls == []