fix(alerts): an alert on a count says how many of what, and names them (#6400)

"swarm services short on swarm is 6." becomes "6 swarm services are
short of their replicas.", with a Which line naming each service and how
many of its tasks run. The swarm figures get labels, and the counting
figures (swarm, mounts, drives, updates) read as sentences in the email,
its subject and the Alerts page. A plugin names what its count counts
through core.alerts.names_behind; the Docker plugin does for the three
swarm figures, from the stored swarm. Webhooks get alert.which too.

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

steward/alerts/service.py

@@ -163,7 +163,8 @@ def alert_sentence(rule: AlertRule, label: str, value: float | None,
     "tail"}, read lead + resource + tail, e.g. "CPU on " + "web-1" +
     " is 94% (> 90%), firing for 12m". label is alert_label(state); since is
     when that label began (no_data_since for no_data)."""
-    from steward.core.formatting import human_duration, metric_label, metric_value
+    from steward.core.formatting import (WHOLE_RESOURCES, count_words, human_duration, metric_label,
+                                         metric_value)
     metric = rule.metric_name
     said = metric_label(metric)
     held = human_duration((now - since).total_seconds()) if since else None
@@ -174,6 +175,15 @@ def alert_sentence(rule: AlertRule, label: str, value: float | None,
     # not how long ago someone acknowledged it (the page says that apart).
     word = "breaching" if label == "acknowledged" else label
     for_how_long = f", {word} for {held}" if held else f", {word}"
+    condition = f" ({rule.operator.value} {metric_value(metric, rule.threshold)})"
+    counted = count_words(metric, value)
+    if counted and rule.resource_name in WHOLE_RESOURCES:
+        # The count is the whole of it, and links where the swarm is shown.
+        return {"lead": "", "resource": counted[:1].upper() + counted[1:],
+                "tail": f"{condition}{for_how_long}"}
+    if counted:
+        return {"lead": f"{counted[:1].upper() + counted[1:]} on ", "resource": rule.resource_name,
+                "tail": f"{condition}{for_how_long}"}
     if value is None:
         reading = " has no reading yet"
     elif metric in _BINARY:
@@ -181,7 +191,6 @@ def alert_sentence(rule: AlertRule, label: str, value: float | None,
                 "tail": f" is {metric_value(metric, value)}{for_how_long}"}
     else:
         reading = f" is {metric_value(metric, value)}"
-    condition = f" ({rule.operator.value} {metric_value(metric, rule.threshold)})"
     return {"lead": f"{said} on ", "resource": rule.resource_name,
             "tail": f"{reading}{condition}{for_how_long}"}
 
@@ -787,11 +796,14 @@ async def test_alert(db, rule: AlertRule) -> dict:
         target = (await db.execute(
             select(AlertRule).where(AlertRule.parent_id == rule.id)
             .order_by(AlertRule.resource_name).limit(1))).scalar_one_or_none() or rule
+    from steward.core.alerts import which
     return {"rule_name": rule.name, "state": "TEST", "metric": rule.metric_name,
             "value": await latest_value(db, target) if target.resource_name else None,
             "operator": rule.operator.value, "threshold": rule.threshold,
             "resource": target.resource_name or covers_text(rule),
             "source_module": rule.source_module,
+            "which": await which(db, rule.source_module, rule.metric_name, target.resource_name)
+            if target.resource_name else [],
             "note": "A test of this rule's notifications. Nothing has fired.",
             "timestamp": datetime.now(timezone.utc).isoformat()}