Merge pull request 'Ansible schedule form: structured playbook-variable fields + first-item dropdown defaults' (#2) from dev into main

The schedule and host run forms opened their Source/Playbook <select>s on a
"— choose … —" placeholder, forcing an extra click. Default them to the first
item and cascade the dependent fields so the initial state is real:

- Remove the placeholder options (schedules source+playbook, host source+
  playbook, and the shared _playbook_options fragment).
- schedules(): pass sel_source (first source, or the edited schedule's) and
  that source's pre-rendered playbooks so the playbook dropdown starts populated
  (create mode too, no longer the all-sources union).
- Cascade on load/change: the source <select> fires playbook-options then
  (hx-on::after-settle) re-triggers the playbook <select>, which loads that
  playbook's variable fields. Host form drives the whole chain from a load
  trigger; schedule create mode loads vars via the playbook <select>'s own load
  trigger, while edit mode keeps its server-rendered prefill untouched.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Dockerfile

@@ -19,8 +19,10 @@ WORKDIR /app
 
 COPY pyproject.toml .
 COPY steward/ steward/
-# .[ansible] pulls the full Ansible package so the playbook runner works in-image.
-RUN pip install --no-cache-dir '.[ansible]'
+# Bundle the extras whose first-party plugins ship in the image: [ansible] for
+# the playbook runner, [snmp] (pysnmp) for the SNMP poller. Without them those
+# bundled plugins load but silently no-op for want of their runtime dependency.
+RUN pip install --no-cache-dir '.[ansible,snmp]'
 
 COPY alembic.ini .
 # First-party plugins ship inside the image (bundled root at /app/plugins).

compose.deploy.yml

@@ -22,12 +22,22 @@ services:
       # /data holds the auto-generated secret key, third-party plugins, and the
       # Ansible playbook cache — all of which must survive image updates. No
       # source bind-mounts: core + first-party plugins live inside the image.
+      #
+      # The container runs as the unprivileged 'app' user (uid 1000), so /data
+      # MUST be writable by uid 1000. A named docker volume (below) gets that
+      # automatically. If you bind-mount a host/NFS path instead, chown it to
+      # 1000:1000 (and mind NFS root_squash) — otherwise the app can't persist
+      # its secret key, mints a fresh one each boot, and every encrypted secret
+      # (managed SSH key, SMTP/OIDC/LDAP creds) becomes unrecoverable. The app
+      # now refuses to start in that state rather than silently degrade.
       - app_data:/data
     environment:
       - STEWARD_DATABASE_URL=postgresql+asyncpg://steward:${POSTGRES_PASSWORD:-steward}@db/steward
       # STEWARD_SECRET_KEY is intentionally absent — the app generates a random
       # key on first boot and persists it to /data/secret.key, which the app_data
-      # volume keeps across image updates.
+      # volume keeps across image updates. On multi-node Swarm with a shared bind
+      # mount, pin STEWARD_SECRET_KEY (env or a docker secret) instead, so the key
+      # never depends on filesystem ownership being correct on every node.
     depends_on:
       db:
         condition: service_healthy

plugins/docker/__init__.py

@@ -47,3 +47,8 @@ def get_scheduled_tasks() -> list:
 def get_blueprint():
     from .routes import docker_bp
     return docker_bp
+
+
+def get_nav() -> list[dict]:
+    # Surfaces in the sidebar's Infrastructure group (fleet-wide container view).
+    return [{"label": "Docker", "href": "/plugins/docker/"}]

plugins/docker/dedup.py

@@ -0,0 +1,27 @@
+"""Model-free Docker view helpers.
+
+Kept separate from routes.py/models.py so tests can import it directly: importing
+a module that re-imports plugins.docker.models AFTER the app has loaded the docker
+plugin raises "Table 'docker_containers' is already defined". This module touches
+no ORM models, so it's safe to import in either environment.
+"""
+from __future__ import annotations
+
+
+def dedup_by_container_id(containers: list) -> list:
+    """Collapse containers double-reported across hosts. Swarm-aware agents on
+    every manager list cluster-wide tasks, so the same container (identical
+    container_id) arrives once per manager and would otherwise be counted N times.
+    Keep the first occurrence per non-empty container_id — callers order
+    running-first, so a running report wins over a stale stopped one. Rows without
+    a container_id (older agents) can't be matched, so they're kept as-is."""
+    seen: set[str] = set()
+    out = []
+    for c in containers:
+        cid = (c.container_id or "").strip()
+        if cid:
+            if cid in seen:
+                continue
+            seen.add(cid)
+        out.append(c)
+    return out

plugins/docker/ingest.py

@@ -321,3 +321,16 @@ async def persist_host_docker(session, host, snapshots, swarm=None, disk=None) -
                 resource_name=resource, metric_name="is_healthy",
                 value=1.0 if health == "healthy" else 0.0,
             )
+
+    # Reap containers that vanished from the host's latest listing. Without this,
+    # removed one-shot containers (CI job runners, buildkit builders, codex jobs)
+    # accumulate forever as permanent "stopped" rows and inflate every count.
+    # The listing is authoritative (event derivation already treats absence as
+    # removal), so anything not in the newest snapshot no longer exists on the
+    # host. Mirrors the image/swarm/disk persisters. An empty snapshot legitimately
+    # means the host has no containers, so the unfiltered delete is correct.
+    seen_names = {c["name"] for c in latest_containers if c.get("name")}
+    reap = delete(DockerContainer).where(DockerContainer.host_id == host.id)
+    if seen_names:
+        reap = reap.where(DockerContainer.name.notin_(seen_names))
+    await session.execute(reap)

plugins/docker/migrations/versions/docker_008_bigint_mem.py

@@ -0,0 +1,33 @@
+"""Widen memory byte columns to BIGINT
+
+docker_metrics.mem_usage_bytes and docker_containers.mem_usage_bytes /
+mem_limit_bytes were int4, which overflows for any container using >2.1 GB
+(asyncpg: "value out of int32 range"), failing the whole ingest batch. Widen to
+BIGINT — a safe in-place int4→int8 promotion, no data rewrite.
+
+Revision ID: docker_008_bigint_mem
+Revises: docker_007_disk_usage
+Create Date: 2026-06-20
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "docker_008_bigint_mem"
+down_revision: Union[str, None] = "docker_007_disk_usage"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.alter_column("docker_metrics", "mem_usage_bytes", type_=sa.BigInteger())
+    op.alter_column("docker_containers", "mem_usage_bytes", type_=sa.BigInteger())
+    op.alter_column("docker_containers", "mem_limit_bytes", type_=sa.BigInteger())
+
+
+def downgrade() -> None:
+    # Narrowing back to int4 will error if any stored value exceeds 2^31 — that's
+    # the very condition this migration fixed, so downgrade is best-effort.
+    op.alter_column("docker_containers", "mem_limit_bytes", type_=sa.Integer())
+    op.alter_column("docker_containers", "mem_usage_bytes", type_=sa.Integer())
+    op.alter_column("docker_metrics", "mem_usage_bytes", type_=sa.Integer())

plugins/docker/models.py

@@ -29,8 +29,10 @@ class DockerContainer(Base):
     status: Mapped[str] = mapped_column(String(32), nullable=False, default="unknown")
     # running | stopped | paused | exited | dead
     cpu_pct: Mapped[float | None] = mapped_column(Float, nullable=True)
-    mem_usage_bytes: Mapped[int | None] = mapped_column(Integer, nullable=True)
-    mem_limit_bytes: Mapped[int | None] = mapped_column(Integer, nullable=True)
+    # BigInteger: a container's memory usage/limit routinely exceeds 2^31 bytes
+    # (~2.1 GB), which overflows int4 and fails the insert.
+    mem_usage_bytes: Mapped[int | None] = mapped_column(BigInteger, nullable=True)
+    mem_limit_bytes: Mapped[int | None] = mapped_column(BigInteger, nullable=True)
     mem_pct: Mapped[float | None] = mapped_column(Float, nullable=True)
     restart_count: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
     ports_json: Mapped[str] = mapped_column(Text, nullable=False, default="[]")
@@ -78,7 +80,8 @@ class DockerMetric(Base):
     )
     cpu_pct: Mapped[float] = mapped_column(Float, nullable=False, default=0.0)
     mem_pct: Mapped[float] = mapped_column(Float, nullable=False, default=0.0)
-    mem_usage_bytes: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+    # BigInteger: per-container memory usage exceeds 2^31 bytes (~2.1 GB) routinely.
+    mem_usage_bytes: Mapped[int] = mapped_column(BigInteger, nullable=False, default=0)
 
     # Per-container history lookups filter on (host_id, container_name) then sort
     # by time — a composite index keeps the rows() sparkline queries cheap.

plugins/docker/routes.py

@@ -1,20 +1,39 @@
 # plugins/docker/routes.py
 from __future__ import annotations
 import json
-from datetime import datetime, timezone
+from datetime import datetime, timedelta, timezone
 
 from quart import Blueprint, current_app, render_template, request
-from sqlalchemy import select
+from sqlalchemy import func, select
 
 from steward.auth.middleware import require_role
 from steward.models.users import UserRole
 from steward.models.hosts import Host
 from steward.core.time_range import parse_range, DEFAULT_RANGE
-from .models import DockerContainer, DockerMetric
+from .dedup import dedup_by_container_id
+from .swarm_view import build_swarm_services
+from .models import (
+    DockerContainer, DockerDiskUsage, DockerEvent, DockerImage, DockerMetric,
+    DockerSwarmNode, DockerSwarmService,
+)
 
 docker_bp = Blueprint("docker", __name__, template_folder="templates")
 
 
+def _human_bytes(n: int | None) -> str:
+    """Compact binary size string (e.g. '1.4 GiB', '512 MiB', '0 B')."""
+    if n is None:
+        return "—"
+    size = float(n)
+    for unit in ("B", "KiB", "MiB", "GiB", "TiB"):
+        if abs(size) < 1024.0 or unit == "TiB":
+            if unit == "B":
+                return f"{int(size)} {unit}"
+            return f"{size:.1f} {unit}"
+        size /= 1024.0
+    return f"{size:.1f} TiB"
+
+
 def _human_uptime(started_at: datetime | None) -> str | None:
     """Compact 'how long running' string (e.g. '3d 4h', '5h 12m', '8m')."""
     if started_at is None:
@@ -92,10 +111,19 @@ async def _container_history(db, host_id: str, name: str, since) -> tuple[list,
 async def index():
     poll_interval = current_app.config.get("MONITORS_POLL_INTERVAL", 60)
     current_range = request.args.get("range", DEFAULT_RANGE)
+    # Show the Swarm link only when a manager has actually reported topology, so
+    # non-swarm installs aren't offered an always-empty page.
+    async with current_app.db_sessionmaker() as db:
+        has_swarm = (await db.execute(
+            select(DockerSwarmService.host_id).limit(1))).first() is not None
+        has_disk = (await db.execute(
+            select(DockerDiskUsage.host_id).limit(1))).first() is not None
     return await render_template(
         "docker/index.html",
         poll_interval=poll_interval,
         current_range=current_range,
+        has_swarm=has_swarm,
+        has_disk=has_disk,
     )
 
 
@@ -111,7 +139,8 @@ async def _host_map(db, host_ids: set[str]) -> dict[str, Host]:
 @docker_bp.get("/rows")
 @require_role(UserRole.viewer)
 async def rows():
-    """HTMX fragment: containers grouped by host, with resource sparklines."""
+    """HTMX fragment: swarm services (collapsed, cluster-complete) on top, then
+    non-swarm containers grouped by host with resource sparklines."""
     since, range_key = parse_range(request.args.get("range"))
 
     async with current_app.db_sessionmaker() as db:
@@ -121,12 +150,24 @@ async def rows():
                 DockerContainer.name,
             )
         )).scalars())
-        hosts = await _host_map(db, {c.host_id for c in containers})
+        services = list((await db.execute(select(DockerSwarmService))).scalars())
+        nodes = list((await db.execute(select(DockerSwarmNode))).scalars())
+        hosts = await _host_map(
+            db,
+            {c.host_id for c in containers}
+            | {n.host_id for n in nodes} | {s.host_id for s in services},
+        )
 
-        # Group by host so each container is clearly attributed to the box it
-        # runs on (names are only unique within a host now).
+        # Swarm tasks collapse into per-service rows (each replica labelled with
+        # its host, plus "ghost" replicas for tasks on agent-less nodes, drawn
+        # from the managers' placement). Non-swarm containers keep the host view.
+        node_hostname = {n.node_id: n.hostname for n in nodes if n.hostname}
+        host_name = {h.id: h.name for h in hosts.values()}
+        swarm_view = build_swarm_services(containers, services, node_hostname, host_name)
+
+        standalone = [c for c in containers if not c.service_name]
         groups: dict[str, dict] = {}
-        for c in containers:
+        for c in standalone:
             g = groups.get(c.host_id)
             if g is None:
                 host = hosts.get(c.host_id)
@@ -149,15 +190,34 @@ async def rows():
             else:
                 g["stopped"] += 1
 
+    # Sub-group each host's containers by compose project, preserving the
+    # running-first ordering. Containers with no project fall into an unlabelled
+    # bucket rendered flat, so plain hosts look unchanged.
+    for g in groups.values():
+        subs: dict[str, dict] = {}
+        order: list[str] = []
+        for item in g["containers"]:
+            c = item["container"]
+            label = c.compose_project or None
+            key = label or ""
+            if key not in subs:
+                subs[key] = {"label": label, "containers": []}
+                order.append(key)
+            subs[key]["containers"].append(item)
+        g["subgroups"] = [subs[k] for k in order]
+        g["grouped"] = any(s["label"] for s in g["subgroups"])
+
     host_groups = sorted(groups.values(), key=lambda g: g["host_name"].lower())
-    running = sum(g["running"] for g in host_groups)
-    total = len(containers)
+    swarm_services = swarm_view["services"]
+    standalone_running = sum(g["running"] for g in host_groups)
+    ghost_total = sum(r["count"] for s in swarm_services for r in s["replicas"] if r["ghost"])
     return await render_template(
         "docker/rows.html",
         host_groups=host_groups,
-        running=running,
-        stopped=total - running,
-        total=total,
+        swarm_services=swarm_services,
+        running=standalone_running + sum(s["running"] for s in swarm_services),
+        stopped=len(standalone) - standalone_running,
+        total=len(containers) + ghost_total,
         range_key=range_key,
     )
 
@@ -168,26 +228,46 @@ async def widget():
     """HTMX dashboard widget: container status overview, grouped by host."""
     show_stopped = request.args.get("show_stopped", "no") == "yes"
     widget_id = request.args.get("wid", "0")
+    since_24h = datetime.now(timezone.utc) - timedelta(hours=24)
 
     async with current_app.db_sessionmaker() as db:
-        all_containers = list((await db.execute(
+        all_containers = dedup_by_container_id(list((await db.execute(
             select(DockerContainer).order_by(
                 (DockerContainer.status == "running").desc(),
                 DockerContainer.name,
             )
-        )).scalars())
-        hosts = await _host_map(db, {c.host_id for c in all_containers})
+        )).scalars()))
+        services = list((await db.execute(select(DockerSwarmService))).scalars())
+        nodes = list((await db.execute(select(DockerSwarmNode))).scalars())
+        # Recent failures are more actionable than a static "stopped" tally. Count
+        # DISTINCT container names so the two managers' duplicate die/oom events
+        # don't double it.
+        failed_24h = (await db.execute(
+            select(func.count(func.distinct(DockerEvent.container_name)))
+            .where(DockerEvent.event.in_(("die", "oom")), DockerEvent.at >= since_24h)
+        )).scalar() or 0
+        hosts = await _host_map(
+            db,
+            {c.host_id for c in all_containers}
+            | {n.host_id for n in nodes} | {s.host_id for s in services},
+        )
 
-    running = [c for c in all_containers if c.status == "running"]
-    stopped = [c for c in all_containers if c.status != "running"]
-    display = all_containers if show_stopped else running
+    # Swarm tasks collapse into per-service rows (each replica tagged with its
+    # host); non-swarm containers stay grouped by host.
+    node_hostname = {n.node_id: n.hostname for n in nodes if n.hostname}
+    host_name = {h.id: h.name for h in hosts.values()}
+    swarm_services = build_swarm_services(all_containers, services, node_hostname, host_name)["services"]
+
+    standalone = [c for c in all_containers if not c.service_name]
+    running = [c for c in standalone if c.status == "running"]
+    display = standalone if show_stopped else running
 
     # Group for display so multi-host fleets read clearly; single-host stays flat.
     groups: dict[str, dict] = {}
     for c in display:
         g = groups.setdefault(c.host_id, {
             "host_id": c.host_id,
-            "host_name": hosts[c.host_id].name if c.host_id in hosts else c.host_id,
+            "host_name": host_name.get(c.host_id, c.host_id),
             "containers": [],
         })
         g["containers"].append(c)
@@ -197,8 +277,10 @@ async def widget():
         "docker/widget.html",
         host_groups=host_groups,
         multi_host=len(host_groups) > 1,
-        running_count=len(running),
-        stopped_count=len(stopped),
+        swarm_services=swarm_services,
+        running_count=len(running) + sum(s["running"] for s in swarm_services),
+        failed_24h=failed_24h,
+        total_count=len(all_containers) + len(swarm_services),
         show_stopped=show_stopped,
         widget_id=widget_id,
     )
@@ -212,11 +294,13 @@ async def widget_resources():
     widget_id = request.args.get("wid", "0")
 
     async with current_app.db_sessionmaker() as db:
-        containers = list((await db.execute(
+        # Dedup before the limit, else swarm tasks reported by both managers can
+        # fill the list with duplicates of the same container.
+        containers = dedup_by_container_id(list((await db.execute(
             select(DockerContainer)
             .where(DockerContainer.status == "running")
             .order_by(DockerContainer.cpu_pct.desc().nullslast())
-        )).scalars())[:limit]
+        )).scalars()))[:limit]
         hosts = await _host_map(db, {c.host_id for c in containers})
 
     rows_data = [
@@ -258,3 +342,224 @@ async def host_panel(host_id: str):
         running=running,
         stopped=len(containers) - running,
     )
+
+
+# Glyph + colour per lifecycle event, so the timeline reads at a glance.
+_EVENT_STYLE = {
+    "start":         ("▲", "var(--green)"),
+    "stop":          ("■", "var(--text-muted)"),
+    "die":           ("✕", "var(--red)"),
+    "oom":           ("☠", "var(--red)"),
+    "health_change": ("◐", "var(--orange)"),
+}
+
+
+@docker_bp.get("/container/<host_id>/<name>")
+@require_role(UserRole.viewer)
+async def container_detail(host_id: str, name: str):
+    """Full detail page for one container: facts, resource history, lifecycle."""
+    async with current_app.db_sessionmaker() as db:
+        c = await db.get(DockerContainer, (host_id, name))
+        host = await db.get(Host, host_id)
+        events = []
+        if c is not None:
+            events = list((await db.execute(
+                select(DockerEvent)
+                .where(DockerEvent.host_id == host_id)
+                .where(DockerEvent.container_name == name)
+                .order_by(DockerEvent.at.desc())
+                .limit(50)
+            )).scalars())
+
+    if c is None:
+        return await render_template(
+            "docker/container_detail.html",
+            container=None, host_id=host_id, name=name,
+        ), 404
+
+    return await render_template(
+        "docker/container_detail.html",
+        container=c,
+        host=host,
+        host_id=host_id,
+        name=name,
+        ports=json.loads(c.ports_json) if c.ports_json else [],
+        uptime=_human_uptime(c.started_at) if c.status == "running" else None,
+        net_rx=_human_bytes(c.net_rx_bytes), net_tx=_human_bytes(c.net_tx_bytes),
+        blk_read=_human_bytes(c.blk_read_bytes), blk_write=_human_bytes(c.blk_write_bytes),
+        events=[
+            {"event": e.event, "detail": e.detail, "at": e.at,
+             "glyph": _EVENT_STYLE.get(e.event, ("•", "var(--text-muted)"))[0],
+             "colour": _EVENT_STYLE.get(e.event, ("•", "var(--text-muted)"))[1]}
+            for e in events
+        ],
+        current_range=request.args.get("range", DEFAULT_RANGE),
+    )
+
+
+@docker_bp.get("/swarm")
+@require_role(UserRole.viewer)
+async def swarm():
+    """Swarm topology page: services with replica health, nodes, placement.
+
+    Swarm services/nodes are cluster-global — every manager's API returns the
+    same list — but each manager reports them independently (rows keyed by
+    host_id). So we group the reporting managers into swarms (managers that share
+    any node_id are the same cluster) and dedup within each: one row per service
+    (by name) and per node (by node_id), keeping the freshest. Without this, two
+    managers in one cluster list every service/node twice.
+    """
+    async with current_app.db_sessionmaker() as db:
+        services = list((await db.execute(
+            select(DockerSwarmService).order_by(DockerSwarmService.service_name)
+        )).scalars())
+        nodes = list((await db.execute(
+            select(DockerSwarmNode).order_by(
+                DockerSwarmNode.role, DockerSwarmNode.hostname)
+        )).scalars())
+        hosts = await _host_map(db, {s.host_id for s in services} | {n.host_id for n in nodes})
+
+    # ── Group reporting managers into swarms by shared node membership ──────────
+    manager_ids = {s.host_id for s in services} | {n.host_id for n in nodes}
+    nodes_by_mgr: dict[str, set[str]] = {}
+    for n in nodes:
+        nodes_by_mgr.setdefault(n.host_id, set()).add(n.node_id)
+
+    parent = {m: m for m in manager_ids}
+
+    def _find(x: str) -> str:
+        while parent[x] != x:
+            parent[x] = parent[parent[x]]
+            x = parent[x]
+        return x
+
+    mgrs = list(manager_ids)
+    for i in range(len(mgrs)):
+        for j in range(i + 1, len(mgrs)):
+            a, b = mgrs[i], mgrs[j]
+            if nodes_by_mgr.get(a) and nodes_by_mgr.get(b) and (nodes_by_mgr[a] & nodes_by_mgr[b]):
+                parent[_find(a)] = _find(b)
+
+    swarm_members: dict[str, set[str]] = {}
+    for m in manager_ids:
+        swarm_members.setdefault(_find(m), set()).add(m)
+
+    # ── Dedup services (by name) + nodes (by node_id) within each swarm ─────────
+    swarms = []
+    for members in swarm_members.values():
+        svc_by_name: dict[str, DockerSwarmService] = {}
+        for s in services:
+            if s.host_id in members and (
+                s.service_name not in svc_by_name
+                or s.updated_at > svc_by_name[s.service_name].updated_at
+            ):
+                svc_by_name[s.service_name] = s
+        node_by_id: dict[str, DockerSwarmNode] = {}
+        for n in nodes:
+            if n.host_id in members and (
+                n.node_id not in node_by_id
+                or n.updated_at > node_by_id[n.node_id].updated_at
+            ):
+                node_by_id[n.node_id] = n
+
+        node_names = {nid: (nrow.hostname or nid) for nid, nrow in node_by_id.items()}
+        svc_dicts = []
+        for s in sorted(svc_by_name.values(), key=lambda s: s.service_name):
+            placement = []
+            for p in (json.loads(s.placement_json) if s.placement_json else []):
+                nid = p.get("node_id", "")
+                placement.append({
+                    "node": node_names.get(nid, nid[:12] or "?"),
+                    "running": p.get("running", 0),
+                })
+            svc_dicts.append({
+                "name": s.service_name, "mode": s.mode,
+                "running": s.running, "desired": s.desired, "image": s.image,
+                "healthy": s.running >= s.desired and s.desired > 0,
+                "degraded": 0 < s.running < s.desired,
+                "down": s.running == 0 and s.desired > 0,
+                "placement": placement,
+            })
+        node_rows = sorted(
+            node_by_id.values(), key=lambda n: (n.role, (n.hostname or n.node_id).lower())
+        )
+        manager_names = sorted(hosts[m].name if m in hosts else m for m in members)
+        swarms.append({
+            "managers": manager_names,
+            "services": svc_dicts,
+            "nodes": node_rows,
+        })
+
+    swarms.sort(key=lambda g: g["managers"][0].lower() if g["managers"] else "")
+    return await render_template("docker/swarm.html", swarms=swarms)
+
+
+@docker_bp.get("/disk")
+@require_role(UserRole.viewer)
+async def disk():
+    """Image/disk usage page: reclaimable space, per-image sizes, stopped count.
+
+    Read-only — prune actions are deferred to the cleanup-actions milestone, so
+    this surfaces the numbers and notes where reclaim lives.
+    """
+    async with current_app.db_sessionmaker() as db:
+        summaries = list((await db.execute(select(DockerDiskUsage))).scalars())
+        images = list((await db.execute(
+            select(DockerImage).order_by(DockerImage.size.desc())
+        )).scalars())
+        # Stopped-container count per host, surfaced alongside reclaimable space.
+        stopped_rows = (await db.execute(
+            select(DockerContainer.host_id)
+            .where(DockerContainer.status != "running")
+        )).scalars()
+        stopped_by_host: dict[str, int] = {}
+        for hid in stopped_rows:
+            stopped_by_host[hid] = stopped_by_host.get(hid, 0) + 1
+        hosts = await _host_map(db, {s.host_id for s in summaries})
+
+    images_by_host: dict[str, list] = {}
+    for im in images:
+        images_by_host.setdefault(im.host_id, []).append({
+            "repo_tag": im.repo_tag, "size": _human_bytes(im.size),
+            "shared": _human_bytes(im.shared_size), "containers": im.containers,
+            "reclaimable": im.containers == 0,
+        })
+
+    host_groups = [
+        {
+            "host_id": s.host_id,
+            "host_name": hosts[s.host_id].name if s.host_id in hosts else s.host_id,
+            "summary": {
+                "images_total": s.images_total, "images_active": s.images_active,
+                "images_size": _human_bytes(s.images_size),
+                "images_reclaimable": _human_bytes(s.images_reclaimable),
+                "layers_size": _human_bytes(s.layers_size),
+                "containers_count": s.containers_count,
+                "containers_size": _human_bytes(s.containers_size),
+                "volumes_count": s.volumes_count,
+                "volumes_size": _human_bytes(s.volumes_size),
+                "build_cache_size": _human_bytes(s.build_cache_size),
+            },
+            "stopped": stopped_by_host.get(s.host_id, 0),
+            "images": images_by_host.get(s.host_id, []),
+        }
+        for s in summaries
+    ]
+    host_groups.sort(key=lambda g: g["host_name"].lower())
+    return await render_template("docker/disk.html", host_groups=host_groups)
+
+
+@docker_bp.get("/container/<host_id>/<name>/history")
+@require_role(UserRole.viewer)
+async def container_history(host_id: str, name: str):
+    """HTMX fragment: CPU + memory sparklines for the selected time range."""
+    since, range_key = parse_range(request.args.get("range"))
+    async with current_app.db_sessionmaker() as db:
+        cpu_hist, mem_hist = await _container_history(db, host_id, name, since)
+    return await render_template(
+        "docker/_container_history.html",
+        sparkline_cpu=_sparkline(cpu_hist, width=320, height=48),
+        sparkline_mem=_sparkline(mem_hist, width=320, height=48),
+        have_data=len(cpu_hist) >= 2,
+        range_key=range_key,
+    )

plugins/docker/swarm_view.py

@@ -0,0 +1,140 @@
+"""Model-free builder for the swarm-aware container view.
+
+Merges two sources into one service-grouped, cluster-complete picture:
+  • docker_containers — real container rows, collected LOCAL-per-node, so each
+    carries the host (= node) it runs on plus its swarm service/node labels.
+  • DockerSwarmService.placement_json — the managers' cluster-wide task→node
+    counts, which include tasks on nodes that run no Steward agent (and so have
+    no container row of their own).
+
+For every swarm service we list the real replicas we have detail for, then add
+"ghost" replicas for the remaining placement count on each node (agent-less
+nodes). Non-swarm containers pass through grouped by host, unchanged.
+
+Kept model-free (no ORM imports) so it's unit-testable and safe to import in
+tests without the plugin-loader "table already defined" gotcha.
+"""
+from __future__ import annotations
+
+import json
+
+
+def _service_state(running: int, desired: int) -> str:
+    if desired > 0 and running >= desired:
+        return "healthy"
+    if running > 0:
+        return "degraded"
+    return "down"
+
+
+def _placement(service) -> list[dict]:
+    raw = getattr(service, "placement_json", None) or "[]"
+    try:
+        data = json.loads(raw)
+    except (ValueError, TypeError):
+        return []
+    return [p for p in data if isinstance(p, dict)]
+
+
+def build_swarm_services(containers, services, node_hostname: dict, host_name: dict) -> dict:
+    """Return {"services": [...], "standalone": [...]}.
+
+    services[i] = {name, mode, desired, running, state, replicas:[...]}
+      real replica  = {ghost:False, host, host_id, name, status, cpu_pct, mem_pct,
+                       health, restart_count}
+      ghost replica = {ghost:True, host, count}   # placement with no local row
+    standalone[i] = {host_id, host, containers:[...]}  # non-swarm, by host
+    """
+    containers = list(containers)
+    swarm_cs = [c for c in containers if getattr(c, "service_name", None)]
+    standalone_cs = [c for c in containers if not getattr(c, "service_name", None)]
+
+    # Dedup service rows by name (every manager reports the same cluster-global
+    # set); keep the first — placement is identical across managers.
+    svc_by_name: dict[str, object] = {}
+    for s in services:
+        name = getattr(s, "service_name", None)
+        if name and name not in svc_by_name:
+            svc_by_name[name] = s
+
+    # Real containers grouped by service name.
+    cs_by_service: dict[str, list] = {}
+    for c in swarm_cs:
+        cs_by_service.setdefault(c.service_name, []).append(c)
+
+    def _host_of(c):
+        return (host_name.get(getattr(c, "host_id", None))
+                or node_hostname.get(getattr(c, "node_id", None))
+                or getattr(c, "host_id", None) or "?")
+
+    out_services = []
+    for name in sorted(set(svc_by_name) | set(cs_by_service)):
+        svc = svc_by_name.get(name)
+        reals = cs_by_service.get(name, [])
+
+        replicas = []
+        for c in sorted(reals, key=lambda c: (_host_of(c).lower(), c.name)):
+            replicas.append({
+                "ghost": False,
+                "host": _host_of(c),
+                "host_id": getattr(c, "host_id", None),
+                "name": c.name,
+                "status": getattr(c, "status", "unknown"),
+                "cpu_pct": getattr(c, "cpu_pct", None),
+                "mem_pct": getattr(c, "mem_pct", None),
+                "health": getattr(c, "health", None),
+                "restart_count": getattr(c, "restart_count", 0) or 0,
+            })
+
+        # Real running replicas per node, to subtract from placement.
+        real_running_by_node: dict[str, int] = {}
+        for c in reals:
+            if (getattr(c, "status", "") or "").lower() == "running":
+                nid = getattr(c, "node_id", None)
+                if nid:
+                    real_running_by_node[nid] = real_running_by_node.get(nid, 0) + 1
+
+        ghosts = []
+        for p in _placement(svc) if svc is not None else []:
+            nid = p.get("node_id", "")
+            placed = int(p.get("running", 0) or 0)
+            ghost = placed - real_running_by_node.get(nid, 0)
+            if ghost > 0:
+                ghosts.append({
+                    "ghost": True,
+                    "host": node_hostname.get(nid) or (nid[:12] if nid else "?"),
+                    "count": ghost,
+                })
+        ghosts.sort(key=lambda g: g["host"].lower())
+        replicas.extend(ghosts)
+
+        running = getattr(svc, "running", None) if svc is not None else None
+        desired = getattr(svc, "desired", None) if svc is not None else None
+        # No service row (manager didn't report it): infer from what we can see.
+        if running is None:
+            running = sum(1 for r in replicas
+                          if (not r["ghost"] and (r["status"] or "").lower() == "running"))
+        if desired is None:
+            desired = len(replicas)
+
+        out_services.append({
+            "name": name,
+            "mode": getattr(svc, "mode", "replicated") if svc is not None else "replicated",
+            "running": running,
+            "desired": desired,
+            "state": _service_state(int(running or 0), int(desired or 0)),
+            "replicas": replicas,
+        })
+
+    # Non-swarm containers grouped by host, running-first order preserved.
+    groups: dict[str, dict] = {}
+    for c in standalone_cs:
+        hid = getattr(c, "host_id", None)
+        g = groups.get(hid)
+        if g is None:
+            g = groups[hid] = {"host_id": hid, "host": host_name.get(hid) or hid or "?",
+                               "containers": []}
+        g["containers"].append(c)
+    standalone = sorted(groups.values(), key=lambda g: g["host"].lower())
+
+    return {"services": out_services, "standalone": standalone}

plugins/docker/templates/docker/_container_history.html

@@ -0,0 +1,17 @@
+{# docker/_container_history.html — CPU/mem sparklines for the selected range #}
+{% if have_data %}
+<div style="display:grid;grid-template-columns:repeat(auto-fit,minmax(280px,1fr));gap:1.5rem;">
+  <div>
+    <div class="section-title" style="margin-bottom:0.4rem;">CPU %</div>
+    {{ sparkline_cpu | safe }}
+  </div>
+  <div>
+    <div class="section-title" style="margin-bottom:0.4rem;">Memory %</div>
+    {{ sparkline_mem | safe }}
+  </div>
+</div>
+{% else %}
+<div style="color:var(--text-muted);font-size:0.85rem;">
+  Not enough samples in this range yet — history fills in as the agent reports.
+</div>
+{% endif %}

plugins/docker/templates/docker/container_detail.html

@@ -0,0 +1,110 @@
+{# docker/container_detail.html — full detail page for one container #}
+{% extends "base.html" %}
+{% from "_macros.html" import crumbs %}
+{% block title %}{{ name }} — Docker — Steward{% endblock %}
+{% block breadcrumb %}{{ crumbs([("Docker", "/plugins/docker/"), (name, "")]) }}{% endblock %}
+{% block content %}
+
+{% if container is none %}
+<div class="card" style="text-align:center;padding:3rem;">
+  <div style="font-weight:600;margin-bottom:0.4rem;">Container not found</div>
+  <div style="color:var(--text-muted);font-size:0.9rem;">
+    No container named <code>{{ name }}</code> is currently reported for this host.
+    It may have been removed, or the host agent hasn't reported recently.
+  </div>
+</div>
+{% else %}
+
+{# ── Header ──────────────────────────────────────────────────────────────── #}
+<div style="display:flex;align-items:center;gap:0.6rem;margin-bottom:0.35rem;flex-wrap:wrap;">
+  <span class="dot {% if container.status == 'running' %}dot-up{% elif container.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
+  <h1 class="page-title" style="margin-bottom:0;">{{ container.name }}</h1>
+  {% if container.health == 'healthy' %}<span style="font-size:0.72rem;color:var(--green);border:1px solid var(--green);border-radius:3px;padding:0.05rem 0.4rem;">healthy</span>
+  {% elif container.health == 'unhealthy' %}<span style="font-size:0.72rem;color:var(--red);border:1px solid var(--red);border-radius:3px;padding:0.05rem 0.4rem;">unhealthy</span>
+  {% elif container.health == 'starting' %}<span style="font-size:0.72rem;color:var(--orange);border:1px solid var(--orange);border-radius:3px;padding:0.05rem 0.4rem;">starting</span>{% endif %}
+</div>
+<div style="font-size:0.82rem;color:var(--text-muted);margin-bottom:1.5rem;">
+  {{ container.status }}{% if uptime %} · up {{ uptime }}{% endif %}
+  {% if host %} · on <a href="/hosts/{{ host.id }}" style="color:var(--text-muted);">{{ host.name }}</a>{% endif %}
+</div>
+
+{# ── Facts grid ──────────────────────────────────────────────────────────── #}
+{% macro fact(label, value, colour="") %}
+<div class="card" style="margin-bottom:0;">
+  <div class="section-title" style="margin-bottom:0.3rem;">{{ label }}</div>
+  <div style="font-size:0.95rem;{% if colour %}color:{{ colour }};{% endif %}font-family:ui-monospace,monospace;">{{ value }}</div>
+</div>
+{% endmacro %}
+
+<div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(150px,1fr));gap:0.75rem;margin-bottom:1.5rem;">
+  {{ fact("CPU", "%.1f%%" | format(container.cpu_pct) if container.cpu_pct is not none else "—") }}
+  {{ fact("Memory", "%.1f%%" | format(container.mem_pct) if container.mem_pct is not none else "—") }}
+  {{ fact("Restarts", container.restart_count, "var(--orange)" if container.restart_count else "") }}
+  {{ fact("Last exit", (container.exit_code ~ (" (OOM)" if container.oom_killed else "")) if container.exit_code is not none else "—",
+          "var(--red)" if (container.exit_code is not none and container.exit_code != 0) else "") }}
+  {{ fact("Net in", net_rx) }}
+  {{ fact("Net out", net_tx) }}
+  {{ fact("Block read", blk_read) }}
+  {{ fact("Block write", blk_write) }}
+</div>
+
+{# ── Image / placement ───────────────────────────────────────────────────── #}
+<div class="card" style="margin-bottom:1.5rem;">
+  <div style="display:grid;grid-template-columns:max-content 1fr;gap:0.4rem 1.25rem;font-size:0.86rem;">
+    <span style="color:var(--text-muted);">Image</span>
+    <span style="font-family:ui-monospace,monospace;word-break:break-all;">{{ container.image or "—" }}</span>
+    {% if container.compose_project %}
+    <span style="color:var(--text-muted);">Compose project</span><span>{{ container.compose_project }}</span>
+    {% endif %}
+    {% if container.service_name %}
+    <span style="color:var(--text-muted);">Swarm service</span><span>{{ container.service_name }}</span>
+    {% endif %}
+    {% if container.node_id %}
+    <span style="color:var(--text-muted);">Node</span><span style="font-family:ui-monospace,monospace;">{{ container.node_id }}</span>
+    {% endif %}
+    {% if ports %}
+    <span style="color:var(--text-muted);">Ports</span>
+    <span style="font-family:ui-monospace,monospace;">{% for p in ports %}{{ p.host_port }}:{{ p.container_port }}/{{ p.protocol }}{% if not loop.last %}, {% endif %}{% endfor %}</span>
+    {% endif %}
+  </div>
+</div>
+
+{# ── Resource history (range-toggled) ────────────────────────────────────── #}
+<div class="card" style="margin-bottom:1.5rem;">
+  <div style="display:flex;align-items:center;justify-content:space-between;gap:1rem;margin-bottom:0.75rem;flex-wrap:wrap;">
+    <h3 class="section-title" style="margin-bottom:0;">Resource history</h3>
+    {% include "_time_range.html" %}
+  </div>
+  <div id="container-history"
+       hx-get="/plugins/docker/container/{{ host_id }}/{{ name }}/history"
+       hx-trigger="load, rangeChange from:body"
+       hx-include="#time-range"
+       hx-swap="innerHTML">
+    <div style="color:var(--text-muted);font-size:0.85rem;">Loading…</div>
+  </div>
+</div>
+
+{# ── Lifecycle timeline ──────────────────────────────────────────────────── #}
+<div class="card">
+  <h3 class="section-title" style="margin-bottom:0.75rem;">Lifecycle</h3>
+  {% if events %}
+  <div style="display:grid;gap:0.5rem;">
+    {% for e in events %}
+    <div style="display:flex;align-items:baseline;gap:0.6rem;font-size:0.85rem;">
+      <span style="color:{{ e.colour }};width:1rem;text-align:center;flex-shrink:0;">{{ e.glyph }}</span>
+      <span style="font-weight:500;width:6.5rem;flex-shrink:0;">{{ e.event }}</span>
+      <span style="color:var(--text-muted);flex:1;min-width:0;">{{ e.detail or "" }}</span>
+      <span style="color:var(--text-dim);font-size:0.78rem;white-space:nowrap;" title="{{ e.at }}">{{ e.at.strftime("%Y-%m-%d %H:%M") }}</span>
+    </div>
+    {% endfor %}
+  </div>
+  {% else %}
+  <div style="color:var(--text-muted);font-size:0.85rem;">
+    No lifecycle events recorded yet. Start/stop/health changes appear here as the
+    agent reports them over time.
+  </div>
+  {% endif %}
+</div>
+
+{% endif %}
+{% endblock %}

plugins/docker/templates/docker/disk.html

@@ -0,0 +1,83 @@
+{# docker/disk.html — image/disk usage: reclaimable space, per-image sizes #}
+{% extends "base.html" %}
+{% from "_macros.html" import crumbs %}
+{% block title %}Disk — Docker — Steward{% endblock %}
+{% block breadcrumb %}{{ crumbs([("Docker", "/plugins/docker/"), ("Image & disk usage", "")]) }}{% endblock %}
+{% block content %}
+
+<h1 class="page-title" style="margin-bottom:0.4rem;">Image &amp; disk usage</h1>
+<p style="font-size:0.82rem;color:var(--text-muted);margin-bottom:1.5rem;">
+  Reclaimable = space held by images no container references. Cleanup actions
+  (prune) arrive in a later release — these are read-only figures for now.
+</p>
+
+{% if host_groups %}
+{% for g in host_groups %}
+<div style="margin-bottom:2rem;">
+  <div style="font-weight:600;font-size:0.95rem;margin-bottom:0.75rem;">{{ g.host_name }}</div>
+
+  {# ── Summary stats ────────────────────────────────────────────────────── #}
+  <div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(150px,1fr));gap:0.75rem;margin-bottom:1rem;">
+    <div class="card" style="margin-bottom:0;">
+      <div class="section-title" style="margin-bottom:0.3rem;">Reclaimable</div>
+      <span class="stat-val" style="color:{% if g.summary.images_reclaimable != '0 B' %}var(--orange){% else %}var(--green){% endif %};">{{ g.summary.images_reclaimable }}</span>
+    </div>
+    <div class="card" style="margin-bottom:0;">
+      <div class="section-title" style="margin-bottom:0.3rem;">Images</div>
+      <span class="stat-val">{{ g.summary.images_size }}</span>
+      <div style="font-size:0.74rem;color:var(--text-muted);">{{ g.summary.images_active }}/{{ g.summary.images_total }} in use</div>
+    </div>
+    <div class="card" style="margin-bottom:0;">
+      <div class="section-title" style="margin-bottom:0.3rem;">Stopped</div>
+      <span class="stat-val" style="{% if g.stopped %}color:var(--text-muted){% endif %};">{{ g.stopped }}</span>
+    </div>
+    <div class="card" style="margin-bottom:0;">
+      <div class="section-title" style="margin-bottom:0.3rem;">Writable layers</div>
+      <span class="stat-val">{{ g.summary.containers_size }}</span>
+      <div style="font-size:0.74rem;color:var(--text-muted);">{{ g.summary.containers_count }} containers</div>
+    </div>
+    <div class="card" style="margin-bottom:0;">
+      <div class="section-title" style="margin-bottom:0.3rem;">Volumes</div>
+      <span class="stat-val">{{ g.summary.volumes_size }}</span>
+      <div style="font-size:0.74rem;color:var(--text-muted);">{{ g.summary.volumes_count }} volumes</div>
+    </div>
+    <div class="card" style="margin-bottom:0;">
+      <div class="section-title" style="margin-bottom:0.3rem;">Build cache</div>
+      <span class="stat-val">{{ g.summary.build_cache_size }}</span>
+    </div>
+  </div>
+
+  {# ── Per-image table ──────────────────────────────────────────────────── #}
+  <div class="card-flush">
+    <table class="table">
+      <thead>
+        <tr><th>Image</th><th>Size</th><th>Shared</th><th>Containers</th></tr>
+      </thead>
+      <tbody>
+        {% for im in g.images %}
+        <tr>
+          <td style="font-size:0.84rem;font-family:ui-monospace,monospace;max-width:340px;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">
+            {{ im.repo_tag }}
+            {% if im.reclaimable %}<span title="not referenced by any container" style="font-size:0.68rem;color:var(--orange);border:1px solid var(--orange);border-radius:3px;padding:0 0.3rem;margin-left:0.4rem;">reclaimable</span>{% endif %}
+          </td>
+          <td style="font-family:ui-monospace,monospace;font-size:0.86rem;">{{ im.size }}</td>
+          <td style="font-family:ui-monospace,monospace;font-size:0.86rem;color:var(--text-muted);">{{ im.shared }}</td>
+          <td style="font-size:0.86rem;color:{% if im.containers %}var(--text){% else %}var(--text-muted){% endif %};">{{ im.containers }}</td>
+        </tr>
+        {% else %}
+        <tr><td colspan="4" style="color:var(--text-muted);font-size:0.85rem;text-align:center;padding:1.5rem;">No images reported.</td></tr>
+        {% endfor %}
+      </tbody>
+    </table>
+  </div>
+</div>
+{% endfor %}
+{% else %}
+<div class="card" style="text-align:center;padding:3rem;">
+  <div style="color:var(--text-muted);font-size:0.9rem;">
+    No disk usage reported yet. The host agent reports image/disk usage from
+    <code>docker system df</code> on hosts running Docker.
+  </div>
+</div>
+{% endif %}
+{% endblock %}

plugins/docker/templates/docker/host_panel.html

@@ -11,7 +11,7 @@
     {% for c in containers %}
     <div style="display:flex;align-items:center;gap:0.5rem;font-size:0.85rem;padding:0.2rem 0;">
       <span class="dot {% if c.status == 'running' %}dot-up{% elif c.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
-      <span style="flex:1;min-width:0;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;" title="{{ c.image }}">{{ c.name }}</span>
+      <a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}" style="flex:1;min-width:0;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;color:inherit;text-decoration:none;" title="{{ c.image }}">{{ c.name }}</a>
       <span style="font-size:0.74rem;color:var(--text-muted);font-family:ui-monospace,monospace;flex-shrink:0;">
         {% if c.cpu_pct is not none %}{{ "%.1f" | format(c.cpu_pct) }}%{% endif %}
         {% if c.mem_pct is not none %} · {{ "%.1f" | format(c.mem_pct) }}%{% endif %}

plugins/docker/templates/docker/index.html

@@ -2,7 +2,15 @@
 {% block title %}Docker — Steward{% endblock %}
 {% block content %}
 <div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:1.5rem;gap:1rem;flex-wrap:wrap;">
-  <h1 class="page-title" style="margin-bottom:0;">Docker</h1>
+  <div style="display:flex;align-items:baseline;gap:1rem;">
+    <h1 class="page-title" style="margin-bottom:0;">Docker</h1>
+    {% if has_swarm %}
+    <a href="/plugins/docker/swarm" style="font-size:0.85rem;color:var(--accent);text-decoration:none;">Swarm →</a>
+    {% endif %}
+    {% if has_disk %}
+    <a href="/plugins/docker/disk" style="font-size:0.85rem;color:var(--accent);text-decoration:none;">Disk →</a>
+    {% endif %}
+  </div>
   {% include "_time_range.html" %}
 </div>
 

plugins/docker/templates/docker/rows.html

@@ -14,12 +14,61 @@
     <div class="section-title" style="margin-bottom:0.35rem;">Total</div>
     <span class="stat-val">{{ total }}</span>
   </div>
+  {% if swarm_services %}
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.35rem;">Services</div>
+    <span class="stat-val">{{ swarm_services | length }}</span>
+  </div>
+  {% endif %}
   <div class="card" style="margin-bottom:0;">
     <div class="section-title" style="margin-bottom:0.35rem;">Hosts</div>
     <span class="stat-val">{{ host_groups | length }}</span>
   </div>
 </div>
 
+{# ── Swarm services: collapsed per-service, each replica labelled with its host.
+   Ghost replicas (tasks on nodes with no Steward agent) come from the managers'
+   placement data so the picture is cluster-complete. ──────────────────────── #}
+{% if swarm_services %}
+<div class="card-flush" style="margin-bottom:1.5rem;">
+  <div style="display:flex;align-items:baseline;gap:0.6rem;padding:0.5rem 0.75rem;border-bottom:1px solid var(--border);">
+    <span style="font-weight:600;font-size:0.95rem;">Swarm services</span>
+    <a href="/plugins/docker/swarm" style="font-size:0.78rem;color:var(--text-muted);">Topology →</a>
+  </div>
+  <div style="padding:0.5rem 0.75rem;display:grid;gap:0.85rem;">
+    {% for s in swarm_services %}
+    <div>
+      <div style="display:flex;align-items:center;gap:0.55rem;flex-wrap:wrap;margin-bottom:0.35rem;">
+        <span class="dot {% if s.state == 'healthy' %}dot-up{% elif s.state == 'degraded' %}dot-warn{% else %}dot-down{% endif %}"></span>
+        <span style="font-weight:600;font-size:0.9rem;">{{ s.name }}</span>
+        <span style="font-size:0.72rem;color:var(--text-dim);">{{ s.mode }}</span>
+        <span style="font-size:0.78rem;color:{% if s.state == 'healthy' %}var(--green){% elif s.state == 'degraded' %}var(--orange){% else %}var(--red){% endif %};font-variant-numeric:tabular-nums;">
+          {{ s.running }}/{{ s.desired }} running
+        </span>
+      </div>
+      <div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(210px,1fr));gap:0.4rem;padding-left:1.1rem;">
+        {% for r in s.replicas %}
+        {% if r.ghost %}
+        <div style="display:flex;align-items:center;gap:0.4rem;font-size:0.8rem;background:var(--bg-elevated);border:1px dashed var(--border-mid);border-radius:5px;padding:0.3rem 0.55rem;" title="Running on a node with no Steward agent — detail unavailable">
+          <span class="dot dot-warn" style="opacity:0.6;"></span>
+          <span style="font-weight:500;">{{ r.host }}</span>
+          <span style="color:var(--text-dim);margin-left:auto;">{{ r.count }} · no agent</span>
+        </div>
+        {% else %}
+        <div style="display:flex;align-items:center;gap:0.4rem;font-size:0.8rem;background:var(--bg-elevated);border:1px solid var(--border);border-radius:5px;padding:0.3rem 0.55rem;">
+          <span class="dot {% if r.status == 'running' %}dot-up{% elif r.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
+          <a href="/plugins/docker/container/{{ r.host_id }}/{{ r.name }}" style="font-weight:500;color:inherit;text-decoration:none;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;" title="{{ r.name }}">{{ r.host }}</a>
+          {% if r.cpu_pct is not none %}<span style="color:var(--text-muted);font-family:ui-monospace,monospace;margin-left:auto;">{{ "%.0f" | format(r.cpu_pct) }}%</span>{% endif %}
+        </div>
+        {% endif %}
+        {% endfor %}
+      </div>
+    </div>
+    {% endfor %}
+  </div>
+</div>
+{% endif %}
+
 {# ── Container tables, one per host ───────────────────────────────────────── #}
 {% if host_groups %}
 {% for g in host_groups %}
@@ -47,7 +96,11 @@
       </tr>
     </thead>
     <tbody>
-      {% for item in g.containers %}
+      {% for sub in g.subgroups %}
+      {% if g.grouped and sub.label %}
+      <tr><td colspan="7" style="padding:0.4rem 0.75rem 0.2rem;font-size:0.72rem;color:var(--text-dim);text-transform:uppercase;letter-spacing:0.05em;font-weight:600;">{{ sub.label }}</td></tr>
+      {% endif %}
+      {% for item in sub.containers %}
       {% set c = item.container %}
       <tr>
         <td>
@@ -55,7 +108,7 @@
             <span class="dot {% if c.status == 'running' %}dot-up{% elif c.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
             <div>
               <div style="font-weight:500;font-size:0.9rem;">
-                {{ c.name }}
+                <a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}" style="color:inherit;text-decoration:none;border-bottom:1px dotted var(--border-mid);">{{ c.name }}</a>
                 {% if c.health == 'healthy' %}<span title="healthy" style="color:var(--green);font-size:0.7rem;">●</span>
                 {% elif c.health == 'unhealthy' %}<span title="unhealthy" style="color:var(--red);font-size:0.7rem;">●</span>
                 {% elif c.health == 'starting' %}<span title="health: starting" style="color:var(--orange);font-size:0.7rem;">◐</span>{% endif %}
@@ -105,11 +158,12 @@
         <td>{{ item.sparkline_mem | safe }}</td>
       </tr>
       {% endfor %}
+      {% endfor %}
     </tbody>
   </table>
 </div>
 {% endfor %}
-{% else %}
+{% elif not swarm_services %}
 <div class="card" style="text-align:center;padding:3rem;">
   <div style="color:var(--text-muted);font-size:0.9rem;">
     No containers reported yet. Containers are collected by the Steward host agent —

plugins/docker/templates/docker/swarm.html

@@ -0,0 +1,87 @@
+{# docker/swarm.html — Swarm topology: services, replica health, nodes, placement #}
+{% extends "base.html" %}
+{% from "_macros.html" import crumbs %}
+{% block title %}Swarm — Docker — Steward{% endblock %}
+{% block breadcrumb %}{{ crumbs([("Docker", "/plugins/docker/"), ("Swarm", "")]) }}{% endblock %}
+{% block content %}
+
+<h1 class="page-title" style="margin-bottom:1.5rem;">Swarm</h1>
+
+{% if swarms %}
+{% for g in swarms %}
+<div style="margin-bottom:2rem;">
+  <div style="display:flex;align-items:baseline;gap:0.6rem;margin-bottom:0.75rem;flex-wrap:wrap;">
+    <span style="font-weight:600;font-size:0.95rem;">Swarm</span>
+    <span style="font-size:0.78rem;color:var(--text-muted);">
+      reported by {{ g.managers | length }} manager{{ 's' if g.managers | length != 1 }} · {{ g.managers | join(", ") }}
+    </span>
+  </div>
+
+  {# ── Services ─────────────────────────────────────────────────────────── #}
+  <div class="card-flush" style="margin-bottom:1rem;">
+    <table class="table">
+      <thead>
+        <tr>
+          <th>Service</th><th>Mode</th><th>Replicas</th><th>Image</th><th>Placement</th>
+        </tr>
+      </thead>
+      <tbody>
+        {% for s in g.services %}
+        <tr>
+          <td style="font-weight:500;font-size:0.9rem;">{{ s.name }}</td>
+          <td style="font-size:0.82rem;color:var(--text-muted);">{{ s.mode }}</td>
+          <td style="font-family:ui-monospace,monospace;font-size:0.88rem;">
+            <span style="color:{% if s.healthy %}var(--green){% elif s.down %}var(--red){% elif s.degraded %}var(--orange){% else %}var(--text-muted){% endif %};">
+              {{ s.running }}/{{ s.desired }}
+            </span>
+          </td>
+          <td style="font-size:0.8rem;color:var(--text-muted);max-width:220px;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">{{ s.image or "—" }}</td>
+          <td style="font-size:0.78rem;color:var(--text-muted);">
+            {% for p in s.placement %}{{ p.node }} ({{ p.running }}){% if not loop.last %}, {% endif %}{% else %}—{% endfor %}
+          </td>
+        </tr>
+        {% else %}
+        <tr><td colspan="5" style="color:var(--text-muted);font-size:0.85rem;text-align:center;padding:1.5rem;">No services in this swarm.</td></tr>
+        {% endfor %}
+      </tbody>
+    </table>
+  </div>
+
+  {# ── Nodes ────────────────────────────────────────────────────────────── #}
+  <div class="card-flush">
+    <table class="table">
+      <thead>
+        <tr><th>Node</th><th>Role</th><th>Availability</th><th>Status</th></tr>
+      </thead>
+      <tbody>
+        {% for n in g.nodes %}
+        <tr>
+          <td style="font-weight:500;font-size:0.9rem;">
+            {{ n.hostname or n.node_id }}
+            {% if n.leader %}<span title="cluster leader" style="font-size:0.68rem;color:var(--accent);border:1px solid var(--accent);border-radius:3px;padding:0 0.3rem;margin-left:0.3rem;">leader</span>{% endif %}
+          </td>
+          <td style="font-size:0.82rem;color:var(--text-muted);">{{ n.role }}</td>
+          <td style="font-size:0.82rem;color:{% if n.availability == 'active' %}var(--text){% else %}var(--orange){% endif %};">{{ n.availability }}</td>
+          <td style="font-size:0.82rem;">
+            <span class="dot {% if n.status == 'ready' %}dot-up{% else %}dot-down{% endif %}"></span>
+            <span style="color:{% if n.status == 'ready' %}var(--green){% else %}var(--red){% endif %};">{{ n.status }}</span>
+          </td>
+        </tr>
+        {% else %}
+        <tr><td colspan="4" style="color:var(--text-muted);font-size:0.85rem;text-align:center;padding:1.5rem;">No nodes reported.</td></tr>
+        {% endfor %}
+      </tbody>
+    </table>
+  </div>
+</div>
+{% endfor %}
+{% else %}
+<div class="card" style="text-align:center;padding:3rem;">
+  <div style="color:var(--text-muted);font-size:0.9rem;">
+    No Swarm topology reported. A Steward host agent running on a Swarm
+    <strong>manager</strong> reports services, nodes, and task placement here —
+    workers and non-swarm hosts contribute only their local containers.
+  </div>
+</div>
+{% endif %}
+{% endblock %}

plugins/docker/templates/docker/widget.html

@@ -1,21 +1,45 @@
 {# docker/widget.html — dashboard widget: container status overview, by host #}
-{% if running_count == 0 and stopped_count == 0 %}
+{% if total_count == 0 %}
 <p class="empty" style="padding:0.5rem 0;font-size:0.85rem;">No containers reported yet.</p>
 {% else %}
 
-<div style="display:flex;gap:1rem;margin-bottom:0.65rem;">
+<div style="display:flex;gap:1.25rem;margin-bottom:0.65rem;">
   <div>
     <span style="font-size:1.4rem;font-weight:700;line-height:1;color:var(--green);">{{ running_count }}</span>
     <span style="font-size:0.78rem;color:var(--text-muted);margin-left:0.25rem;">running</span>
   </div>
-  {% if stopped_count %}
-  <div>
-    <span style="font-size:1.4rem;font-weight:700;line-height:1;color:var(--text-muted);">{{ stopped_count }}</span>
-    <span style="font-size:0.78rem;color:var(--text-muted);margin-left:0.25rem;">stopped</span>
+  <div title="Distinct containers with a die or OOM event in the last 24h">
+    <span style="font-size:1.4rem;font-weight:700;line-height:1;color:{% if failed_24h %}var(--red){% else %}var(--text-muted){% endif %};">{{ failed_24h }}</span>
+    <span style="font-size:0.78rem;color:var(--text-muted);margin-left:0.25rem;">failed (24h)</span>
   </div>
-  {% endif %}
 </div>
 
+{# Swarm services — collapsed, each with its replicas' host chips (dashed = a
+   node with no Steward agent, count-only from placement). #}
+{% if swarm_services %}
+<div style="font-size:0.72rem;color:var(--text-dim);text-transform:uppercase;letter-spacing:0.05em;margin:0.5rem 0 0.2rem;">Swarm services</div>
+<div style="display:grid;gap:0.4rem;margin-bottom:0.4rem;">
+  {% for s in swarm_services %}
+  <div>
+    <div style="display:flex;align-items:center;gap:0.4rem;font-size:0.82rem;">
+      <span class="dot {% if s.state == 'healthy' %}dot-up{% elif s.state == 'degraded' %}dot-warn{% else %}dot-down{% endif %}"></span>
+      <span style="font-weight:500;min-width:0;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">{{ s.name }}</span>
+      <span style="color:var(--text-muted);font-family:ui-monospace,monospace;font-size:0.74rem;margin-left:auto;flex-shrink:0;">{{ s.running }}/{{ s.desired }}</span>
+    </div>
+    <div style="display:flex;flex-wrap:wrap;gap:0.25rem;margin:0.2rem 0 0 1rem;">
+      {% for r in s.replicas %}
+      {% if r.ghost %}
+      <span title="no Steward agent on this node — count from swarm placement" style="font-size:0.68rem;color:var(--text-dim);background:var(--bg-elevated);border:1px dashed var(--border-mid);border-radius:4px;padding:0.05rem 0.35rem;">{{ r.host }} ×{{ r.count }}</span>
+      {% else %}
+      <span style="font-size:0.68rem;color:var(--text-muted);background:var(--bg-elevated);border:1px solid var(--border);border-radius:4px;padding:0.05rem 0.35rem;">{{ r.host }}</span>
+      {% endif %}
+      {% endfor %}
+    </div>
+  </div>
+  {% endfor %}
+</div>
+{% endif %}
+
 {% for g in host_groups %}
 {% if multi_host %}
 <div style="font-size:0.72rem;color:var(--text-dim);text-transform:uppercase;letter-spacing:0.05em;margin:0.5rem 0 0.2rem;">{{ g.host_name }}</div>
@@ -24,9 +48,12 @@
   {% for c in g.containers %}
   <div style="display:flex;align-items:center;gap:0.5rem;font-size:0.82rem;padding:0.15rem 0;">
     <span class="dot {% if c.status == 'running' %}dot-up{% elif c.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
-    <span style="flex:1;min-width:0;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">
-      {{ c.name }}
-    </span>
+    <a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}" style="flex:1;min-width:0;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;color:inherit;text-decoration:none;">
+      {{ c.name }}{% if c.health == 'unhealthy' %}<span title="unhealthy" style="color:var(--red);"> ●</span>{% elif c.health == 'healthy' %}<span title="healthy" style="color:var(--green);"> ●</span>{% endif %}{% if c.restart_count %}<span title="restarts" style="color:var(--orange);font-size:0.7rem;"> ⟳{{ c.restart_count }}</span>{% endif %}
+    </a>
+    {% if c.status != 'running' and c.exit_code is not none and c.exit_code != 0 %}
+    <span title="last exit code" style="font-size:0.7rem;color:var(--red);flex-shrink:0;">exit {{ c.exit_code }}</span>
+    {% endif %}
     {% if c.cpu_pct is not none %}
     <span style="font-size:0.72rem;color:var(--text-muted);font-family:ui-monospace,monospace;flex-shrink:0;">
       {{ "%.1f" | format(c.cpu_pct) }}%

plugins/docker/templates/docker/widget_resources.html

@@ -1,36 +1,28 @@
-{# docker/widget_resources.html — dashboard widget: CPU + memory bars, busiest containers #}
+{# docker/widget_resources.html — dashboard widget: the busiest running containers
+   by CPU, each with labeled CPU + memory utilisation bars. #}
 {% if not rows %}
 <div style="color:var(--text-muted);font-size:0.85rem;padding:0.5rem 0;">No running containers.</div>
 {% else %}
-<div style="display:grid;gap:0.5rem;">
+{% macro bar(label, pct, warn, crit) %}
+<div style="display:flex;align-items:center;gap:0.45rem;margin-bottom:2px;">
+  <span style="font-size:0.62rem;color:var(--text-dim);width:1.9rem;flex-shrink:0;letter-spacing:0.03em;">{{ label }}</span>
+  <div style="flex:1;height:6px;background:var(--bg-elevated);border-radius:3px;overflow:hidden;">
+    <div style="height:100%;border-radius:3px;width:{{ [pct, 100] | min }}%;
+                background:{% if pct > crit %}var(--red){% elif pct > warn %}var(--orange){% else %}var(--accent){% endif %};
+                transition:width 0.4s;"></div>
+  </div>
+  <span style="font-size:0.7rem;color:var(--text-muted);font-family:ui-monospace,monospace;width:3rem;text-align:right;flex-shrink:0;">{{ "%.1f" | format(pct) }}%</span>
+</div>
+{% endmacro %}
+<div style="display:grid;gap:0.6rem;">
   {% for r in rows %}
   {% set c = r.c %}
   <div>
-    <div style="display:flex;justify-content:space-between;align-items:baseline;margin-bottom:0.15rem;">
-      <span style="font-size:0.8rem;font-weight:500;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;max-width:55%;">
-        {{ c.name }}{% if multi_host %}<span style="color:var(--text-dim);font-weight:normal;"> · {{ r.host_name }}</span>{% endif %}
-      </span>
-      <span style="font-size:0.72rem;color:var(--text-muted);font-family:ui-monospace,monospace;flex-shrink:0;margin-left:0.5rem;">
-        {% if c.cpu_pct is not none %}CPU {{ "%.1f" | format(c.cpu_pct) }}%{% endif %}
-        {% if c.mem_pct is not none %} · Mem {{ "%.1f" | format(c.mem_pct) }}%{% endif %}
-      </span>
+    <div style="font-size:0.8rem;font-weight:500;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;margin-bottom:0.25rem;">
+      {{ c.name }}{% if multi_host %}<span style="color:var(--text-dim);font-weight:normal;"> · {{ r.host_name }}</span>{% endif %}
     </div>
-    {% if c.cpu_pct is not none %}
-    <div style="height:3px;background:var(--bg-elevated);border-radius:2px;margin-bottom:2px;">
-      <div style="height:100%;border-radius:2px;width:{{ [c.cpu_pct, 100] | min }}%;
-                  background:{% if c.cpu_pct > 80 %}var(--red){% elif c.cpu_pct > 50 %}var(--orange){% else %}var(--accent){% endif %};
-                  transition:width 0.4s;">
-      </div>
-    </div>
-    {% endif %}
-    {% if c.mem_pct is not none %}
-    <div style="height:3px;background:var(--bg-elevated);border-radius:2px;">
-      <div style="height:100%;border-radius:2px;width:{{ [c.mem_pct, 100] | min }}%;
-                  background:{% if c.mem_pct > 90 %}var(--red){% elif c.mem_pct > 70 %}var(--orange){% else %}var(--green){% endif %};
-                  transition:width 0.4s;">
-      </div>
-    </div>
-    {% endif %}
+    {% if c.cpu_pct is not none %}{{ bar("CPU", c.cpu_pct, 50, 80) }}{% endif %}
+    {% if c.mem_pct is not none %}{{ bar("MEM", c.mem_pct, 70, 90) }}{% endif %}
   </div>
   {% endfor %}
 </div>

plugins/host_agent/metrics_query.py

@@ -0,0 +1,112 @@
+"""Host-metric read helpers (latest snapshot + bucketed history).
+
+Kept separate from routes.py so it imports only the core PluginMetric model — not
+the host_agent ORM models — which lets integration tests import these helpers
+without tripping the plugin-loader's double-registration ("Table already
+defined") guard.
+"""
+from __future__ import annotations
+
+from datetime import datetime, timedelta, timezone
+
+from sqlalchemy import func, or_, select
+
+from steward.core.time_range import bucket_seconds
+from steward.models.metrics import PluginMetric, PluginMetricHourly
+
+SOURCE_MODULE = "host_agent"
+
+# Host-level metrics charted on the detail page (sub-resources are shown as
+# current-value lists, not time series, to keep the page readable).
+HISTORY_METRICS = (
+    "cpu_pct", "mem_used_pct", "disk_used_pct_worst", "load_1m",
+    "net_rx_bps", "net_tx_bps", "disk_read_bps", "disk_write_bps",
+    "temp_c_max", "psi_mem_some_avg10",
+)
+
+
+async def _latest_metrics_for_host(session, host_name: str) -> dict[str, dict[str, float]]:
+    """{resource_name: {metric: value}} — latest sample per (resource, metric) for
+    a host + its sub-resources.
+
+    DISTINCT ON picks the newest row per group in one index-ordered pass over
+    ix_plugin_metrics_module_resource_metric_recorded, instead of a GROUP-BY-max
+    subquery self-joined back to the table (two passes over the whole history).
+    """
+    rows = (await session.execute(
+        select(PluginMetric)
+        .where(
+            PluginMetric.source_module == SOURCE_MODULE,
+            or_(
+                PluginMetric.resource_name == host_name,
+                PluginMetric.resource_name.like(host_name + ":%"),
+            ),
+        )
+        .distinct(PluginMetric.resource_name, PluginMetric.metric_name)
+        .order_by(
+            PluginMetric.resource_name,
+            PluginMetric.metric_name,
+            PluginMetric.recorded_at.desc(),
+        )
+    )).scalars().all()
+    out: dict[str, dict[str, float]] = {}
+    for r in rows:
+        out.setdefault(r.resource_name, {})[r.metric_name] = r.value
+    return out
+
+
+async def _history_for_host(session, host_name: str, since, *, raw_days: int = 7) -> dict[str, list]:
+    """{metric: [[epoch_ms, avg_value], …]} host-level series since `since`.
+
+    Retention rolls raw plugin_metrics older than `raw_days` into hourly averages
+    (plugin_metrics_hourly) and deletes the raw rows. So we read the rollup for the
+    part of the range older than that boundary and raw (bucket-averaged in SQL,
+    capped to ≤1h to match the rollup) for the recent part — never shipping raw
+    samples to Python. The two windows don't overlap, so appending hourly-then-raw
+    keeps each metric's series time-ordered. Epoch-ms x feeds a linear chart axis.
+    """
+    now = datetime.now(timezone.utc)
+    raw_cutoff = (now - timedelta(days=raw_days)).replace(minute=0, second=0, microsecond=0)
+    series: dict[str, list] = {m: [] for m in HISTORY_METRICS}
+
+    # ── Older than the raw window: the hourly rollup ──
+    if since < raw_cutoff:
+        hourly = (await session.execute(
+            select(PluginMetricHourly.metric_name, PluginMetricHourly.bucket,
+                   PluginMetricHourly.value_avg)
+            .where(
+                PluginMetricHourly.source_module == SOURCE_MODULE,
+                PluginMetricHourly.resource_name == host_name,
+                PluginMetricHourly.metric_name.in_(HISTORY_METRICS),
+                PluginMetricHourly.bucket >= since,
+                PluginMetricHourly.bucket < raw_cutoff,
+            )
+            .order_by(PluginMetricHourly.bucket)
+        )).all()
+        for metric_name, bucket, avg in hourly:
+            series[metric_name].append([int(bucket.timestamp() * 1000), round(float(avg), 2)])
+
+    # ── Recent part: raw, bucket-averaged in SQL ──
+    raw_since = since if since >= raw_cutoff else raw_cutoff
+    width_s = bucket_seconds(raw_since, 120)
+    if since < raw_cutoff:
+        width_s = min(width_s, 3600)  # ≤ 1h so it matches the hourly portion
+    rbucket = func.date_bin(
+        func.make_interval(0, 0, 0, 0, 0, 0, width_s),
+        PluginMetric.recorded_at,
+        func.to_timestamp(0),                            # epoch origin
+    ).label("bucket")
+    raw = (await session.execute(
+        select(PluginMetric.metric_name, rbucket, func.avg(PluginMetric.value))
+        .where(
+            PluginMetric.source_module == SOURCE_MODULE,
+            PluginMetric.resource_name == host_name,
+            PluginMetric.metric_name.in_(HISTORY_METRICS),
+            PluginMetric.recorded_at >= raw_since,
+        )
+        .group_by(PluginMetric.metric_name, rbucket)
+        .order_by(rbucket)
+    )).all()
+    for metric_name, b, avg in raw:
+        series[metric_name].append([int(b.timestamp() * 1000), round(float(avg), 2)])
+    return series

plugins/host_agent/routes.py

@@ -14,16 +14,21 @@ from steward.models.users import UserRole
 from sqlalchemy import select, func, or_, and_
 from datetime import timedelta
 
-from steward.core.settings import public_base_url
+from steward.core.settings import get_setting, public_base_url
 from steward.core.time_range import parse_range, RANGE_OPTIONS
 from steward.models.hosts import Host
 from steward.models.metrics import PluginMetric
 from .models import HostAgentRegistration
+# Query helpers live in a model-free module so integration tests can import them
+# without the plugin-loader double-registration guard tripping (see metrics_query).
+from .metrics_query import (
+    SOURCE_MODULE,
+    _history_for_host,
+    _latest_metrics_for_host,
+)
 
 host_agent_bp = Blueprint("host_agent", __name__, template_folder="templates")
 
-SOURCE_MODULE = "host_agent"
-
 
 def _hash_token(raw: str) -> str:
     return hashlib.sha256(raw.encode("utf-8")).hexdigest()
@@ -540,89 +545,25 @@ def _downsample(series: list[list], target: int = 120) -> list[list]:
     return out
 
 
-# Host-level metrics charted on the detail page (sub-resources are shown as
-# current-value lists, not time series, to keep the page readable).
-HISTORY_METRICS = (
-    "cpu_pct", "mem_used_pct", "disk_used_pct_worst", "load_1m",
-    "net_rx_bps", "net_tx_bps", "disk_read_bps", "disk_write_bps",
-    "temp_c_max", "psi_mem_some_avg10",
-)
+# Full-metrics refresh cadences. The agent pushes every ~30s by default, so
+# polling the current snapshot at 15s keeps it ≤ one cadence stale; the history
+# charts move slowly over a multi-hour/day window, so they refresh less often.
+_DETAIL_POLL_SECONDS = 15
+_CHART_POLL_SECONDS = 60
 
 
-async def _latest_metrics_for_host(session, host_name: str) -> dict[str, dict[str, float]]:
-    """{resource_name: {metric: value}} of the latest sample for a host + sub-resources."""
-    subq = (
-        select(
-            PluginMetric.resource_name,
-            PluginMetric.metric_name,
-            func.max(PluginMetric.recorded_at).label("max_ts"),
-        )
-        .where(PluginMetric.source_module == SOURCE_MODULE)
-        .where(or_(
-            PluginMetric.resource_name == host_name,
-            PluginMetric.resource_name.like(host_name + ":%"),
-        ))
-        .group_by(PluginMetric.resource_name, PluginMetric.metric_name)
-    ).subquery()
-    rows = (await session.execute(
-        select(PluginMetric).join(
-            subq,
-            (PluginMetric.resource_name == subq.c.resource_name) &
-            (PluginMetric.metric_name == subq.c.metric_name) &
-            (PluginMetric.recorded_at == subq.c.max_ts),
-        ).where(PluginMetric.source_module == SOURCE_MODULE)
-    )).scalars().all()
-    out: dict[str, dict[str, float]] = {}
-    for r in rows:
-        out.setdefault(r.resource_name, {})[r.metric_name] = r.value
-    return out
-
-
-async def _history_for_host(session, host_name: str, since) -> dict[str, list]:
-    """{metric: [[epoch_ms, value], …]} host-level series since `since`.
-
-    Epoch-ms x values let the charts use a linear axis (no Chart.js date adapter).
-    """
-    rows = (await session.execute(
-        select(PluginMetric).where(
-            PluginMetric.source_module == SOURCE_MODULE,
-            PluginMetric.resource_name == host_name,
-            PluginMetric.metric_name.in_(HISTORY_METRICS),
-            PluginMetric.recorded_at >= since,
-        ).order_by(PluginMetric.recorded_at)
-    )).scalars().all()
-    series: dict[str, list] = {m: [] for m in HISTORY_METRICS}
-    for p in rows:
-        series[p.metric_name].append([int(p.recorded_at.timestamp() * 1000), round(p.value, 2)])
-    # Downsample to a readable point count (see _downsample) — raw agent cadence
-    # is too dense to read over a multi-hour window.
-    return {m: _downsample(v) for m, v in series.items()}
-
-
-@host_agent_bp.get("/<host_id>/")
-@require_role(UserRole.viewer)
-async def host_detail(host_id: str):
-    """Netdata-style per-host detail: current gauges + history charts."""
-    since, range_key = parse_range(request.args.get("range"))
-    async with current_app.db_sessionmaker() as session:
-        host = (await session.execute(
-            select(Host).where(Host.id == host_id))).scalar_one_or_none()
-        if host is None:
-            return _error(404, "not_found")
-        reg = (await session.execute(select(HostAgentRegistration).where(
-            HostAgentRegistration.host_id == host_id))).scalar_one_or_none()
-        latest = await _latest_metrics_for_host(session, host.name)
-        series = await _history_for_host(session, host.name, since)
-
-    hostlvl = latest.get(host.name, {})
+def _split_host_metrics(host_name: str, latest: dict) -> tuple:
+    """Split a host's latest snapshot into
+    (hostlvl, cores, nets, disks_io, temps, mounts) for the current-state fragment."""
+    hostlvl = latest.get(host_name, {})
     cores: list = []
     nets: dict = {}
     disks_io: dict = {}
     temps: dict = {}
     mounts: dict = {}
-    plen = len(host.name) + 1
+    plen = len(host_name) + 1
     for res, metrics in latest.items():
-        if res == host.name:
+        if res == host_name:
             continue
         suffix = res[plen:]
         if suffix.startswith("core"):
@@ -640,35 +581,80 @@ async def host_detail(host_id: str):
         else:
             mounts[suffix] = metrics
     cores.sort(key=lambda c: c[0])
+    return hostlvl, cores, nets, disks_io, temps, mounts
 
-    ls = reg.last_seen_at if reg else None
-    stale = ls is None or (datetime.now(timezone.utc) - ls).total_seconds() > _stale_after_seconds()
 
+@host_agent_bp.get("/<host_id>/")
+@require_role(UserRole.viewer)
+async def host_detail(host_id: str):
+    """Full-metrics shell: header + range toggle only. The current-state and
+    history sections stream in via HTMX (host_detail_metrics / host_detail_charts)
+    so the heavy history query never blocks first paint and the data live-updates."""
+    _, range_key = parse_range(request.args.get("range"))
+    async with current_app.db_sessionmaker() as session:
+        host = (await session.execute(
+            select(Host).where(Host.id == host_id))).scalar_one_or_none()
+        if host is None:
+            return _error(404, "not_found")
     return await render_template(
         "host_detail.html",
+        host=host, current_range=range_key, range_options=RANGE_OPTIONS,
+        poll_seconds=_DETAIL_POLL_SECONDS, chart_poll_seconds=_CHART_POLL_SECONDS,
+    )
+
+
+@host_agent_bp.get("/<host_id>/metrics")
+@require_role(UserRole.viewer)
+async def host_detail_metrics(host_id: str):
+    """Current-state fragment (gauges/cores/filesystems/IO/temps) — polled live so
+    the page shows the latest snapshot the server holds without a full reload."""
+    async with current_app.db_sessionmaker() as session:
+        host = (await session.execute(
+            select(Host).where(Host.id == host_id))).scalar_one_or_none()
+        if host is None:
+            return "", 404
+        reg = (await session.execute(select(HostAgentRegistration).where(
+            HostAgentRegistration.host_id == host_id))).scalar_one_or_none()
+        latest = await _latest_metrics_for_host(session, host.name)
+    hostlvl, cores, nets, disks_io, temps, mounts = _split_host_metrics(host.name, latest)
+    ls = reg.last_seen_at if reg else None
+    stale = ls is None or (datetime.now(timezone.utc) - ls).total_seconds() > _stale_after_seconds()
+    return await render_template(
+        "_host_metrics.html",
         host=host, reg=reg, stale=stale, hostlvl=hostlvl,
         cores=cores, nets=nets, disks_io=disks_io, temps=temps, mounts=mounts,
-        series=series, range_key=range_key, range_options=RANGE_OPTIONS,
     )
 
 
+@host_agent_bp.get("/<host_id>/charts")
+@require_role(UserRole.viewer)
+async def host_detail_charts(host_id: str):
+    """History-charts fragment — lazy-loaded so the date_bin query never blocks
+    the page shell; refreshed on a slow cadence and on range change."""
+    since, range_key = parse_range(request.args.get("range"))
+    async with current_app.db_sessionmaker() as session:
+        host = (await session.execute(
+            select(Host).where(Host.id == host_id))).scalar_one_or_none()
+        if host is None:
+            return "", 404
+        raw_days = int(await get_setting(session, "metrics.retention.raw_days") or 7)
+        series = await _history_for_host(session, host.name, since, raw_days=raw_days)
+    return await render_template("_host_charts.html", series=series, range_key=range_key)
+
+
 def _new_token_pair() -> tuple[str, str]:
     raw = secrets.token_urlsafe(32)
     return raw, _hash_token(raw)
 
 
-@host_agent_bp.get("/panel/<host_id>")
+@host_agent_bp.get("/vitals/<host_id>")
 @require_role(UserRole.viewer)
-async def host_panel(host_id: str):
-    """Per-host agent panel embedded into the core Hosts hub via HTMX.
+async def host_vitals(host_id: str):
+    """Compact live vitals strip (CPU/MEM/DISK/LOAD + pressure) for the Hosts hub.
 
-    Shows live agent metrics if the host reports, otherwise the provisioning
-    actions — tied to the host's linked Ansible target (the SSH connection).
-    Self-contained fragment (no base.html) so it can be hx-swapped in.
+    Self-contained fragment, polled by hosts/detail.html. Renders nothing until
+    the agent reports — the Agent panel below then carries provisioning.
     """
-    from steward.core.capabilities import has_capability
-    from steward.models.ansible_inventory import AnsibleTarget
-
     async with current_app.db_sessionmaker() as db:
         host = (await db.execute(
             select(Host).where(Host.id == host_id))).scalar_one_or_none()
@@ -677,8 +663,6 @@ async def host_panel(host_id: str):
         reg = (await db.execute(select(HostAgentRegistration).where(
             HostAgentRegistration.host_id == host_id))).scalar_one_or_none()
         latest = await _latest_metrics_for_host(db, host.name) if reg else {}
-        target = (await db.execute(select(AnsibleTarget).where(
-            AnsibleTarget.host_id == host_id))).scalar_one_or_none()
 
         # Recent series for the at-a-glance sparklines (cpu/mem/load host-level,
         # disk from the root mount sub-resource).
@@ -721,6 +705,40 @@ async def host_panel(host_id: str):
         "mem": hostlvl.get("psi_mem_some_avg10"),
         "io": hostlvl.get("psi_io_some_avg10"),
     }
+    ls = reg.last_seen_at if reg else None
+    reporting = bool(reg) and ls is not None
+    stale = reporting and (datetime.now(timezone.utc) - ls).total_seconds() > _stale_after_seconds()
+    return await render_template(
+        "_host_vitals.html",
+        reg=reg, cpu=hostlvl.get("cpu_pct"), mem=hostlvl.get("mem_used_pct"),
+        disk_root=disk_root, load_per_core=load_per_core, sparks=sparks, psi=psi,
+        reporting=reporting, stale=stale,
+    )
+
+
+@host_agent_bp.get("/panel/<host_id>")
+@require_role(UserRole.viewer)
+async def host_panel(host_id: str):
+    """Per-host agent management panel embedded into the Hosts hub via HTMX.
+
+    Lifecycle actions (update / rotate / remove / re-provision) when the host
+    reports, otherwise provisioning — tied to the host's linked Ansible target.
+    The live vitals are a separate strip (host_vitals); this panel is management
+    only. Self-contained fragment (no base.html) so it can be hx-swapped in.
+    """
+    from steward.core.capabilities import has_capability
+    from steward.models.ansible_inventory import AnsibleTarget
+
+    async with current_app.db_sessionmaker() as db:
+        host = (await db.execute(
+            select(Host).where(Host.id == host_id))).scalar_one_or_none()
+        if host is None:
+            return "", 404
+        reg = (await db.execute(select(HostAgentRegistration).where(
+            HostAgentRegistration.host_id == host_id))).scalar_one_or_none()
+        target = (await db.execute(select(AnsibleTarget).where(
+            AnsibleTarget.host_id == host_id))).scalar_one_or_none()
+
     ls = reg.last_seen_at if reg else None
     # "Deployed" is gated on the agent actually checking in (last_seen_at), NOT on
     # the registration row — that row is minted before the playbook runs, so a
@@ -731,9 +749,7 @@ async def host_panel(host_id: str):
     ansible_cfg = current_app.config.get("ANSIBLE", {})
     return await render_template(
         "panel.html",
-        host=host, reg=reg, hostlvl=hostlvl, disk_root=disk_root,
-        sparks=sparks, cores=cores, load1=load1, load_per_core=load_per_core, psi=psi,
-        reporting=reporting, stale=stale, target=target,
+        host=host, reg=reg, reporting=reporting, stale=stale, target=target,
         ansible_available=has_capability("ansible.run_playbook"),
         managed_key_set=bool((ansible_cfg.get("ssh_public_key") or "").strip()),
     )

plugins/host_agent/templates/_host_charts.html

@@ -0,0 +1,6 @@
+{# History data-only fragment. Swapped into the hidden #hm-charts-data poller;
+   the script runs and feeds the persistent charts created by host_detail.html,
+   updating them in place (no canvas swap → no flicker / no re-animation). #}
+<script>
+  if (window.applyHostSeries) window.applyHostSeries({{ series|tojson }}, {{ range_key|tojson }});
+</script>

plugins/host_agent/templates/_host_metrics.html

@@ -0,0 +1,182 @@
+{# Current-state fragment for the full-metrics page — polled live via HTMX.
+   Self-contained (defines its own format macros) so each poll re-renders the
+   latest snapshot the server has. #}
+{% macro fmt_bps(v) %}
+{%- if v is none -%}—
+{%- elif v >= 1048576 -%}{{ "%.1f"|format(v / 1048576) }} MB/s
+{%- elif v >= 1024 -%}{{ "%.1f"|format(v / 1024) }} KB/s
+{%- else -%}{{ "%.0f"|format(v) }} B/s
+{%- endif -%}
+{% endmacro %}
+{% macro fmt_bytes(v) %}
+{%- if v is none -%}—
+{%- elif v >= 1125899906842624 -%}{{ "%.1f"|format(v / 1125899906842624) }} PB
+{%- elif v >= 1099511627776 -%}{{ "%.1f"|format(v / 1099511627776) }} TB
+{%- elif v >= 1073741824 -%}{{ "%.1f"|format(v / 1073741824) }} GB
+{%- elif v >= 1048576 -%}{{ "%.0f"|format(v / 1048576) }} MB
+{%- else -%}{{ "%.0f"|format(v / 1024) }} KB
+{%- endif -%}
+{% endmacro %}
+{% macro bar(pct) %}
+{%- set p = pct if pct is not none else 0 -%}
+<div style="height:8px;background:var(--bg-elevated);border-radius:4px;overflow:hidden;border:1px solid var(--border);">
+  <div style="height:100%;width:{{ p|round(1) }}%;background:{% if pct is none %}var(--text-dim){% elif pct < 70 %}var(--green){% elif pct < 90 %}var(--yellow){% else %}var(--red){% endif %};"></div>
+</div>
+{% endmacro %}
+
+{# ── Identity / metadata ─────────────────────────────────────────────────── #}
+<div class="card" style="display:flex;flex-wrap:wrap;gap:0.5rem 1.25rem;align-items:center;font-size:0.82rem;color:var(--text-muted);">
+  <span>{% if stale %}<span class="badge badge-red">stale</span>{% else %}<span class="badge badge-green">live</span>{% endif %}</span>
+  <span><span style="color:var(--text-dim);">Address</span> {{ host.address or "—" }}</span>
+  {% if reg %}
+  {% if reg.distro %}<span><span style="color:var(--text-dim);">OS</span> {{ reg.distro }}</span>{% endif %}
+  {% if reg.kernel %}<span><span style="color:var(--text-dim);">Kernel</span> {{ reg.kernel }}</span>{% endif %}
+  {% if reg.arch %}<span><span style="color:var(--text-dim);">Arch</span> {{ reg.arch }}</span>{% endif %}
+  {% if reg.agent_version %}<span><span style="color:var(--text-dim);">Agent</span> v{{ reg.agent_version }}</span>{% endif %}
+  {% set up = hostlvl.get('uptime_secs') %}
+  {% if up %}<span><span style="color:var(--text-dim);">Uptime</span> {{ (up // 86400)|int }}d {{ ((up % 86400) // 3600)|int }}h</span>{% endif %}
+  <span><span style="color:var(--text-dim);">Last seen</span> {{ reg.last_seen_at.strftime("%Y-%m-%d %H:%M:%S") ~ " UTC" if reg.last_seen_at else "never" }}</span>
+  {% else %}
+  <span>No agent registration found for this host.</span>
+  {% endif %}
+</div>
+
+{% if not hostlvl %}
+<div class="card empty">No metrics reported yet. Once the agent checks in, CPU, memory, network, disk I/O, temperatures, and pressure will appear here.</div>
+{% else %}
+
+{# ── Current headline gauges ─────────────────────────────────────────────── #}
+<div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(150px,1fr));gap:0.75rem;margin-bottom:1rem;">
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.35rem;">CPU</div>
+    {% set c = hostlvl.get('cpu_pct') %}
+    <span class="stat-val" style="font-size:1.6rem;color:{% if c is none %}var(--text-dim){% elif c < 70 %}var(--green){% elif c < 90 %}var(--yellow){% else %}var(--red){% endif %};">{% if c is not none %}{{ "%.0f"|format(c) }}%{% else %}—{% endif %}</span>
+    <div style="margin-top:0.5rem;">{{ bar(c) }}</div>
+  </div>
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.35rem;">Memory</div>
+    {% set mp = hostlvl.get('mem_used_pct') %}
+    <span class="stat-val" style="font-size:1.6rem;color:{% if mp is none %}var(--text-dim){% elif mp < 70 %}var(--green){% elif mp < 90 %}var(--yellow){% else %}var(--red){% endif %};">{% if mp is not none %}{{ "%.0f"|format(mp) }}%{% else %}—{% endif %}</span>
+    <div style="margin-top:0.5rem;">{{ bar(mp) }}</div>
+    <div style="margin-top:0.45rem;font-size:0.72rem;color:var(--text-dim);">
+      avail {{ fmt_bytes(hostlvl.get('mem_available_bytes')) }} · cache {{ fmt_bytes(hostlvl.get('mem_cached_bytes')) }} · swap {{ fmt_bytes(hostlvl.get('swap_used_bytes')) }}
+    </div>
+  </div>
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.35rem;">Load</div>
+    <span class="stat-val" style="font-size:1.6rem;">{% if hostlvl.get('load_1m') is not none %}{{ "%.2f"|format(hostlvl.get('load_1m')) }}{% else %}—{% endif %}</span>
+    <div style="margin-top:0.45rem;font-size:0.72rem;color:var(--text-dim);">
+      5m {% if hostlvl.get('load_5m') is not none %}{{ "%.2f"|format(hostlvl.get('load_5m')) }}{% else %}—{% endif %} ·
+      15m {% if hostlvl.get('load_15m') is not none %}{{ "%.2f"|format(hostlvl.get('load_15m')) }}{% else %}—{% endif %}
+    </div>
+  </div>
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.35rem;">Network</div>
+    <div style="font-size:0.95rem;font-variant-numeric:tabular-nums;">
+      <div><span style="color:var(--green);">↓</span> {{ fmt_bps(hostlvl.get('net_rx_bps')) }}</div>
+      <div><span style="color:var(--red);">↑</span> {{ fmt_bps(hostlvl.get('net_tx_bps')) }}</div>
+    </div>
+  </div>
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.35rem;">Disk I/O</div>
+    <div style="font-size:0.95rem;font-variant-numeric:tabular-nums;">
+      <div><span style="color:var(--text-dim);">rd</span> {{ fmt_bps(hostlvl.get('disk_read_bps')) }}</div>
+      <div><span style="color:var(--text-dim);">wr</span> {{ fmt_bps(hostlvl.get('disk_write_bps')) }}</div>
+    </div>
+  </div>
+  {% if hostlvl.get('temp_c_max') is not none %}
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.35rem;">Temp (max)</div>
+    {% set t = hostlvl.get('temp_c_max') %}
+    <span class="stat-val" style="font-size:1.6rem;color:{% if t < 70 %}var(--green){% elif t < 85 %}var(--yellow){% else %}var(--red){% endif %};">{{ "%.0f"|format(t) }}°C</span>
+  </div>
+  {% endif %}
+  {% if hostlvl.get('psi_mem_some_avg10') is not none %}
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.35rem;" title="Pressure stall — % of time stalled, 10s avg">Pressure (10s)</div>
+    <div style="font-size:0.78rem;font-variant-numeric:tabular-nums;color:var(--text-muted);">
+      <div>mem {{ "%.1f"|format(hostlvl.get('psi_mem_some_avg10')) }}%</div>
+      {% if hostlvl.get('psi_cpu_some_avg10') is not none %}<div>cpu {{ "%.1f"|format(hostlvl.get('psi_cpu_some_avg10')) }}%</div>{% endif %}
+      {% if hostlvl.get('psi_io_some_avg10') is not none %}<div>io {{ "%.1f"|format(hostlvl.get('psi_io_some_avg10')) }}%</div>{% endif %}
+    </div>
+  </div>
+  {% endif %}
+</div>
+
+{# ── Per-core CPU ─────────────────────────────────────────────────────────── #}
+{% if cores %}
+<div class="card">
+  <div class="section-title" style="margin-bottom:0.6rem;">Per-core CPU</div>
+  <div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(120px,1fr));gap:0.5rem 0.9rem;">
+    {% for idx, pct in cores %}
+    <div>
+      <div style="display:flex;justify-content:space-between;font-size:0.72rem;color:var(--text-dim);margin-bottom:0.2rem;">
+        <span>core {{ idx }}</span><span>{% if pct is not none %}{{ "%.0f"|format(pct) }}%{% else %}—{% endif %}</span>
+      </div>
+      {{ bar(pct) }}
+    </div>
+    {% endfor %}
+  </div>
+</div>
+{% endif %}
+
+{# ── Filesystems ──────────────────────────────────────────────────────────── #}
+{% if mounts %}
+<div class="card">
+  <div class="section-title" style="margin-bottom:0.6rem;">Filesystems</div>
+  {% for mount, m in mounts.items() %}
+  <div style="margin-bottom:0.6rem;">
+    <div style="display:flex;justify-content:space-between;font-size:0.8rem;margin-bottom:0.2rem;">
+      <span style="font-family:ui-monospace,monospace;">{{ mount }}</span>
+      <span style="color:var(--text-muted);">{{ fmt_bytes(m.get('disk_used_bytes')) }} / {{ fmt_bytes(m.get('disk_total_bytes')) }}{% if m.get('disk_used_pct') is not none %} · {{ "%.0f"|format(m.get('disk_used_pct')) }}%{% endif %}</span>
+    </div>
+    {{ bar(m.get('disk_used_pct')) }}
+  </div>
+  {% endfor %}
+</div>
+{% endif %}
+
+{# ── Interfaces / disks (short panels, side by side at natural height) ─────── #}
+{% if nets or disks_io %}
+<div style="display:grid;grid-template-columns:repeat(auto-fit,minmax(260px,1fr));gap:1rem;align-items:start;">
+  {% if nets %}
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.5rem;">Interfaces</div>
+    {% for iface, m in nets.items() %}
+    <div style="display:flex;justify-content:space-between;gap:0.75rem;font-size:0.8rem;padding:0.15rem 0;">
+      <span style="font-family:ui-monospace,monospace;">{{ iface }}</span>
+      <span style="color:var(--text-muted);font-variant-numeric:tabular-nums;white-space:nowrap;">↓ {{ fmt_bps(m.get('net_rx_bps')) }} · ↑ {{ fmt_bps(m.get('net_tx_bps')) }}</span>
+    </div>
+    {% endfor %}
+  </div>
+  {% endif %}
+  {% if disks_io %}
+  <div class="card" style="margin-bottom:0;">
+    <div class="section-title" style="margin-bottom:0.5rem;">Disks</div>
+    {% for dev, m in disks_io.items() %}
+    <div style="display:flex;justify-content:space-between;gap:0.75rem;font-size:0.8rem;padding:0.15rem 0;">
+      <span style="font-family:ui-monospace,monospace;">{{ dev }}</span>
+      <span style="color:var(--text-muted);font-variant-numeric:tabular-nums;white-space:nowrap;">rd {{ fmt_bps(m.get('disk_read_bps')) }} · wr {{ fmt_bps(m.get('disk_write_bps')) }}</span>
+    </div>
+    {% endfor %}
+  </div>
+  {% endif %}
+</div>
+{% endif %}
+
+{# ── Temperatures (full width; cores flow into a compact multi-column grid) ── #}
+{% if temps %}
+<div class="card" style="margin-top:1rem;margin-bottom:0;">
+  <div class="section-title" style="margin-bottom:0.5rem;">Temperatures</div>
+  <div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(130px,1fr));gap:0.1rem 1.25rem;">
+    {% for label, c in temps.items() %}
+    <div style="display:flex;justify-content:space-between;gap:0.5rem;font-size:0.8rem;padding:0.15rem 0;">
+      <span style="white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">{{ label }}</span>
+      <span style="color:{% if c is none %}var(--text-dim){% elif c < 70 %}var(--green){% elif c < 85 %}var(--yellow){% else %}var(--red){% endif %};font-variant-numeric:tabular-nums;">{% if c is not none %}{{ "%.0f"|format(c) }}°C{% else %}—{% endif %}</span>
+    </div>
+    {% endfor %}
+  </div>
+</div>
+{% endif %}
+
+{% endif %}

plugins/host_agent/templates/_host_vitals.html

@@ -0,0 +1,35 @@
+{# Host vitals strip — compact CPU/MEM/DISK/LOAD + pressure, live-polled into the
+   host hub. Renders nothing until the agent reports (the Agent panel below then
+   shows provisioning). #}
+{% if reporting %}
+{% set lbl = "font-size:0.68rem;color:var(--text-muted);text-transform:uppercase;letter-spacing:0.04em;" %}
+{% macro vital(label, value, text, kind, spark) %}
+<div style="min-width:88px;">
+  <div style="{{ lbl }}">{{ label }}</div>
+  <div style="font-size:1.5rem;font-weight:700;line-height:1.1;{{ threshold_style(value, kind) }}">{{ text }}</div>
+  <div style="margin-top:0.2rem;">{{ spark | safe }}</div>
+</div>
+{% endmacro %}
+<div class="card" style="display:flex;flex-wrap:wrap;align-items:flex-start;gap:1rem 2rem;margin-bottom:1rem;">
+  {{ vital("CPU", cpu, ('%.0f%%'|format(cpu) if cpu is not none else '—'), 'cpu', sparks.cpu) }}
+  {{ vital("Memory", mem, ('%.0f%%'|format(mem) if mem is not none else '—'), 'mem', sparks.mem) }}
+  {{ vital("Disk /", disk_root, ('%.0f%%'|format(disk_root) if disk_root is not none else '—'), 'disk', sparks.disk) }}
+  {{ vital("Load /core", load_per_core, ('%d%%'|format(load_per_core) if load_per_core is not none else '—'), 'load', sparks.load) }}
+
+  <div style="margin-left:auto;display:flex;flex-direction:column;align-items:flex-end;gap:0.3rem;font-size:0.75rem;color:var(--text-muted);text-align:right;">
+    <span>
+      {% if stale %}<span class="badge badge-red">stale</span>{% else %}<span class="badge badge-green">live</span>{% endif %}
+      {% if reg and reg.agent_version %}<span style="margin-left:0.3rem;">v{{ reg.agent_version }}</span>{% endif %}
+    </span>
+    {% if psi.cpu is not none or psi.mem is not none or psi.io is not none %}
+    <span title="Pressure stall — % of the last 10s tasks waited for the resource">
+      <span style="color:var(--text-dim);text-transform:uppercase;font-size:0.66rem;letter-spacing:0.04em;">Pressure 10s</span>
+      cpu {{ '%.0f%%'|format(psi.cpu) if psi.cpu is not none else '—' }}
+      · mem {{ '%.0f%%'|format(psi.mem) if psi.mem is not none else '—' }}
+      · io {{ '%.0f%%'|format(psi.io) if psi.io is not none else '—' }}
+    </span>
+    {% endif %}
+    {% if reg and reg.last_seen_at %}<span>seen {{ reg.last_seen_at.strftime("%H:%M:%S") }} UTC</span>{% endif %}
+  </div>
+</div>
+{% endif %}

plugins/host_agent/templates/host_detail.html

@@ -1,215 +1,56 @@
 {% extends "base.html" %}
 {% from "_macros.html" import crumbs %}
-{% block title %}{{ host.name }} — Host Agent — Steward{% endblock %}
+{% block title %}{{ host.name }} — Metrics — Steward{% endblock %}
 {% block breadcrumb %}{{ crumbs([("Hosts", "/hosts/"), (host.name, "/hosts/" ~ host.id), ("Metrics", "")]) }}{% endblock %}
-
-{% macro fmt_bps(v) %}
-{%- if v is none -%}—
-{%- elif v >= 1048576 -%}{{ "%.1f"|format(v / 1048576) }} MB/s
-{%- elif v >= 1024 -%}{{ "%.1f"|format(v / 1024) }} KB/s
-{%- else -%}{{ "%.0f"|format(v) }} B/s
-{%- endif -%}
-{% endmacro %}
-{% macro fmt_bytes(v) %}
-{%- if v is none -%}—
-{%- elif v >= 1073741824 -%}{{ "%.1f"|format(v / 1073741824) }} GB
-{%- elif v >= 1048576 -%}{{ "%.0f"|format(v / 1048576) }} MB
-{%- else -%}{{ "%.0f"|format(v / 1024) }} KB
-{%- endif -%}
-{% endmacro %}
-{% macro bar(pct) %}
-{%- set p = pct if pct is not none else 0 -%}
-<div style="height:8px;background:var(--bg-elevated);border-radius:4px;overflow:hidden;border:1px solid var(--border);">
-  <div style="height:100%;width:{{ p|round(1) }}%;background:{% if pct is none %}var(--text-dim){% elif pct < 70 %}var(--green){% elif pct < 90 %}var(--yellow){% else %}var(--red){% endif %};"></div>
-</div>
-{% endmacro %}
-
 {% block content %}
+{# Shell only: current-state + history data stream in via HTMX so the heavy
+   history query never blocks first paint and the data refreshes live (≈ the
+   agent's push cadence). The chart CANVASES live here permanently — only their
+   data is polled and applied in place, so refreshes never re-create/flicker. #}
 <div style="display:flex;align-items:center;justify-content:space-between;gap:1rem;flex-wrap:wrap;margin-bottom:0.75rem;">
-  <div style="display:flex;align-items:center;gap:0.6rem;">
-    <a href="/hosts/{{ host.id }}" class="btn btn-ghost btn-sm">← Host</a>
-    <h1 class="page-title" style="margin-bottom:0;">{{ host.name }}</h1>
-    {% if stale %}<span class="badge badge-red">stale</span>{% else %}<span class="badge badge-green">live</span>{% endif %}
-  </div>
-  <div style="display:flex;gap:0.3rem;">
-    {% for r in range_options %}
-    <a class="range-btn {% if r == range_key %}active{% endif %}" href="?range={{ r }}">{{ r }}</a>
-    {% endfor %}
-  </div>
+  <h1 class="page-title" style="margin-bottom:0;">{{ host.name }}</h1>
+  {% include "_time_range.html" %}
 </div>
 
-{# ── Identity / metadata ─────────────────────────────────────────────────── #}
-<div class="card" style="display:flex;flex-wrap:wrap;gap:0.5rem 1.25rem;align-items:center;font-size:0.82rem;color:var(--text-muted);">
-  <span><span style="color:var(--text-dim);">Address</span> {{ host.address or "—" }}</span>
-  {% if reg %}
-  {% if reg.distro %}<span><span style="color:var(--text-dim);">OS</span> {{ reg.distro }}</span>{% endif %}
-  {% if reg.kernel %}<span><span style="color:var(--text-dim);">Kernel</span> {{ reg.kernel }}</span>{% endif %}
-  {% if reg.arch %}<span><span style="color:var(--text-dim);">Arch</span> {{ reg.arch }}</span>{% endif %}
-  {% if reg.agent_version %}<span><span style="color:var(--text-dim);">Agent</span> v{{ reg.agent_version }}</span>{% endif %}
-  {% set up = hostlvl.get('uptime_secs') %}
-  {% if up %}<span><span style="color:var(--text-dim);">Uptime</span> {{ (up // 86400)|int }}d {{ ((up % 86400) // 3600)|int }}h</span>{% endif %}
-  <span><span style="color:var(--text-dim);">Last seen</span> {{ reg.last_seen_at.strftime("%Y-%m-%d %H:%M:%S") ~ " UTC" if reg.last_seen_at else "never" }}</span>
-  {% else %}
-  <span>No agent registration found for this host.</span>
-  {% endif %}
+{# Current state — lazy + polled live. Atomic innerHTML swap of fixed-height
+   cards, so values update without a layout jump. #}
+<div id="hm-current"
+     hx-get="/plugins/host_agent/{{ host.id }}/metrics"
+     hx-trigger="load, every {{ poll_seconds }}s"
+     hx-swap="innerHTML">
+  <div class="card empty">Loading metrics…</div>
 </div>
 
-{% if not hostlvl %}
-<div class="card empty">No metrics reported yet. Once the agent checks in, CPU, memory, network, disk I/O, temperatures, and pressure will appear here.</div>
-{% else %}
-
-{# ── Current headline gauges ─────────────────────────────────────────────── #}
-<div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(150px,1fr));gap:0.75rem;margin-bottom:1rem;">
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">CPU</div>
-    {% set c = hostlvl.get('cpu_pct') %}
-    <span class="stat-val" style="font-size:1.6rem;color:{% if c is none %}var(--text-dim){% elif c < 70 %}var(--green){% elif c < 90 %}var(--yellow){% else %}var(--red){% endif %};">{% if c is not none %}{{ "%.0f"|format(c) }}%{% else %}—{% endif %}</span>
-    <div style="margin-top:0.5rem;">{{ bar(c) }}</div>
-  </div>
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Memory</div>
-    {% set mp = hostlvl.get('mem_used_pct') %}
-    <span class="stat-val" style="font-size:1.6rem;color:{% if mp is none %}var(--text-dim){% elif mp < 70 %}var(--green){% elif mp < 90 %}var(--yellow){% else %}var(--red){% endif %};">{% if mp is not none %}{{ "%.0f"|format(mp) }}%{% else %}—{% endif %}</span>
-    <div style="margin-top:0.5rem;">{{ bar(mp) }}</div>
-    <div style="margin-top:0.45rem;font-size:0.72rem;color:var(--text-dim);">
-      avail {{ fmt_bytes(hostlvl.get('mem_available_bytes')) }} · cache {{ fmt_bytes(hostlvl.get('mem_cached_bytes')) }} · swap {{ fmt_bytes(hostlvl.get('swap_used_bytes')) }}
-    </div>
-  </div>
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Load</div>
-    <span class="stat-val" style="font-size:1.6rem;">{% if hostlvl.get('load_1m') is not none %}{{ "%.2f"|format(hostlvl.get('load_1m')) }}{% else %}—{% endif %}</span>
-    <div style="margin-top:0.45rem;font-size:0.72rem;color:var(--text-dim);">
-      5m {% if hostlvl.get('load_5m') is not none %}{{ "%.2f"|format(hostlvl.get('load_5m')) }}{% else %}—{% endif %} ·
-      15m {% if hostlvl.get('load_15m') is not none %}{{ "%.2f"|format(hostlvl.get('load_15m')) }}{% else %}—{% endif %}
-    </div>
-  </div>
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Network</div>
-    <div style="font-size:0.95rem;font-variant-numeric:tabular-nums;">
-      <div><span style="color:var(--green);">↓</span> {{ fmt_bps(hostlvl.get('net_rx_bps')) }}</div>
-      <div><span style="color:var(--red);">↑</span> {{ fmt_bps(hostlvl.get('net_tx_bps')) }}</div>
-    </div>
-  </div>
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Disk I/O</div>
-    <div style="font-size:0.95rem;font-variant-numeric:tabular-nums;">
-      <div><span style="color:var(--text-dim);">rd</span> {{ fmt_bps(hostlvl.get('disk_read_bps')) }}</div>
-      <div><span style="color:var(--text-dim);">wr</span> {{ fmt_bps(hostlvl.get('disk_write_bps')) }}</div>
-    </div>
-  </div>
-  {% if hostlvl.get('temp_c_max') is not none %}
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Temp (max)</div>
-    {% set t = hostlvl.get('temp_c_max') %}
-    <span class="stat-val" style="font-size:1.6rem;color:{% if t < 70 %}var(--green){% elif t < 85 %}var(--yellow){% else %}var(--red){% endif %};">{{ "%.0f"|format(t) }}°C</span>
-  </div>
-  {% endif %}
-  {% if hostlvl.get('psi_mem_some_avg10') is not none %}
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;" title="Pressure stall — % of time stalled, 10s avg">Pressure (10s)</div>
-    <div style="font-size:0.78rem;font-variant-numeric:tabular-nums;color:var(--text-muted);">
-      <div>mem {{ "%.1f"|format(hostlvl.get('psi_mem_some_avg10')) }}%</div>
-      {% if hostlvl.get('psi_cpu_some_avg10') is not none %}<div>cpu {{ "%.1f"|format(hostlvl.get('psi_cpu_some_avg10')) }}%</div>{% endif %}
-      {% if hostlvl.get('psi_io_some_avg10') is not none %}<div>io {{ "%.1f"|format(hostlvl.get('psi_io_some_avg10')) }}%</div>{% endif %}
-    </div>
-  </div>
-  {% endif %}
-</div>
-
-{# ── Per-core CPU ─────────────────────────────────────────────────────────── #}
-{% if cores %}
-<div class="card">
-  <div class="section-title" style="margin-bottom:0.6rem;">Per-core CPU</div>
-  <div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(120px,1fr));gap:0.5rem 0.9rem;">
-    {% for idx, pct in cores %}
-    <div>
-      <div style="display:flex;justify-content:space-between;font-size:0.72rem;color:var(--text-dim);margin-bottom:0.2rem;">
-        <span>core {{ idx }}</span><span>{% if pct is not none %}{{ "%.0f"|format(pct) }}%{% else %}—{% endif %}</span>
-      </div>
-      {{ bar(pct) }}
-    </div>
-    {% endfor %}
-  </div>
-</div>
-{% endif %}
-
-{# ── Filesystems ──────────────────────────────────────────────────────────── #}
-{% if mounts %}
-<div class="card">
-  <div class="section-title" style="margin-bottom:0.6rem;">Filesystems</div>
-  {% for mount, m in mounts.items() %}
-  <div style="margin-bottom:0.6rem;">
-    <div style="display:flex;justify-content:space-between;font-size:0.8rem;margin-bottom:0.2rem;">
-      <span style="font-family:ui-monospace,monospace;">{{ mount }}</span>
-      <span style="color:var(--text-muted);">{{ fmt_bytes(m.get('disk_used_bytes')) }} / {{ fmt_bytes(m.get('disk_total_bytes')) }}{% if m.get('disk_used_pct') is not none %} · {{ "%.0f"|format(m.get('disk_used_pct')) }}%{% endif %}</span>
-    </div>
-    {{ bar(m.get('disk_used_pct')) }}
-  </div>
-  {% endfor %}
-</div>
-{% endif %}
-
-{# ── Interfaces / disks / sensors current detail ──────────────────────────── #}
-{% if nets or disks_io or temps %}
-<div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(240px,1fr));gap:1rem;">
-  {% if nets %}
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.5rem;">Interfaces</div>
-    {% for iface, m in nets.items() %}
-    <div style="display:flex;justify-content:space-between;font-size:0.8rem;padding:0.15rem 0;">
-      <span style="font-family:ui-monospace,monospace;">{{ iface }}</span>
-      <span style="color:var(--text-muted);font-variant-numeric:tabular-nums;">↓ {{ fmt_bps(m.get('net_rx_bps')) }} · ↑ {{ fmt_bps(m.get('net_tx_bps')) }}</span>
-    </div>
-    {% endfor %}
-  </div>
-  {% endif %}
-  {% if disks_io %}
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.5rem;">Disks</div>
-    {% for dev, m in disks_io.items() %}
-    <div style="display:flex;justify-content:space-between;font-size:0.8rem;padding:0.15rem 0;">
-      <span style="font-family:ui-monospace,monospace;">{{ dev }}</span>
-      <span style="color:var(--text-muted);font-variant-numeric:tabular-nums;">rd {{ fmt_bps(m.get('disk_read_bps')) }} · wr {{ fmt_bps(m.get('disk_write_bps')) }}</span>
-    </div>
-    {% endfor %}
-  </div>
-  {% endif %}
-  {% if temps %}
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.5rem;">Temperatures</div>
-    {% for label, c in temps.items() %}
-    <div style="display:flex;justify-content:space-between;font-size:0.8rem;padding:0.15rem 0;">
-      <span>{{ label }}</span>
-      <span style="color:{% if c is none %}var(--text-dim){% elif c < 70 %}var(--green){% elif c < 85 %}var(--yellow){% else %}var(--red){% endif %};font-variant-numeric:tabular-nums;">{% if c is not none %}{{ "%.0f"|format(c) }}°C{% else %}—{% endif %}</span>
-    </div>
-    {% endfor %}
-  </div>
-  {% endif %}
-</div>
-{% endif %}
-
-{# ── History charts ───────────────────────────────────────────────────────── #}
+{# History charts — persistent canvases (created once below); only data is polled. #}
 <div style="display:grid;grid-template-columns:repeat(auto-fit,minmax(340px,1fr));gap:1rem;margin-top:1rem;">
   <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.5rem;">Utilization % — last {{ range_key }}</div>
+    <div class="section-title" style="margin-bottom:0.5rem;">Utilization % — last <span class="hm-chart-range">{{ current_range }}</span></div>
     <div style="height:220px;"><canvas id="chart-util"></canvas></div>
   </div>
   <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.5rem;">Throughput (B/s) — last {{ range_key }}</div>
+    <div class="section-title" style="margin-bottom:0.5rem;">Throughput (B/s) — last <span class="hm-chart-range">{{ current_range }}</span></div>
     <div style="height:220px;"><canvas id="chart-net"></canvas></div>
   </div>
   <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.5rem;">Load &amp; pressure — last {{ range_key }}</div>
+    <div class="section-title" style="margin-bottom:0.5rem;">Load &amp; pressure — last <span class="hm-chart-range">{{ current_range }}</span></div>
     <div style="height:220px;"><canvas id="chart-pressure"></canvas></div>
   </div>
 </div>
+{# Hidden poller: fetches the history series and applies it to the charts above
+   (no visible swap). Lazy on load → never blocks paint; refreshed slowly + on
+   range change. #}
+<div id="hm-charts-data" style="display:none;"
+     hx-get="/plugins/host_agent/{{ host.id }}/charts"
+     hx-trigger="load, every {{ chart_poll_seconds }}s, rangeChange from:body"
+     hx-include="#time-range"
+     hx-swap="innerHTML"></div>
 
 <script>
 (function () {
-  const series = {{ series|tojson }};
   const fmtTime = (v) => { const d = new Date(v); return ("0" + d.getHours()).slice(-2) + ":" + ("0" + d.getMinutes()).slice(-2); };
   const baseOpts = (ymax) => ({
     responsive: true, maintainAspectRatio: false,
+    animation: false,                       // no grow-in / re-animate on refresh
     interaction: { mode: "index", intersect: false },
     elements: { point: { radius: 0 } },
     scales: {
@@ -218,39 +59,50 @@
     },
     plugins: {
       legend: { labels: { color: "#b8b8b0", boxWidth: 10, font: { size: 11 } } },
-      tooltip: { callbacks: { title: function (items) { return items.length ? fmtTime(items[0].parsed.x) : ""; } } },
+      tooltip: { callbacks: { title: (items) => items.length ? fmtTime(items[0].parsed.x) : "" } },
     },
   });
-  const mk = (id, datasets, ymax) => {
-    const el = document.getElementById(id);
-    if (!el) return;
-    new Chart(el.getContext("2d"), {
-      type: "line",
-      data: { datasets: datasets.map((d) => ({
-        label: d.label,
-        data: (series[d.key] || []).map((p) => ({ x: p[0], y: p[1] })),
-        borderColor: d.color, backgroundColor: d.color, borderWidth: 1.5, tension: 0.25,
-      })) },
-      options: baseOpts(ymax),
-    });
+  const defs = {
+    "chart-util": { ymax: 100, ds: [
+      { key: "cpu_pct", label: "CPU %", color: "#c8a840" },
+      { key: "mem_used_pct", label: "Mem %", color: "#4aa86a" },
+      { key: "disk_used_pct_worst", label: "Disk % (worst)", color: "#c87840" },
+    ] },
+    "chart-net": { ymax: undefined, ds: [
+      { key: "net_rx_bps", label: "Net RX", color: "#4aa86a" },
+      { key: "net_tx_bps", label: "Net TX", color: "#c84048" },
+      { key: "disk_read_bps", label: "Disk read", color: "#6aa0d0" },
+      { key: "disk_write_bps", label: "Disk write", color: "#c8a840" },
+    ] },
+    "chart-pressure": { ymax: undefined, ds: [
+      { key: "load_1m", label: "Load 1m", color: "#c8a840" },
+      { key: "psi_mem_some_avg10", label: "Mem PSI %", color: "#c84048" },
+      { key: "temp_c_max", label: "Temp °C", color: "#c87840" },
+    ] },
+  };
+  const charts = {};
+  for (const id of Object.keys(defs)) {
+    const el = document.getElementById(id);
+    if (!el) continue;
+    charts[id] = new Chart(el.getContext("2d"), {
+      type: "line",
+      data: { datasets: defs[id].ds.map((d) => ({ label: d.label, data: [], borderColor: d.color, backgroundColor: d.color, borderWidth: 1.5, tension: 0.25 })) },
+      options: baseOpts(defs[id].ymax),
+    });
+  }
+  // Called by the polled /charts data fragment. Updates data in place (no
+  // re-create, no animation), so refreshes never flicker or shift layout.
+  window.applyHostSeries = function (series, rangeKey) {
+    for (const id of Object.keys(defs)) {
+      const ch = charts[id];
+      if (!ch) continue;
+      defs[id].ds.forEach((d, i) => {
+        ch.data.datasets[i].data = (series[d.key] || []).map((p) => ({ x: p[0], y: p[1] }));
+      });
+      ch.update("none");
+    }
+    if (rangeKey) document.querySelectorAll(".hm-chart-range").forEach((s) => { s.textContent = rangeKey; });
   };
-  mk("chart-util", [
-    { key: "cpu_pct", label: "CPU %", color: "#c8a840" },
-    { key: "mem_used_pct", label: "Mem %", color: "#4aa86a" },
-    { key: "disk_used_pct_worst", label: "Disk % (worst)", color: "#c87840" },
-  ], 100);
-  mk("chart-net", [
-    { key: "net_rx_bps", label: "Net RX", color: "#4aa86a" },
-    { key: "net_tx_bps", label: "Net TX", color: "#c84048" },
-    { key: "disk_read_bps", label: "Disk read", color: "#6aa0d0" },
-    { key: "disk_write_bps", label: "Disk write", color: "#c8a840" },
-  ], undefined);
-  mk("chart-pressure", [
-    { key: "load_1m", label: "Load 1m", color: "#c8a840" },
-    { key: "psi_mem_some_avg10", label: "Mem PSI %", color: "#c84048" },
-    { key: "temp_c_max", label: "Temp °C", color: "#c87840" },
-  ], undefined);
 })();
 </script>
-{% endif %}
 {% endblock %}

plugins/host_agent/templates/panel.html

@@ -14,39 +14,8 @@
   </div>
 
   {% if reporting %}
-  {# ── Reporting: at-a-glance metrics (+ sparkline trend) + lifecycle ── #}
-  {% set cpu = hostlvl.get('cpu_pct') %}
-  {% set mem = hostlvl.get('mem_used_pct') %}
-  {% set lbl = "font-size:0.72rem;color:var(--text-muted);text-transform:uppercase;" %}
-  <div style="display:flex;gap:1.5rem;flex-wrap:wrap;margin-bottom:0.6rem;">
-    <div title="Average CPU utilization across all cores">
-      <div style="{{ lbl }}">CPU</div>
-      <div style="font-weight:600;{{ threshold_style(cpu, 'cpu') }}">{{ '%.0f%%'|format(cpu) if cpu is not none else '—' }}</div>
-      {{ sparks.cpu | safe }}</div>
-    <div title="RAM in use (total minus available; cache/buffers count as free)">
-      <div style="{{ lbl }}">Memory</div>
-      <div style="font-weight:600;{{ threshold_style(mem, 'mem') }}">{{ '%.0f%%'|format(mem) if mem is not none else '—' }}</div>
-      {{ sparks.mem | safe }}</div>
-    <div title="Root filesystem (/) usage — see Full metrics for every mount">
-      <div style="{{ lbl }}">Disk /</div>
-      <div style="font-weight:600;{{ threshold_style(disk_root, 'disk') }}">{{ '%.0f%%'|format(disk_root) if disk_root is not none else '—' }}</div>
-      {{ sparks.disk | safe }}</div>
-    {# Load /core: 100% = run queue matches CPU capacity, so warn 80 / crit 100. #}
-    <div title="1-minute load average ÷ {{ cores or '?' }} CPU cores (100% = run queue matches capacity). Raw 1m load: {{ '%.2f'|format(load1) if load1 is not none else '—' }}.">
-      <div style="{{ lbl }}">Load /core</div>
-      <div style="font-weight:600;{{ threshold_style(load_per_core, 'load') }}">{{ '%d%%'|format(load_per_core) if load_per_core is not none else ('%.2f'|format(load1) if load1 is not none else '—') }}</div>
-      {{ sparks.load | safe }}</div>
-  </div>
-  {% if psi.cpu is not none or psi.mem is not none or psi.io is not none %}
-  <div style="font-size:0.75rem;color:var(--text-muted);margin-bottom:0.9rem;"
-       title="Pressure Stall Information — % of the last 10s that tasks were stalled waiting for the resource. Hardware-independent, comparable across hosts.">
-    <span style="text-transform:uppercase;font-size:0.7rem;letter-spacing:0.04em;color:var(--text-dim);">Pressure 10s</span>
-    &nbsp;CPU {{ '%.0f%%'|format(psi.cpu) if psi.cpu is not none else '—' }}
-    · Mem {{ '%.0f%%'|format(psi.mem) if psi.mem is not none else '—' }}
-    · IO {{ '%.0f%%'|format(psi.io) if psi.io is not none else '—' }}
-  </div>
-  {% endif %}
-
+  {# Reporting: live vitals are shown by the vitals strip above; this panel is
+     lifecycle/management only. #}
   <div style="display:flex;gap:0.5rem;flex-wrap:wrap;align-items:center;">
     <a href="/plugins/host_agent/{{ host.id }}/" class="btn btn-sm btn-ghost">Full metrics →</a>
     {% if session.user_role == 'admin' %}

plugins/host_agent/templates/settings_list.html

@@ -5,7 +5,6 @@
 {% block content %}
 <div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:1rem;gap:1rem;flex-wrap:wrap;">
   <h1 class="page-title" style="margin-bottom:0;">Agent fleet</h1>
-  <a href="/hosts/" class="btn btn-ghost btn-sm">← Hosts</a>
 </div>
 <p style="color:var(--text-muted);font-size:0.84rem;margin:-0.4rem 0 1.25rem;">
   Bulk agent operations across your inventory. For a single host, use its

plugins/snmp/__init__.py

@@ -21,3 +21,8 @@ def get_scheduled_tasks() -> list:
 def get_blueprint():
     from .routes import snmp_bp
     return snmp_bp
+
+
+def get_nav() -> list[dict]:
+    # Surfaces in the sidebar's Infrastructure group (SNMP device readings).
+    return [{"label": "SNMP", "href": "/plugins/snmp/"}]

plugins/snmp/plugin.yaml

@@ -15,6 +15,16 @@ tags:
 config:
   poll_interval_seconds: 60
   # Each device is polled independently. OIDs must resolve to numeric values.
+  #
+  # `host` is the SNMP poll target (IP/hostname we query). A device also shows
+  # up on a Steward Host's detail page when it maps to that host. By default
+  # that mapping is implicit: `host` is matched against the Host's address or
+  # name. To bind explicitly instead — e.g. when you poll by IP but the Host is
+  # recorded by DNS name, or for an SNMP-only switch/PDU/UPS — add ONE of:
+  #   host_id:      "<steward-host-uuid>"   # exact Host.id match
+  #   steward_host: "switch01"              # Host name OR address (case-insensitive)
+  # An explicit binding is exclusive: the device then maps ONLY to that host,
+  # never implicitly to another by a coincidental address string.
   devices:
     - name: "core-switch"
       host: "192.168.1.1"

plugins/snmp/poller.py

@@ -1,11 +1,13 @@
 # plugins/snmp/poller.py
 """
-Synchronous SNMP GET helper, run via executor.
+Asynchronous SNMP GET helper.
 
-Requires pysnmp-lextudio (maintained pysnmp fork):
+Requires pysnmp-lextudio (the maintained pysnmp fork), bundled into the Docker
+image via the `snmp` extra (`pip install .[snmp]`):
     pip install 'steward[snmp]'
 
-If pysnmp is not installed, poll_device() returns an empty dict and logs a warning.
+If pysnmp is not installed, poll_device() returns an empty dict and logs a
+warning — SNMP polling is then simply disabled, nothing else breaks.
 """
 from __future__ import annotations
 import logging
@@ -22,73 +24,148 @@ def _pysnmp_available() -> bool:
 
 
 def _mp_model(version: str) -> int:
-    """Map version string to pysnmp mpModel integer."""
+    """Map an SNMP version string to pysnmp's mpModel int (0 = v1, 1 = v2c)."""
     return 0 if version == "1" else 1
 
 
-def poll_device_sync(
+def _close_engine(engine) -> None:
+    """Release the engine's UDP transport socket.
+
+    pysnmp opens a UDP socket per ``SnmpEngine`` and never closes it on its own.
+    Since we build a fresh engine for every poll, an unclosed engine leaks one
+    file descriptor each scheduler tick; over hours of polling that exhausts the
+    process fd limit (``OSError: [Errno 24] Too many open files``), which also
+    takes down the app's listening socket. So close it explicitly here.
+
+    The close method/attribute names differ across pysnmp majors (6.2.x lextudio
+    is camelCase, canonical 7.x is snake_case), so probe for whatever exists.
+    All paths are best-effort — a failed close must never break the poll loop.
+    """
+    # 7.x exposes a convenience close directly on the engine.
+    for meth in ("close_dispatcher", "closeDispatcher"):
+        fn = getattr(engine, meth, None)
+        if callable(fn):
+            try:
+                fn()
+            except Exception:
+                pass
+            return
+    # 6.2.x: go through the transport dispatcher.
+    for attr in ("transport_dispatcher", "transportDispatcher"):
+        dispatcher = getattr(engine, attr, None)
+        if dispatcher is None:
+            continue
+        for meth in ("close_dispatcher", "closeDispatcher"):
+            fn = getattr(dispatcher, meth, None)
+            if callable(fn):
+                try:
+                    fn()
+                except Exception:
+                    pass
+                return
+        return
+
+
+async def poll_device(
     host: str,
     port: int,
     community: str,
     version: str,
     oids: list[dict],
 ) -> dict[str, float]:
-    """
-    Perform SNMP GET for each OID and return {label: float_value}.
-    Non-numeric OIDs (strings, etc.) are skipped.
-    Returns empty dict on any error.
+    """Perform an SNMP GET for each OID and return ``{label: float_value}``.
+
+    Non-numeric OIDs (strings, etc.) are skipped. Returns an empty dict on any
+    error (unreachable host, wrong community, …) so a flaky device never breaks
+    the poll loop.
+
+    pysnmp's HLAPI is asyncio-only as of v6. The import path moved between major
+    versions, so we support both rather than let a dependency bump silently
+    re-break polling:
+      • pysnmp-lextudio 6.2.x → ``pysnmp.hlapi.asyncio`` (``getCmd`` + a directly
+        constructed ``UdpTransportTarget``).
+      • canonical pysnmp 7.x  → ``pysnmp.hlapi.v3arch.asyncio`` (``get_cmd`` + the
+        async ``UdpTransportTarget.create``).
     """
     if not _pysnmp_available():
         logger.warning("pysnmp not installed — SNMP polling disabled. "
                        "Install with: pip install 'steward[snmp]'")
         return {}
 
-    from pysnmp.hlapi import (
-        CommunityData,
-        ContextData,
-        ObjectIdentity,
-        ObjectType,
-        SnmpEngine,
-        UdpTransportTarget,
-        getCmd,
-    )
+    try:
+        # canonical pysnmp 7.x
+        from pysnmp.hlapi.v3arch.asyncio import (
+            CommunityData,
+            ContextData,
+            ObjectIdentity,
+            ObjectType,
+            SnmpEngine,
+            UdpTransportTarget,
+            get_cmd as _get_cmd,
+        )
+        _transport_is_async = True
+    except ImportError:
+        # pysnmp-lextudio 6.2.x
+        from pysnmp.hlapi.asyncio import (
+            CommunityData,
+            ContextData,
+            ObjectIdentity,
+            ObjectType,
+            SnmpEngine,
+            UdpTransportTarget,
+            getCmd as _get_cmd,
+        )
+        _transport_is_async = False
 
-    results: dict[str, float] = {}
     engine = SnmpEngine()
 
-    for oid_cfg in oids:
-        oid = oid_cfg["oid"]
-        label = oid_cfg.get("label") or oid
-        scale = float(oid_cfg.get("scale", 1.0))
-
+    # Always release the engine's UDP socket — see _close_engine for why.
+    try:
+        # Same host/port for every OID on this device, so build the transport once.
         try:
-            error_indication, error_status, error_index, var_binds = next(
-                getCmd(
+            if _transport_is_async:
+                transport = await UdpTransportTarget.create((host, port), timeout=5, retries=1)
+            else:
+                transport = UdpTransportTarget((host, port), timeout=5, retries=1)
+        except Exception as exc:
+            logger.debug("SNMP transport setup failed for %s:%s: %s", host, port, exc)
+            return {}
+
+        results: dict[str, float] = {}
+
+        for oid_cfg in oids:
+            oid = oid_cfg["oid"]
+            label = oid_cfg.get("label") or oid
+            scale = float(oid_cfg.get("scale", 1.0))
+
+            try:
+                error_indication, error_status, error_index, var_binds = await _get_cmd(
                     engine,
                     CommunityData(community, mpModel=_mp_model(version)),
-                    UdpTransportTarget((host, port), timeout=5, retries=1),
+                    transport,
                     ContextData(),
                     ObjectType(ObjectIdentity(oid)),
                 )
-            )
-        except Exception as exc:
-            logger.debug("SNMP GET %s@%s OID %s failed: %s", host, port, oid, exc)
-            continue
+            except Exception as exc:
+                logger.debug("SNMP GET %s@%s OID %s failed: %s", host, port, oid, exc)
+                continue
 
-        if error_indication:
-            logger.debug("SNMP error %s@%s OID %s: %s", host, port, oid, error_indication)
-            continue
-        if error_status:
-            logger.debug("SNMP status %s@%s OID %s: %s at %s",
-                         host, port, oid, error_status.prettyPrint(),
-                         error_index and var_binds[int(error_index) - 1][0] or "?")
-            continue
+            if error_indication:
+                logger.debug("SNMP error %s@%s OID %s: %s", host, port, oid, error_indication)
+                continue
+            if error_status:
+                logger.debug("SNMP status %s@%s OID %s: %s at %s",
+                             host, port, oid, error_status.prettyPrint(),
+                             error_index and var_binds[int(error_index) - 1][0] or "?")
+                continue
 
-        for _, val in var_binds:
-            try:
-                results[label] = float(val) * scale
-            except (TypeError, ValueError):
-                # Non-numeric type (e.g. OctetString description) — skip
-                logger.debug("SNMP non-numeric value for %s label=%s: %r", oid, label, val)
+            for _, val in var_binds:
+                try:
+                    results[label] = float(val) * scale
+                except (TypeError, ValueError):
+                    # Non-numeric type (e.g. OctetString description) — skip.
+                    logger.debug("SNMP non-numeric value for %s label=%s: %r", oid, label, val)
 
-    return results
+        return results
+    finally:
+        _close_engine(engine)

plugins/snmp/routes.py

@@ -93,6 +93,75 @@ async def index():
                                  poll_interval=poll_interval)
 
 
+def _device_binds_to_host(d: dict, keys: set[str], host_id: str) -> bool:
+    """Does config device `d` belong on the host identified by `keys`
+    (lowercased {address, name}) / `host_id`?
+
+    A device may declare an **explicit** binding that decouples "which Steward
+    host this belongs to" from "where to poll" (the `host` field):
+      • `host_id`       — the Steward Host UUID (exact match), the explicit link.
+      • `steward_host`  — a friendly bind by Host name or address (case-insensitive).
+    An explicit binding is **exclusive**: a device bound to host A must not also
+    match host B by a coincidental poll-target string. Only when no explicit
+    binding is present do we fall back to the implicit `host`-string match
+    (backward compatible with configs that only set `host`).
+    """
+    explicit_id = str(d.get("host_id", "")).strip()
+    explicit_name = str(d.get("steward_host", "")).strip().lower()
+    if explicit_id or explicit_name:
+        if explicit_id and host_id and explicit_id == host_id:
+            return True
+        return bool(explicit_name and explicit_name in keys)
+    return str(d.get("host", "")).strip().lower() in keys
+
+
+def _devices_for_host(devices_cfg: list, address: str | None, name: str | None,
+                      host_id: str | None = None) -> list[dict]:
+    """SNMP devices that map onto a Steward host's page. Prefers an explicit
+    per-device binding (`host_id` / `steward_host`); otherwise falls back to
+    matching the device's `host` (poll target) against the host's address or
+    name (case-insensitive). See `_device_binds_to_host` for the precedence."""
+    keys = {(address or "").strip().lower(), (name or "").strip().lower()}
+    keys.discard("")
+    hid = (host_id or "").strip()
+    return [
+        d for d in devices_cfg
+        if isinstance(d, dict) and _device_binds_to_host(d, keys, hid)
+    ]
+
+
+@snmp_bp.get("/host/<host_id>")
+@require_role(UserRole.viewer)
+async def host_panel(host_id: str):
+    """Per-host SNMP fragment for the Hosts hub. Surfaces any configured SNMP
+    device that maps to this host (by address or name) with its latest readings.
+    Renders nothing when no device maps, so hosts without SNMP carry no empty card.
+    """
+    from steward.models.hosts import Host
+
+    devices_cfg: list[dict] = current_app.config["PLUGINS"]["snmp"].get("devices", [])
+    async with current_app.db_sessionmaker() as db:
+        host = (await db.execute(
+            select(Host).where(Host.id == host_id))).scalar_one_or_none()
+        if host is None:
+            return ""
+        matched = _devices_for_host(devices_cfg, host.address, host.name, host.id)
+        if not matched:
+            return ""
+        names = [d.get("name") or d.get("host", "?") for d in matched]
+        latest = await _latest_readings(db, names)
+
+    devices = [{
+        "name": d.get("name") or d.get("host", "?"),
+        "host": d.get("host", ""),
+        "oids": d.get("oids", []),
+        "readings": latest.get(d.get("name") or d.get("host", "?"), {}),
+        "bound": bool(str(d.get("host_id", "")).strip()
+                      or str(d.get("steward_host", "")).strip()),
+    } for d in matched]
+    return await render_template("snmp/host_panel.html", devices=devices)
+
+
 @snmp_bp.get("/widget")
 @require_role(UserRole.viewer)
 async def widget():

plugins/snmp/scheduler.py

@@ -1,6 +1,5 @@
 # plugins/snmp/scheduler.py
 from __future__ import annotations
-import asyncio
 import logging
 from typing import TYPE_CHECKING
 
@@ -27,15 +26,13 @@ def make_poll_task(app: "Quart") -> ScheduledTask:
 
 
 async def _do_poll(app: "Quart") -> None:
-    from .poller import poll_device_sync
+    from .poller import poll_device
     from steward.core.alerts import record_metric
 
     devices: list[dict] = app.config["PLUGINS"]["snmp"].get("devices", [])
     if not devices:
         return
 
-    loop = asyncio.get_event_loop()
-
     async with app.db_sessionmaker() as session:
         async with session.begin():
             for device in devices:
@@ -52,11 +49,7 @@ async def _do_poll(app: "Quart") -> None:
                     continue
 
                 try:
-                    readings = await loop.run_in_executor(
-                        None,
-                        poll_device_sync,
-                        host, port, community, version, oids,
-                    )
+                    readings = await poll_device(host, port, community, version, oids)
                 except Exception:
                     logger.exception("SNMP poll failed for device %s (%s)", name, host)
                     continue

plugins/snmp/templates/snmp/device.html

@@ -1,9 +1,10 @@
 {# plugins/snmp/templates/snmp/device.html #}
 {% extends "base.html" %}
+{% from "_macros.html" import crumbs %}
 {% block title %}SNMP — {{ device_name }} — Steward{% endblock %}
+{% block breadcrumb %}{{ crumbs([("SNMP", "/plugins/snmp/"), (device_name, "")]) }}{% endblock %}
 {% block content %}
 <div style="display:flex;align-items:center;gap:1rem;margin-bottom:1.5rem;flex-wrap:wrap;">
-  <a href="/plugins/snmp/" style="color:var(--text-muted);font-size:0.85rem;">&#8592; SNMP</a>
   <h1 class="page-title" style="margin:0;">{{ device_name }}</h1>
   <span style="font-size:0.8rem;color:var(--text-dim);">{{ device.host }}</span>
   <div style="margin-left:auto;display:flex;gap:0.5rem;">

plugins/snmp/templates/snmp/host_panel.html

@@ -0,0 +1,51 @@
+{# Per-host SNMP fragment — embedded on the Hosts hub via HTMX. Self-contained.
+   Shows the SNMP device(s) whose address maps to this host + latest readings. #}
+<div class="card">
+  <div style="display:flex;align-items:baseline;justify-content:space-between;margin-bottom:0.6rem;gap:1rem;flex-wrap:wrap;">
+    <h3 class="section-title" style="margin-bottom:0;">SNMP</h3>
+    <a href="/plugins/snmp/" style="font-size:0.8rem;color:var(--text-muted);">All devices →</a>
+  </div>
+
+  {% for device in devices %}
+  <div style="{% if not loop.last %}margin-bottom:1rem;{% endif %}">
+    <div style="display:flex;align-items:center;gap:0.6rem;margin-bottom:0.5rem;flex-wrap:wrap;">
+      <span style="font-weight:600;font-size:0.92rem;">{{ device.name }}</span>
+      <span style="font-size:0.78rem;color:var(--text-dim);font-family:ui-monospace,monospace;">{{ device.host }}</span>
+      {% if device.bound %}
+      <span title="Explicitly bound to this host (host_id / steward_host)"
+            style="font-size:0.68rem;color:var(--text-muted);background:var(--bg-elevated);border:1px solid var(--border);border-radius:3px;padding:0.05rem 0.4rem;">bound</span>
+      {% endif %}
+      {% if device.readings %}
+      <span style="font-size:0.74rem;color:var(--green);">&#9679; reachable</span>
+      {% else %}
+      <span style="font-size:0.74rem;color:var(--text-dim);">&#9675; no data yet</span>
+      {% endif %}
+      <a href="/plugins/snmp/device/{{ device.name }}" class="btn btn-ghost btn-sm"
+         style="margin-left:auto;font-size:0.76rem;padding:0.15rem 0.55rem;">History</a>
+    </div>
+
+    {% if device.readings %}
+    <div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(150px,1fr));gap:0.5rem;">
+      {% for oid_cfg in device.oids %}
+      {% set label = oid_cfg.label if oid_cfg.label else oid_cfg.oid %}
+      {% set val = device.readings.get(label) %}
+      <div style="background:var(--bg-elevated);border-radius:5px;padding:0.5rem 0.7rem;border:1px solid var(--border);">
+        <div style="font-size:0.68rem;color:var(--text-dim);margin-bottom:0.2rem;font-family:ui-monospace,monospace;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">{{ label }}</div>
+        {% if val is not none %}
+        <div style="font-size:1rem;font-weight:600;font-variant-numeric:tabular-nums;">
+          {% if val >= 1000000 %}{{ "%.2f"|format(val / 1000000) }}<span style="font-size:0.72rem;color:var(--text-muted);">M</span>
+          {% elif val >= 1000 %}{{ "%.1f"|format(val / 1000) }}<span style="font-size:0.72rem;color:var(--text-muted);">K</span>
+          {% else %}{{ "%.4g"|format(val) }}{% endif %}
+        </div>
+        {% else %}
+        <div style="font-size:0.85rem;color:var(--text-dim);">—</div>
+        {% endif %}
+      </div>
+      {% endfor %}
+    </div>
+    {% else %}
+    <p style="font-size:0.8rem;color:var(--text-dim);margin:0;">No readings yet — waiting for the next poll.</p>
+    {% endif %}
+  </div>
+  {% endfor %}
+</div>

plugins/traefik/__init__.py

@@ -28,3 +28,8 @@ def get_scheduled_tasks() -> list:
 def get_blueprint():
     from .routes import traefik_bp
     return traefik_bp
+
+
+def get_nav() -> list[dict]:
+    # Surfaces in the sidebar's Infrastructure group (Traefik services view).
+    return [{"label": "Traefik", "href": "/plugins/traefik/"}]

plugins/unifi/__init__.py

@@ -26,3 +26,8 @@ def get_scheduled_tasks() -> list:
 def get_blueprint():
     from .routes import unifi_bp
     return unifi_bp
+
+
+def get_nav() -> list[dict]:
+    # Surfaces in the sidebar's Infrastructure group (UniFi network overview).
+    return [{"label": "UniFi", "href": "/plugins/unifi/"}]

plugins/unifi/scheduler.py

@@ -15,6 +15,25 @@ logger = logging.getLogger(__name__)
 _client = None  # UnifiClient instance, initialised on first scrape
 
 
+async def _drop_client() -> None:
+    """Close and discard the cached client.
+
+    The client holds a long-lived ``httpx.AsyncClient`` (a connection pool over
+    real sockets). Nulling the reference without ``aclose()`` orphans that pool
+    until GC, so every failure tick that re-auths would leak file descriptors —
+    the same class of bug that took the app down via SNMP (Errno 24, "Too many
+    open files"). Close first, then drop. Best-effort: a failed close must not
+    break the poll loop.
+    """
+    global _client
+    if _client is not None:
+        try:
+            await _client.close()
+        except Exception:
+            pass
+        _client = None
+
+
 def make_poll_task(app: "Quart") -> ScheduledTask:
     cfg = app.config["PLUGINS"]["unifi"]
     interval = int(cfg.get("poll_interval_seconds", 60))
@@ -51,7 +70,7 @@ async def _do_poll(app: "Quart") -> None:
             await _client.login()
         except Exception as exc:
             logger.warning("UniFi initial login failed: %s", exc)
-            _client = None
+            await _drop_client()
             return
 
     try:
@@ -60,7 +79,7 @@ async def _do_poll(app: "Quart") -> None:
         devices = await _client.get_devices()
     except Exception as exc:
         logger.warning("UniFi poll failed — will retry next tick: %s", exc)
-        _client = None  # force re-auth on next tick
+        await _drop_client()  # close + force re-auth on next tick
         return
 
     scraped_at = datetime.now(timezone.utc)

steward/ansible/routes.py

@@ -172,6 +172,8 @@ async def playbook_vars():
     (vars/vars_prompt), loaded when a playbook is chosen from a dropdown."""
     source_name = (request.args.get("source_name", "") or "").strip()
     playbook_path = (request.args.get("playbook_path", "") or "").strip()
+    # Schedule form: schedules can't prompt, so secret vars render disabled.
+    schedule = (request.args.get("schedule", "") or "").strip() == "1"
     source = next((s for s in _get_sources() if s["name"] == source_name), None)
     variables: list = []
     meta: dict = {"description": "", "category": "", "confirm": False}
@@ -182,7 +184,8 @@ async def playbook_vars():
             meta = src_module.discover_playbook_meta(contents)
     return await render_template(
         "ansible/_playbook_vars.html", variables=variables,
-        description=meta["description"], category=meta["category"], confirm=meta["confirm"])
+        description=meta["description"], category=meta["category"],
+        confirm=meta["confirm"], schedule=schedule)
 
 
 @ansible_bp.get("/run-form/<source_name>/<path:playbook_path>")
@@ -370,7 +373,6 @@ async def schedules():
         {"name": s["name"], "playbooks": src_module.discover_playbooks(s["path"])}
         for s in _get_sources()
     ]
-    all_playbooks = sorted({p for sd in source_data for p in sd["playbooks"]})
 
     rows = []
     editing = None
@@ -383,11 +385,31 @@ async def schedules():
         if editing_id and s.id == editing_id:
             editing = s
 
+    # Default the source dropdown to the first source (or the edited schedule's
+    # source), and pre-render that source's playbooks so the playbook dropdown
+    # starts on a real item rather than a placeholder.
+    sel_source = editing.source_name if editing else (source_data[0]["name"] if source_data else "")
+    sel_playbooks = next((sd["playbooks"] for sd in source_data if sd["name"] == sel_source), [])
+
+    # Edit mode: pre-render the selected playbook's declared variables so their
+    # saved values show (fresh loads fetch these via HTMX from /playbook-vars).
+    edit_variables: list = []
+    edit_meta: dict = {"description": "", "category": "", "confirm": False}
+    if editing:
+        src = next((s for s in _get_sources() if s["name"] == editing.source_name), None)
+        if src:
+            contents = src_module.read_playbook(src["path"], editing.playbook_path)
+            if contents is not None:
+                edit_variables = src_module.discover_playbook_variables(contents)
+                edit_meta = src_module.discover_playbook_meta(contents)
+
     return await render_template(
         "ansible/schedules.html",
         rows=rows, groups=groups, targets=targets,
-        source_data=source_data, all_playbooks=all_playbooks,
+        source_data=source_data, sel_source=sel_source, sel_playbooks=sel_playbooks,
         editing=editing, interval_presets=INTERVAL_PRESETS,
+        edit_variables=edit_variables, edit_description=edit_meta["description"],
+        edit_category=edit_meta["category"], edit_confirm=edit_meta["confirm"],
     )
 
 

steward/app.py

@@ -1,11 +1,14 @@
 # steward/app.py
 from __future__ import annotations
 import asyncio
+import logging
 from pathlib import Path
 from quart import Quart, render_template
 from .config import load_bootstrap
 from .database import init_db
 
+logger = logging.getLogger(__name__)
+
 
 def create_app(
     config_path: Path | str | None = None,
@@ -50,9 +53,17 @@ def create_app(
     # ── 3b. Encryption-at-rest: bind the encryptor, then encrypt legacy secrets ─
     if not testing:
         from .core.crypto import init_crypto
-        from .core.settings import migrate_plaintext_secrets
+        from .core.settings import migrate_plaintext_secrets, scan_undecryptable_secrets
         init_crypto(app.config["SECRET_KEY"])
         migrate_plaintext_secrets(app.config["DATABASE_URL"])
+        # Flag any secrets sealed under a now-lost key (see the admin banner).
+        undecryptable = scan_undecryptable_secrets(app.config["DATABASE_URL"])
+        if undecryptable:
+            logger.warning(
+                "%d stored secret(s) cannot be decrypted with the current app key "
+                "(re-enter them in Settings): %s",
+                len(undecryptable), ", ".join(undecryptable),
+            )
 
     # ── 4. Load all settings from DB → populate app.config ────────────────────
     if not testing:
@@ -166,7 +177,8 @@ def create_app(
     # ── 10. Template context: inject plugin_failures into every response ───────
     @app.context_processor
     def _inject_plugin_failures():
-        from .core.plugin_manager import get_plugin_failures
+        from .core.plugin_manager import get_plugin_failures, get_plugin_nav
+        from .core.settings import get_undecryptable_secrets
         # enabled_plugins lets templates gate cross-plugin embeds (e.g. the
         # Hosts hub only fetches the docker fragment when docker is enabled),
         # avoiding a 404 to a route whose blueprint isn't registered.
@@ -174,9 +186,16 @@ def create_app(
             name for name, cfg in (app.config.get("PLUGINS") or {}).items()
             if isinstance(cfg, dict) and cfg.get("enabled")
         }
+        # Sidebar entries for enabled plugins that expose a UI (get_nav). Filter
+        # by enabled so a hot-disabled plugin's link can't linger before restart.
+        plugin_nav = [e for e in get_plugin_nav() if e["plugin"] in enabled_plugins]
         return {
             "plugin_failures": get_plugin_failures(),
             "enabled_plugins": enabled_plugins,
+            "plugin_nav": plugin_nav,
+            # Read from an in-memory cache (no DB hit per render); kept current by
+            # the startup scan + set_setting's discard-on-write.
+            "undecryptable_secrets": get_undecryptable_secrets(),
         }
 
     # ── 11. Share-token middleware ─────────────────────────────────────────────
@@ -249,6 +268,10 @@ def _register_core_tasks(app: Quart) -> None:
         from .core.reports import check_and_send
         await check_and_send(app)
 
+    async def run_self_monitor():
+        from .core.self_monitor import record_self_metrics
+        await record_self_metrics(app)
+
     app._task_registry.extend([
         ScheduledTask(
             name="weekly_report_check",
@@ -256,6 +279,15 @@ def _register_core_tasks(app: Quart) -> None:
             interval_seconds=3600,
             run_on_startup=False,
         ),
+        # Watch our own fd usage so a descriptor leak surfaces as a metric/alert
+        # instead of a silent Errno 24 lockup. Cheap; plugin_metrics roll up
+        # hourly so the 60s cadence is storage-bounded.
+        ScheduledTask(
+            name="self_monitor",
+            coro_factory=run_self_monitor,
+            interval_seconds=60,
+            run_on_startup=True,
+        ),
         ScheduledTask(
             name="monitor_check",
             coro_factory=run_monitor_checks,

steward/config.py

@@ -75,13 +75,27 @@ def load_bootstrap(config_path: Path | str | None = None) -> dict[str, Any]:
 
 
 def _resolve_secret_key(raw: dict) -> str:
-    """Resolve secret_key: env var → file → auto-generate."""
+    """Resolve secret_key: env var → file → auto-generate.
+
+    Refuses to start if a new key must be generated but cannot be persisted: an
+    ephemeral key changes on every restart, which silently renders every
+    encrypted secret (managed SSH key, SMTP/OIDC/LDAP credentials) unrecoverable.
+    Failing loudly with a fix beats limping along and losing data on the next
+    boot — exactly the footgun that bit the vdnt-docker02 deployment.
+    """
     from_env = _env("SECRET_KEY") or raw.get("secret_key")
     if from_env:
         return from_env
 
     if _SECRET_KEY_FILE.exists():
-        key = _SECRET_KEY_FILE.read_text().strip()
+        try:
+            key = _SECRET_KEY_FILE.read_text().strip()
+        except OSError as exc:
+            raise RuntimeError(
+                f"App secret key file {_SECRET_KEY_FILE} exists but cannot be read "
+                f"({exc}). Make it readable by the container user (uid 1000), or set "
+                f"STEWARD_SECRET_KEY."
+            ) from exc
         if key:
             return key
 
@@ -89,11 +103,16 @@ def _resolve_secret_key(raw: dict) -> str:
     try:
         _SECRET_KEY_FILE.parent.mkdir(parents=True, exist_ok=True)
         _SECRET_KEY_FILE.write_text(key)
-        logger.info("Generated new secret key and saved to %s", _SECRET_KEY_FILE)
     except OSError as exc:
-        logger.warning(
-            "Could not write secret key to %s (%s). "
-            "Key will not persist across restarts.",
-            _SECRET_KEY_FILE, exc,
-        )
+        raise RuntimeError(
+            f"Generated a new app secret key but could not persist it to "
+            f"{_SECRET_KEY_FILE} ({exc}). An ephemeral key changes on every restart, "
+            f"which makes all encrypted secrets (managed SSH key, SMTP/OIDC/LDAP "
+            f"credentials) unrecoverable. Fix one of:\n"
+            f"  • set STEWARD_SECRET_KEY to a stable value "
+            f"(recommended for Swarm / multi-node), or\n"
+            f"  • make {_SECRET_KEY_FILE.parent} writable by the container user "
+            f"(uid 1000)."
+        ) from exc
+    logger.info("Generated new secret key and saved to %s", _SECRET_KEY_FILE)
     return key

steward/core/cleanup.py

@@ -6,7 +6,6 @@ from typing import TYPE_CHECKING
 from sqlalchemy import delete
 
 from steward.models.monitors import MonitorResult
-from steward.models.metrics import PluginMetric
 from steward.models.ansible import AnsibleRun
 
 if TYPE_CHECKING:
@@ -26,7 +25,6 @@ async def run_cleanup(app: "Quart") -> None:
         async with session.begin():
             for model, ts_col in [
                 (MonitorResult, MonitorResult.checked_at),
-                (PluginMetric, PluginMetric.recorded_at),
                 (AnsibleRun, AnsibleRun.started_at),
             ]:
                 result = await session.execute(
@@ -35,9 +33,29 @@ async def run_cleanup(app: "Quart") -> None:
                 if result.rowcount:
                     logger.info(f"Pruned {result.rowcount} rows from {model.__tablename__}")
 
+        # plugin_metrics is NOT blanket-deleted here — it's rolled up to hourly
+        # then pruned, so multi-week host history stays cheap.
+        await _run_metrics_retention(session, now)
         await _run_docker_retention(session, now)
 
 
+async def _run_metrics_retention(session, now: datetime) -> None:
+    """Roll up + prune plugin_metrics (raw → hourly → gone). Windows read fresh
+    from settings each run (rule 25 — UI change takes effect next cleanup, no
+    restart). get_setting's SELECT autobegins, so read inside the begin block."""
+    from steward.core.metrics_retention import rollup_plugin_metrics
+    from steward.core.settings import get_setting
+
+    async with session.begin():
+        raw_days = int(await get_setting(session, "metrics.retention.raw_days") or 7)
+        rollup_days = int(await get_setting(session, "metrics.retention.rollup_days") or 90)
+        counts = await rollup_plugin_metrics(
+            session, raw_days=raw_days, rollup_days=rollup_days, now=now,
+        )
+    if counts and any(counts.values()):
+        logger.info("Metrics retention: %s", counts)
+
+
 async def _run_docker_retention(session, now: datetime) -> None:
     """Drive the docker plugin's rollup + prune via its capability, if loaded.
 

steward/core/metrics_retention.py

@@ -0,0 +1,112 @@
+"""Bound plugin_metrics growth: roll old raw samples up to hourly, prune the rest.
+
+plugin_metrics grows by (sources × resources × sample cadence) — host agents push
+host-level + per-core/mount/iface sub-resources every ~30s, so a fleet accrues
+millions of rows. We keep raw samples for a short window, aggregate everything
+older into hourly averages (plugin_metrics_hourly) and delete the raw rows, then
+prune hourly beyond a longer window. Charts read raw for the recent part of a
+range and hourly for the older part.
+
+Driven by the core cleanup task (steward.core.cleanup). Runs inside the caller's
+open transaction; never opens or commits its own.
+"""
+from __future__ import annotations
+
+from datetime import datetime, timedelta
+
+
+def _hour_floor(dt: datetime) -> datetime:
+    """Truncate a datetime down to the start of its hour (drops min/sec/µs)."""
+    return dt.replace(minute=0, second=0, microsecond=0)
+
+
+def _rollup_cutoff(now: datetime, raw_days: int) -> datetime:
+    """Hour-aligned boundary below which raw metrics get rolled up + deleted.
+
+    Aligning to the hour means we only roll up *whole* elapsed hours — a bucket
+    is never split across the keep/roll boundary, so a re-run can't produce a
+    partial-then-complete duplicate for the same hour.
+    """
+    return _hour_floor(now - timedelta(days=raw_days))
+
+
+async def rollup_plugin_metrics(
+    session,
+    *,
+    raw_days: int,
+    rollup_days: int,
+    now: datetime | None = None,
+) -> dict:
+    """Roll up + prune plugin_metrics. Returns a counts dict for logging.
+
+    1. Aggregate plugin_metrics older than the (hour-aligned) raw window into
+       plugin_metrics_hourly (avg/max per source/resource/metric/hour), upserting
+       so a re-run is idempotent, then delete those raw rows.
+    2. Prune rolled-up rows older than the rollup window.
+    """
+    from datetime import timezone
+
+    from sqlalchemy import delete, func, select
+    from sqlalchemy.dialects.postgresql import insert as pg_insert
+
+    from steward.models.metrics import PluginMetric, PluginMetricHourly
+
+    if now is None:
+        now = datetime.now(timezone.utc)
+
+    rolled = rolled_rows = rollup_pruned = 0
+
+    # ── 1. Roll up raw metrics older than the raw window into hourly buckets ──
+    raw_cutoff = _rollup_cutoff(now, raw_days)
+    hour = func.date_trunc("hour", PluginMetric.recorded_at)
+    agg = (
+        select(
+            PluginMetric.source_module,
+            PluginMetric.resource_name,
+            PluginMetric.metric_name,
+            hour.label("bucket"),
+            func.avg(PluginMetric.value).label("value_avg"),
+            func.max(PluginMetric.value).label("value_max"),
+            func.count().label("sample_count"),
+        )
+        .where(PluginMetric.recorded_at < raw_cutoff)
+        .group_by(
+            PluginMetric.source_module, PluginMetric.resource_name,
+            PluginMetric.metric_name, hour,
+        )
+    )
+    for r in (await session.execute(agg)).all():
+        avg_v = float(r.value_avg or 0.0)
+        max_v = float(r.value_max or 0.0)
+        cnt = int(r.sample_count or 0)
+        await session.execute(
+            pg_insert(PluginMetricHourly)
+            .values(
+                source_module=r.source_module, resource_name=r.resource_name,
+                metric_name=r.metric_name, bucket=r.bucket,
+                value_avg=avg_v, value_max=max_v, sample_count=cnt,
+            )
+            .on_conflict_do_update(
+                constraint="uq_plugin_metrics_hourly_bucket",
+                set_={"value_avg": avg_v, "value_max": max_v, "sample_count": cnt},
+            )
+        )
+        rolled += 1
+        rolled_rows += cnt
+    if rolled:
+        await session.execute(
+            delete(PluginMetric).where(PluginMetric.recorded_at < raw_cutoff)
+        )
+
+    # ── 2. Prune rolled-up rows beyond the rollup window ──
+    rollup_cutoff = now - timedelta(days=rollup_days)
+    res = await session.execute(
+        delete(PluginMetricHourly).where(PluginMetricHourly.bucket < rollup_cutoff)
+    )
+    rollup_pruned = res.rowcount or 0
+
+    return {
+        "buckets_rolled": rolled,
+        "raw_rows_rolled": rolled_rows,
+        "rollup_pruned": rollup_pruned,
+    }

steward/core/plugin_manager.py

@@ -24,12 +24,52 @@ _LOADED_PLUGINS: set[str] = set()
 # Track plugins that failed to load: name → human-readable reason.
 _FAILED_PLUGINS: dict[str, str] = {}
 
+# Nav entries contributed by plugins via the optional get_nav() export, so a
+# plugin's UI gets a home in the sidebar. Each entry:
+#   {"plugin": <name>, "label": str, "href": str, "section": str}
+_PLUGIN_NAV: list[dict] = []
+
 
 def get_plugin_failures() -> dict[str, str]:
     """Return a copy of the failed-plugin registry (name → error message)."""
     return dict(_FAILED_PLUGINS)
 
 
+def get_plugin_nav() -> list[dict]:
+    """Plugin-contributed sidebar entries, sorted by section then label."""
+    return sorted(_PLUGIN_NAV, key=lambda e: (e.get("section", ""), e.get("label", "")))
+
+
+def _collect_plugin_nav(name: str, module) -> None:
+    """Pull a plugin's optional get_nav() entries into the nav registry.
+
+    Tolerant by design: a missing hook is fine, and a raising or malformed hook
+    is logged and skipped — a bad nav contribution must never break loading.
+    Idempotent per plugin (drops prior entries first) so hot-reload re-runs cleanly.
+    """
+    global _PLUGIN_NAV
+    _PLUGIN_NAV = [e for e in _PLUGIN_NAV if e.get("plugin") != name]
+    if not hasattr(module, "get_nav"):
+        return
+    try:
+        items = module.get_nav() or []
+    except Exception:
+        logger.exception("Plugin %r: get_nav() raised, skipping its nav entries", name)
+        return
+    for item in items:
+        try:
+            label, href = str(item["label"]), str(item["href"])
+        except (TypeError, KeyError):
+            logger.warning("Plugin %r: malformed nav item %r, skipping", name, item)
+            continue
+        _PLUGIN_NAV.append({
+            "plugin": name,
+            "label": label,
+            "href": href,
+            "section": str(item.get("section", "Infrastructure")),
+        })
+
+
 def resolve_plugin_path(plugin_dirs: list[Path], name: str) -> Path | None:
     """Return the first plugin root that contains `name`, else None.
 
@@ -222,6 +262,7 @@ def load_plugins(app: "Quart") -> None:
 
         _FAILED_PLUGINS.pop(name, None)  # clear any stale failure from a previous reload attempt
         _LOADED_PLUGINS.add(name)
+        _collect_plugin_nav(name, module)
         logger.info("Plugin %r loaded (v%s)", name, meta.get("version", "?"))
 
 
@@ -440,6 +481,7 @@ def hot_reload_plugin(app: "Quart", name: str) -> tuple[bool, str]:
         return False, f"get_scheduled_tasks() raised: {exc}"
 
     _LOADED_PLUGINS.add(name)
+    _collect_plugin_nav(name, module)
     logger.info("Plugin %r hot-reloaded (v%s)", name, meta.get("version", "?"))
     return True, f"Plugin activated (v{meta.get('version', '?')})"
 

steward/core/scheduler.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 import asyncio
 import logging
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Callable, Coroutine
 
 logger = logging.getLogger(__name__)
@@ -15,28 +15,65 @@ class ScheduledTask:
     run_on_startup: bool = False
 
 
+@dataclass
+class _DueTracker:
+    """Decides which scheduled tasks are due, with a self-overlap guard.
+
+    Pure (no asyncio, no clock of its own — `now` is passed in) so the
+    scheduling policy is unit-testable without timing races. A task whose prior
+    run is still in flight is NOT re-fired: overlapping poll runs stack up open
+    connections/subprocesses and amplify any per-poll resource use — the same
+    failure mode behind the fd-leak lockups. The skipped task is retried on the
+    next tick once it completes (its last_run isn't advanced while skipped).
+    """
+    last_run: dict[str, float] = field(default_factory=dict)
+    in_flight: set[str] = field(default_factory=set)
+
+    def due(self, tasks: list[ScheduledTask], now: float) -> list[ScheduledTask]:
+        ready: list[ScheduledTask] = []
+        for task in tasks:
+            if now - self.last_run.get(task.name, 0) < task.interval_seconds:
+                continue
+            if task.name in self.in_flight:
+                logger.warning(
+                    "Scheduled task %r still running — skipping this tick",
+                    task.name)
+                continue
+            ready.append(task)
+        return ready
+
+    def mark_started(self, task: ScheduledTask, now: float) -> None:
+        self.in_flight.add(task.name)
+        self.last_run[task.name] = now
+
+    def mark_done(self, name: str) -> None:
+        self.in_flight.discard(name)
+
+
 async def start_scheduler(tasks: list[ScheduledTask]) -> None:
     """Run scheduled tasks in a loop. Call with asyncio.create_task()."""
-    last_run: dict[str, float] = {}
+    tracker = _DueTracker()
+
+    def _spawn(task: ScheduledTask, now: float) -> None:
+        tracker.mark_started(task, now)
+        asyncio.create_task(_run_task(task, tracker))
 
     for task in tasks:
         if task.run_on_startup:
             logger.info(f"Startup task: {task.name}")
-            asyncio.create_task(_run_task(task))
-            last_run[task.name] = asyncio.get_event_loop().time()
+            _spawn(task, asyncio.get_event_loop().time())
 
     while True:
         now = asyncio.get_event_loop().time()
-        for task in tasks:
-            last = last_run.get(task.name, 0)
-            if now - last >= task.interval_seconds:
-                asyncio.create_task(_run_task(task))
-                last_run[task.name] = now
+        for task in tracker.due(tasks, now):
+            _spawn(task, now)
         await asyncio.sleep(1)
 
 
-async def _run_task(task: ScheduledTask) -> None:
+async def _run_task(task: ScheduledTask, tracker: _DueTracker) -> None:
     try:
         await task.coro_factory()
     except Exception:
         logger.exception(f"Scheduled task {task.name!r} raised an exception")
+    finally:
+        tracker.mark_done(task.name)

steward/core/self_monitor.py

@@ -0,0 +1,116 @@
+"""Self-monitoring: record Steward's own open file-descriptor usage.
+
+A leaked socket/file handle in a poll loop (see the SNMP and UniFi fd-leak
+issues) used to fail silently until the process hit its fd ceiling and
+`socket.accept()` started raising `OSError: [Errno 24] Too many open files` —
+taking the whole app down with no early warning.
+
+This turns that failure mode into an observable signal. Steward monitors other
+things; it should monitor itself. We record two metrics each tick under
+`source_module="steward"`:
+
+  • ``open_fds``      — raw count of open descriptors
+  • ``open_fds_pct``  — that count as a percentage of the soft RLIMIT_NOFILE
+
+Both flow through the normal alert pipeline, so the operator can attach an alert
+rule to either via the existing alert-rules UI. As a zero-config floor we also
+log a WARNING once usage crosses ``FD_WARN_PCT`` — a leak becomes visible even
+before any rule is set up.
+"""
+from __future__ import annotations
+import logging
+import os
+
+logger = logging.getLogger(__name__)
+
+# Stdlib-only fd counting via /proc keeps this dependency-free. resource is
+# POSIX-only but always present on the Linux runtime image; guard anyway so
+# imports never explode on a dev machine.
+try:
+    import resource
+except ImportError:  # pragma: no cover - non-POSIX
+    resource = None  # type: ignore[assignment]
+
+# Warn (without needing a configured alert rule) once we're using this fraction
+# of the soft fd limit. 80% leaves headroom to act before accepts start failing.
+FD_WARN_PCT = 80.0
+
+_warned = False  # de-dupe the WARNING so a sustained leak doesn't spam the log
+
+
+def count_open_fds() -> int | None:
+    """Number of open file descriptors for this process, or None if unknown.
+
+    Reads ``/proc/self/fd`` (Linux). Returns None where /proc isn't available
+    (e.g. a macOS dev box) so callers degrade to a no-op rather than guessing.
+    """
+    try:
+        return len(os.listdir("/proc/self/fd"))
+    except OSError:
+        return None
+
+
+def fd_soft_limit() -> int | None:
+    """Soft RLIMIT_NOFILE for this process, or None if it can't be read.
+
+    None when the limit is unknown or 'unlimited' (RLIM_INFINITY) — a percentage
+    against an unbounded ceiling is meaningless, so we skip the pct metric then.
+    """
+    if resource is None:
+        return None
+    try:
+        soft, _hard = resource.getrlimit(resource.RLIMIT_NOFILE)
+    except (ValueError, OSError):
+        return None
+    if soft <= 0 or soft == resource.RLIM_INFINITY:
+        return None
+    return soft
+
+
+async def record_self_metrics(app) -> None:
+    """Record open-fd usage as Steward's own metrics; warn past the floor.
+
+    No-op (logged at debug) when fd accounting isn't available on this platform,
+    so it's safe to schedule unconditionally.
+    """
+    global _warned
+
+    fds = count_open_fds()
+    if fds is None:
+        logger.debug("self_monitor: /proc/self/fd unavailable — skipping fd metrics")
+        return
+
+    from .alerts import record_metric
+
+    soft = fd_soft_limit()
+    pct = (fds / soft * 100.0) if soft else None
+
+    async with app.db_sessionmaker() as session:
+        async with session.begin():
+            await record_metric(
+                session=session,
+                source_module="steward",
+                resource_name="process",
+                metric_name="open_fds",
+                value=float(fds),
+            )
+            if pct is not None:
+                await record_metric(
+                    session=session,
+                    source_module="steward",
+                    resource_name="process",
+                    metric_name="open_fds_pct",
+                    value=pct,
+                )
+
+    if pct is not None and pct >= FD_WARN_PCT:
+        if not _warned:
+            logger.warning(
+                "Open file descriptors at %.0f%% of the soft limit (%d/%d) — "
+                "possible descriptor leak; the app will stop accepting "
+                "connections if this reaches 100%%.", pct, fds, soft)
+            _warned = True
+    else:
+        _warned = False  # recovered — re-arm the warning for the next breach
+
+    logger.debug("self_monitor: open_fds=%d soft_limit=%s pct=%s", fds, soft, pct)

steward/core/settings.py

@@ -84,6 +84,10 @@ DEFAULTS: dict[str, Any] = {
     "docker.retention.metrics_raw_days": 7,
     "docker.retention.metrics_rollup_days": 90,
     "docker.retention.events_days": 30,
+    # Host/plugin metrics retention (plugin_metrics): keep a short raw window at
+    # the agent's ~30s cadence, then roll up to hourly averages kept much longer.
+    "metrics.retention.raw_days": 7,
+    "metrics.retention.rollup_days": 90,
     "plugins.index_url": "https://git.fabledsword.com/bvandeusen/Steward-plugins/raw/branch/main/index.yaml",
     # Default-enabled plugins. These are the generic, non-vendor-specific
     # bundled plugins (protocols/standards, not a single product) — useful on
@@ -150,6 +154,34 @@ def _decode(value: Any, key: str = "") -> Any:
     return decrypt_secret(value, context=key) if is_encrypted(value) else value
 
 
+# Secret settings that are stored as ciphertext but won't decrypt with the
+# current app key (key rotated or lost). Surfaced as an admin banner so the
+# operator knows precisely which secrets to re-enter — instead of finding out
+# only when something that uses one fails. Refreshed at startup
+# (scan_undecryptable_secrets) and kept live: re-entering a secret clears it
+# without a restart (set_setting discards it on a fresh write).
+_undecryptable_secrets: set[str] = set()
+
+
+def get_undecryptable_secrets() -> list[str]:
+    """Sorted keys whose stored ciphertext won't decrypt (for the UI banner)."""
+    return sorted(_undecryptable_secrets)
+
+
+def _is_undecryptable(stored: Any, key: str = "") -> bool:
+    """True iff `stored` is an encrypted token that fails to decrypt.
+
+    A failed decrypt returns the ciphertext unchanged (still enc-prefixed), so a
+    value that is still encrypted after a decrypt attempt is undecryptable.
+    """
+    from steward.core.crypto import decrypt_secret, is_encrypted
+    return (
+        isinstance(stored, str)
+        and is_encrypted(stored)
+        and is_encrypted(decrypt_secret(stored, context=key))
+    )
+
+
 # ─────────────────────────────────────────────────────────────────────────────
 # Async helpers (use inside request handlers / scheduled tasks)
 # ─────────────────────────────────────────────────────────────────────────────
@@ -182,6 +214,12 @@ async def set_setting(session: AsyncSession, key: str, value: Any) -> None:
         row.value_json = json.dumps(to_store)
         row.updated_at = now
 
+    # A fresh write of a secret is encrypted with the current key (or cleared to
+    # plaintext), so it's decryptable now — drop any stale "undecryptable" flag
+    # so the banner clears without a restart.
+    if key in SECRET_KEYS:
+        _undecryptable_secrets.discard(key)
+
 
 async def get_all_settings(session: AsyncSession) -> dict[str, Any]:
     """Return flat key→value dict with defaults filled in for missing keys."""
@@ -359,6 +397,37 @@ def migrate_plaintext_secrets(db_url: str) -> int:
     return asyncio.run(_run())
 
 
+def scan_undecryptable_secrets(db_url: str) -> list[str]:
+    """Populate the undecryptable-secrets cache from the DB; return the keys found.
+
+    Run once at startup, AFTER migrate_plaintext_secrets and init_crypto: any row
+    that's encrypted but won't decrypt with the current key was sealed under a
+    different (lost/rotated) key and must be re-entered. Surfaced via the admin
+    banner (get_undecryptable_secrets).
+    """
+    async def _run() -> list[str]:
+        from sqlalchemy.ext.asyncio import create_async_engine, async_sessionmaker
+        engine = create_async_engine(db_url, echo=False)
+        factory = async_sessionmaker(engine, expire_on_commit=False)
+        bad: list[str] = []
+        try:
+            async with factory() as session:
+                result = await session.execute(
+                    select(AppSetting).where(AppSetting.key.in_(SECRET_KEYS))
+                )
+                for row in result.scalars():
+                    if _is_undecryptable(json.loads(row.value_json), row.key):
+                        bad.append(row.key)
+        finally:
+            await engine.dispose()
+        return bad
+
+    global _undecryptable_secrets
+    found = asyncio.run(_run())
+    _undecryptable_secrets = set(found)
+    return sorted(found)
+
+
 # ─────────────────────────────────────────────────────────────────────────────
 # External URL helper
 # ─────────────────────────────────────────────────────────────────────────────

steward/migrations/versions/0023_plugin_metrics_indexes.py

@@ -0,0 +1,39 @@
+"""Index plugin_metrics for host/dashboard reads
+
+plugin_metrics had only a PK on id, so every host-detail / full-metrics /
+dashboard-widget query (all filter by source_module + resource_name over a
+recorded_at range) sequentially scanned the whole time-series table — slower as
+samples accumulate. Add the two composite indexes that match those query shapes.
+
+The non-concurrent CREATE INDEX takes a brief exclusive lock; it runs once at
+startup migration time, acceptable for this table.
+
+Revision ID: 0023_plugin_metrics_indexes
+Revises: 0022_unify_monitors
+Create Date: 2026-06-20
+"""
+from typing import Sequence, Union
+from alembic import op
+
+revision: str = "0023_plugin_metrics_indexes"
+down_revision: Union[str, None] = "0022_unify_monitors"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.create_index(
+        "ix_plugin_metrics_module_resource_recorded",
+        "plugin_metrics",
+        ["source_module", "resource_name", "recorded_at"],
+    )
+    op.create_index(
+        "ix_plugin_metrics_module_resource_metric_recorded",
+        "plugin_metrics",
+        ["source_module", "resource_name", "metric_name", "recorded_at"],
+    )
+
+
+def downgrade() -> None:
+    op.drop_index("ix_plugin_metrics_module_resource_metric_recorded", "plugin_metrics")
+    op.drop_index("ix_plugin_metrics_module_resource_recorded", "plugin_metrics")

steward/migrations/versions/0024_plugin_metrics_hourly.py

@@ -0,0 +1,43 @@
+"""Hourly rollup table for plugin_metrics
+
+Adds plugin_metrics_hourly — the coarse series that retention rolls raw
+plugin_metrics into before pruning them, so multi-day/week host history stays
+cheap to store. One row per (source_module, resource_name, metric_name, hour);
+the unique constraint is the conflict target for the idempotent rollup upsert.
+
+Revision ID: 0024_plugin_metrics_hourly
+Revises: 0023_plugin_metrics_indexes
+Create Date: 2026-06-20
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "0024_plugin_metrics_hourly"
+down_revision: Union[str, None] = "0023_plugin_metrics_indexes"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "plugin_metrics_hourly",
+        sa.Column("id", sa.String(length=36), nullable=False),
+        sa.Column("source_module", sa.String(length=64), nullable=False),
+        sa.Column("resource_name", sa.String(length=255), nullable=False),
+        sa.Column("metric_name", sa.String(length=128), nullable=False),
+        sa.Column("bucket", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("value_avg", sa.Float(), nullable=False, server_default="0"),
+        sa.Column("value_max", sa.Float(), nullable=False, server_default="0"),
+        sa.Column("sample_count", sa.Integer(), nullable=False, server_default="0"),
+        sa.PrimaryKeyConstraint("id"),
+        sa.UniqueConstraint("source_module", "resource_name", "metric_name", "bucket",
+                            name="uq_plugin_metrics_hourly_bucket"),
+    )
+    op.create_index("ix_plugin_metrics_hourly_lookup", "plugin_metrics_hourly",
+                    ["source_module", "resource_name", "metric_name", "bucket"])
+
+
+def downgrade() -> None:
+    op.drop_index("ix_plugin_metrics_hourly_lookup", table_name="plugin_metrics_hourly")
+    op.drop_table("plugin_metrics_hourly")

steward/models/metrics.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 import uuid
 from datetime import datetime, timezone
-from sqlalchemy import DateTime, Float, String
+from sqlalchemy import DateTime, Float, Index, Integer, String, UniqueConstraint
 from sqlalchemy.orm import Mapped, mapped_column
 from .base import Base
 
@@ -17,3 +17,41 @@ class PluginMetric(Base):
     recorded_at: Mapped[datetime] = mapped_column(
         DateTime(timezone=True), nullable=False, default=lambda: datetime.now(timezone.utc)
     )
+
+    # This time-series table grows by (sources × resources × sample cadence); every
+    # host-detail / full-metrics / dashboard-widget read filters by
+    # (source_module, resource_name) over a recorded_at range. Without these it's a
+    # full sequential scan on every load.
+    __table_args__ = (
+        Index("ix_plugin_metrics_module_resource_recorded",
+              "source_module", "resource_name", "recorded_at"),
+        Index("ix_plugin_metrics_module_resource_metric_recorded",
+              "source_module", "resource_name", "metric_name", "recorded_at"),
+    )
+
+
+class PluginMetricHourly(Base):
+    """Hourly rollup of plugin_metrics — the coarse series that retention rolls
+    raw samples into before pruning them, so multi-day/week history stays cheap.
+
+    One row per (source_module, resource_name, metric_name, hour bucket); the
+    unique constraint is the conflict target for the idempotent rollup upsert.
+    Charts read this for the part of a range older than the raw-retention window.
+    """
+    __tablename__ = "plugin_metrics_hourly"
+
+    id: Mapped[str] = mapped_column(String(36), primary_key=True, default=lambda: str(uuid.uuid4()))
+    source_module: Mapped[str] = mapped_column(String(64), nullable=False)
+    resource_name: Mapped[str] = mapped_column(String(255), nullable=False)
+    metric_name: Mapped[str] = mapped_column(String(128), nullable=False)
+    bucket: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False)
+    value_avg: Mapped[float] = mapped_column(Float, nullable=False, default=0.0)
+    value_max: Mapped[float] = mapped_column(Float, nullable=False, default=0.0)
+    sample_count: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+
+    __table_args__ = (
+        UniqueConstraint("source_module", "resource_name", "metric_name", "bucket",
+                         name="uq_plugin_metrics_hourly_bucket"),
+        Index("ix_plugin_metrics_hourly_lookup",
+              "source_module", "resource_name", "metric_name", "bucket"),
+    )

steward/settings/routes.py

@@ -133,6 +133,8 @@ _RETENTION_FIELDS = [
     ("docker_metrics_raw_days", "docker.retention.metrics_raw_days"),
     ("docker_metrics_rollup_days", "docker.retention.metrics_rollup_days"),
     ("docker_events_days", "docker.retention.events_days"),
+    ("metrics_raw_days", "metrics.retention.raw_days"),
+    ("metrics_rollup_days", "metrics.retention.rollup_days"),
 ]
 
 

steward/templates/alerts/maintenance.html

@@ -6,7 +6,6 @@
 <div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:1.5rem;">
   <h1 class="page-title" style="margin-bottom:0;">Maintenance Windows</h1>
   <div style="display:flex;gap:0.5rem;">
-    <a class="btn btn-ghost" href="/alerts/">← Alerts</a>
     <a class="btn" href="/alerts/maintenance/new">New Window</a>
   </div>
 </div>

steward/templates/ansible/_playbook_options.html

@@ -1,6 +1,7 @@
 {# <option> list for a playbook <select>, populated when a source is chosen.
-   Returned by /ansible/playbook-options. `selected` pre-selects one (edit). #}
-<option value="">— choose playbook —</option>
+   Returned by /ansible/playbook-options. `selected` pre-selects one; with none
+   given the browser selects the first, so the dropdown never sits on a
+   placeholder. #}
 {% for pb in playbooks %}
 <option value="{{ pb }}" {% if pb == selected %}selected{% endif %}>{{ pb }}</option>
 {% endfor %}

steward/templates/ansible/_playbook_vars.html

@@ -1,6 +1,11 @@
 {# Description + discovered variable fields for a selected playbook. Shared by
    the browse run form, host run form, and schedule form. Expects `variables`
-   and `description`. Rendered standalone by /ansible/playbook-vars. #}
+   and `description`. Rendered standalone by /ansible/playbook-vars.
+   Optional: `values` (dict var-name→saved value, prefills fields for edit
+   forms) and `schedule` (True in the schedule form — schedules can't prompt,
+   so secret vars are shown disabled and the confirm gate isn't `required`). #}
+{% set _values = values|default({}) %}
+{% set _schedule = schedule|default(false) %}
 {% if description or category %}
 <div style="background:var(--bg);border-left:3px solid var(--accent);border-radius:3px;
             padding:0.5rem 0.7rem;margin:0.25rem 0 0.6rem;font-size:0.82rem;color:var(--text-muted);">
@@ -13,7 +18,7 @@
 <label style="display:flex;align-items:flex-start;gap:0.5rem;background:color-mix(in srgb,var(--red) 10%,var(--bg-elevated));
               border:1px solid color-mix(in srgb,var(--red) 35%,var(--border));border-radius:6px;
               padding:0.6rem 0.8rem;margin:0 0 0.6rem;font-size:0.82rem;font-weight:normal;cursor:pointer;">
-  <input type="checkbox" name="confirmed" required style="margin-top:0.15rem;">
+  <input type="checkbox" name="confirmed" {% if not _schedule %}required{% endif %} style="margin-top:0.15rem;">
   <span><strong style="color:var(--red);">Confirm:</strong> this playbook is marked as making
   significant or destructive changes. Tick to enable running it.</span>
 </label>
@@ -33,13 +38,22 @@
       ({{ v.name }}{% if v.required %}, required{% endif %}{% if v.secret %}, secret{% endif %})
     </span>
   </label>
+  {% if v.secret and _schedule %}
+  {# Scheduled runs can't prompt, so secret vars are dropped downstream — show
+     the field disabled rather than inviting silent data loss. #}
+  <input type="text" disabled
+         placeholder="secrets can't be scheduled — set via inventory / global creds"
+         style="width:100%;font-family:ui-monospace,monospace;font-size:0.85rem;opacity:0.55;">
+  {% else %}
   {% if v.secret %}<input type="hidden" name="secret__{{ v.name }}" value="1">{% endif %}
   <input type="{{ 'password' if v.secret else 'text' }}"
          name="var__{{ v.name }}"
+         {% if not v.secret %}value="{{ _values.get(v.name, '') }}"{% endif %}
          {% if v.required %}required{% endif %}
          {% if v.secret %}autocomplete="new-password"{% endif %}
          placeholder="{% if v.secret %}(hidden){% elif v.default not in (None, '') %}default: {{ v.default }}{% else %}no default{% endif %}"
          style="width:100%;font-family:ui-monospace,monospace;font-size:0.85rem;">
+  {% endif %}
 </div>
 {% endfor %}
 {% else %}

steward/templates/ansible/browse.html

@@ -17,7 +17,6 @@
       {% endif %}
       <a href="/ansible/inventory/targets" class="btn btn-ghost btn-sm">Inventory</a>
       <a href="/ansible/schedules" class="btn btn-ghost btn-sm">Schedules</a>
-      <a href="/ansible/" class="btn btn-ghost btn-sm">← Run History</a>
     </div>
 </div>
 
@@ -35,7 +34,6 @@
                 <button type="submit" class="btn btn-sm btn-danger">Delete</button>
             </form>
             {% endif %}
-            <a href="/ansible/browse" class="btn btn-ghost btn-sm">← Back to browse</a>
         </div>
     </div>
     <pre style="background:var(--bg);padding:1rem;border-radius:4px;overflow-x:auto;font-size:0.8rem;color:var(--text-muted);max-height:600px;overflow-y:auto;">{{ view_contents }}</pre>

steward/templates/ansible/inventory/group_detail.html

@@ -5,7 +5,6 @@
 {% block content %}
 <div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:1.5rem;">
     <h1 class="page-title" style="margin-bottom:0;">Group: {{ group.name }}</h1>
-    <a href="/ansible/inventory/groups" class="btn btn-ghost btn-sm">← Groups</a>
 </div>
 
 <form method="post" action="/ansible/inventory/groups/{{ group.id }}">

steward/templates/ansible/inventory/groups.html

@@ -7,7 +7,6 @@
     <h1 class="page-title" style="margin-bottom:0;">Inventory Groups</h1>
     <div style="display:flex;gap:0.5rem;">
         <a href="/ansible/inventory/targets" class="btn btn-ghost btn-sm">Targets</a>
-        <a href="/ansible/" class="btn btn-ghost btn-sm">← Ansible</a>
     </div>
 </div>
 

steward/templates/ansible/inventory/target_detail.html

@@ -5,7 +5,6 @@
 {% block content %}
 <div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:1.5rem;">
     <h1 class="page-title" style="margin-bottom:0;">Target: {{ target.name }}</h1>
-    <a href="/ansible/inventory/targets" class="btn btn-ghost btn-sm">← Targets</a>
 </div>
 
 <form method="post" action="/ansible/inventory/targets/{{ target.id }}">

steward/templates/ansible/inventory/targets.html

@@ -7,7 +7,6 @@
     <h1 class="page-title" style="margin-bottom:0;">Inventory Targets</h1>
     <div style="display:flex;gap:0.5rem;">
         <a href="/ansible/inventory/groups" class="btn btn-ghost btn-sm">Groups</a>
-        <a href="/ansible/" class="btn btn-ghost btn-sm">← Ansible</a>
     </div>
 </div>
 

steward/templates/ansible/playbook_editor.html

@@ -5,7 +5,6 @@
 {% block content %}
 <div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:1.25rem;gap:1rem;flex-wrap:wrap;">
   <h1 class="page-title" style="margin-bottom:0;">{{ "Edit playbook" if editing else "New playbook" }}</h1>
-  <a href="/ansible/browse" class="btn btn-ghost btn-sm">← Browse</a>
 </div>
 
 {% if error %}<div class="alert alert-error">{{ error }}</div>{% endif %}

steward/templates/ansible/run_detail.html

@@ -4,7 +4,6 @@
 {% block breadcrumb %}{{ crumbs([("Ansible", "/ansible/"), ("Run " ~ run.id[:8], "")]) }}{% endblock %}
 {% block content %}
 <div style="display:flex;align-items:center;gap:1rem;margin-bottom:1.5rem;">
-    <a href="/ansible/" class="btn btn-ghost btn-sm">← Runs</a>
     <h1 class="page-title" style="margin-bottom:0;">Run Detail</h1>
     <div style="margin-left:auto;display:flex;gap:0.5rem;">
         {% if run.status.value in ("running", "queued") and session.user_role in ("operator", "admin") %}

steward/templates/ansible/schedules.html

@@ -14,7 +14,6 @@
   <div style="display:flex;gap:0.5rem;">
     <a href="/ansible/inventory/targets" class="btn btn-ghost">Inventory</a>
     <a href="/ansible/browse" class="btn btn-ghost">Browse</a>
-    <a href="/ansible/" class="btn btn-ghost">← Runs</a>
   </div>
 </div>
 
@@ -76,6 +75,8 @@
 
 {# ── Create / edit form ───────────────────────────────────────────────────── #}
 {% set p = (editing.params or {}) if editing else {} %}
+{% set saved_vars = p.get('extra_vars_map') or {} %}
+{% set declared_names = (edit_variables or []) | map(attribute='name') | list %}
 <div class="card" id="form">
   <h2 class="section-title" style="margin-bottom:1rem;">{{ "Edit schedule" if editing else "New schedule" }}</h2>
   <form method="post" action="{{ '/ansible/schedules/' ~ editing.id if editing else '/ansible/schedules' }}">
@@ -88,18 +89,21 @@
         <label>Source</label>
         <select name="source_name" required
                 hx-get="/ansible/playbook-options" hx-trigger="change"
-                hx-target="#sch-playbook" hx-swap="innerHTML" hx-include="this">
-          <option value="">— choose source —</option>
+                hx-target="#sch-playbook" hx-swap="innerHTML" hx-include="this"
+                hx-on::after-settle="htmx.trigger('#sch-playbook','change')">
           {% for sd in source_data %}
-          <option value="{{ sd.name }}" {% if editing and editing.source_name == sd.name %}selected{% endif %}>{{ sd.name }}</option>
+          <option value="{{ sd.name }}" {% if sd.name == sel_source %}selected{% endif %}>{{ sd.name }}</option>
           {% endfor %}
+          {% if not source_data %}<option value="" disabled>No Ansible sources — add one in Settings</option>{% endif %}
         </select>
       </div>
       <div class="form-group" style="margin-bottom:0;">
         <label>Playbook</label>
-        <select name="playbook_path" id="sch-playbook" required>
-          <option value="">— choose playbook —</option>
-          {% for pb in all_playbooks %}
+        <select name="playbook_path" id="sch-playbook" required
+                hx-get="/ansible/playbook-vars?schedule=1"
+                hx-trigger="{% if not editing %}load, {% endif %}change"
+                hx-target="#sch-playbook-vars" hx-swap="innerHTML" hx-include="closest form">
+          {% for pb in sel_playbooks %}
           <option value="{{ pb }}" {% if editing and editing.playbook_path == pb %}selected{% endif %}>{{ pb }}</option>
           {% endfor %}
         </select>
@@ -133,11 +137,29 @@
         <input type="text" name="tags" value="{{ p.get('tags', '') }}" placeholder="tag1,tag2">
       </div>
     </div>
-    <div class="form-group" style="margin-top:1rem;">
-      <label>Extra vars (one key=value per line)</label>
-      <textarea name="extra_vars" rows="3" placeholder="prune_volumes=false">{% for k, v in (p.get('extra_vars_map') or {}).items() %}{{ k }}={{ v }}
-{% endfor %}</textarea>
+
+    {# Declared playbook variables (vars:/vars_prompt:) — loaded via HTMX on
+       playbook change; pre-rendered here in edit mode so saved values show. #}
+    <div id="sch-playbook-vars" style="margin-top:1rem;">
+      {% if editing %}
+      {% set variables = edit_variables %}
+      {% set description = edit_description %}
+      {% set category = edit_category %}
+      {% set confirm = edit_confirm %}
+      {% set values = saved_vars %}
+      {% set schedule = true %}
+      {% include "ansible/_playbook_vars.html" %}
+      {% endif %}
     </div>
+
+    <details style="margin-top:0.75rem;">
+      <summary style="cursor:pointer;font-size:0.82rem;color:var(--text-muted);">Extra vars</summary>
+      <div class="form-group" style="margin-top:0.6rem;">
+        <label>Extra vars <span style="color:var(--text-muted);font-weight:normal;">(one key=value per line — for vars not listed above)</span></label>
+        <textarea name="extra_vars" rows="3" placeholder="prune_volumes=false">{% for k, v in saved_vars.items() if k not in declared_names %}{{ k }}={{ v }}
+{% endfor %}</textarea>
+      </div>
+    </details>
     <div class="form-group" style="display:flex;align-items:center;gap:0.5rem;">
       <input type="checkbox" name="check" id="check" style="width:auto;" {% if p.get('check') %}checked{% endif %}>
       <label for="check" style="margin-bottom:0;">Dry run (--check --diff)</label>

steward/templates/base.html

@@ -51,37 +51,90 @@ code { font-family: ui-monospace, monospace; font-size: 0.875em; background: var
     radial-gradient(ellipse 50% 40% at 20% 80%, rgba(200, 168, 64, 0.025) 0%, transparent 65%);
 }
 
-/* Nav */
-nav {
-  background: var(--bg-card);
-  padding: 0 1.5rem;
-  display: flex;
-  align-items: center;
-  gap: 0;
-  border-bottom: 1px solid var(--border-mid);
-  height: 48px;
-  position: relative;
-  z-index: 10;
+/* ── App shell: persistent left sidebar + content column ───────────────────── */
+.app-shell { display: flex; min-height: 100vh; position: relative; z-index: 1; }
+.sidebar {
+  width: 220px; flex-shrink: 0; background: var(--bg-card);
+  border-right: 1px solid var(--border-mid);
+  display: flex; flex-direction: column;
+  position: sticky; top: 0; height: 100vh; z-index: 20;
 }
-nav .brand {
-  font-family: var(--font-serif);
-  font-weight: 700;
-  font-size: 1.1rem;
-  margin-right: 2rem;
-  letter-spacing: 0.01em;
-  display: flex;
-  align-items: center;
-  gap: 0.45rem;
-  color: var(--gold);
-  text-decoration: none;
+.sidebar .brand {
+  font-family: var(--font-serif); font-weight: 700; font-size: 1.15rem;
+  letter-spacing: 0.01em; color: var(--gold); text-decoration: none;
+  display: flex; align-items: center; gap: 0.5rem;
+  padding: 0.85rem 1.25rem; border-bottom: 1px solid var(--border);
 }
-nav .brand svg { flex-shrink: 0; }
-nav a { color: var(--text-muted); font-size: 0.875rem; padding: 0 0.875rem; height: 48px; display: flex; align-items: center; border-bottom: 2px solid transparent; transition: color .15s, border-color .15s; }
-nav a:hover { color: var(--text); border-bottom-color: var(--border-mid); }
-nav a.nav-end { margin-left: auto; }
+.sidebar .brand svg { flex-shrink: 0; }
+/* Only the nav list scrolls when long — the brand and user/logout stay pinned. */
+.side-nav { flex: 1; min-height: 0; overflow-y: auto; padding: 0.75rem 0; }
+.nav-group { margin-bottom: 0.9rem; }
+.nav-group-label {
+  font-size: 0.66rem; text-transform: uppercase; letter-spacing: 0.09em;
+  color: var(--text-dim); font-weight: 600; padding: 0 1.25rem; margin-bottom: 0.3rem;
+}
+.side-nav a {
+  display: flex; align-items: center; gap: 0.5rem;
+  padding: 0.4rem 1.25rem; font-size: 0.9rem; color: var(--text-muted);
+  border-left: 2px solid transparent; text-decoration: none;
+  transition: color .12s, background .12s, border-color .12s;
+}
+.side-nav a:hover { color: var(--text); background: var(--bg-elevated); }
+.side-nav a.active {
+  color: var(--gold); border-left-color: var(--gold);
+  background: color-mix(in srgb, var(--gold) 8%, transparent);
+}
+.side-user { border-top: 1px solid var(--border); padding: 0.7rem 1.25rem; }
+.side-user a {
+  color: var(--text-muted); font-size: 0.84rem; text-decoration: none;
+  display: flex; align-items: center; gap: 0.4rem;
+}
+.side-user a:hover { color: var(--text); }
+/* Keyboard focus: a clear inset ring on every interactive sidebar element. */
+.sidebar .brand:focus-visible,
+.side-nav a:focus-visible,
+.side-user a:focus-visible {
+  outline: 2px solid var(--accent); outline-offset: -2px; color: var(--text);
+}
+.app-main { flex: 1; min-width: 0; display: flex; flex-direction: column; }
+.topbar { display: none; }   /* slim mobile bar; shown only under 900px */
+.nav-scrim { display: none; }
 
 /* Layout */
-main { padding: 1.5rem 2rem; max-width: 1600px; margin: 0 auto; width: 100%; box-sizing: border-box; position: relative; z-index: 1; }
+main { padding: 1.5rem 2rem; max-width: 1400px; margin: 0 auto; width: 100%; box-sizing: border-box; position: relative; z-index: 1; }
+
+/* Responsive: sidebar slides off-canvas under 900px, toggled by ☰ */
+@media (max-width: 900px) {
+  .sidebar {
+    position: fixed; left: 0; top: 0; bottom: 0; height: 100%;
+    transform: translateX(-100%); transition: transform .2s ease;
+    box-shadow: 0 0 30px rgba(0,0,0,0.5);
+  }
+  body.nav-open .sidebar { transform: translateX(0); }
+  .topbar {
+    display: flex; align-items: center; gap: 0.75rem; height: 48px; padding: 0 1rem;
+    background: var(--bg-card); border-bottom: 1px solid var(--border-mid);
+    position: sticky; top: 0; z-index: 15;
+  }
+  .sidebar-toggle {
+    background: none; border: none; color: var(--text-muted); font-size: 1.3rem;
+    cursor: pointer; line-height: 1; padding: 0.25rem; border-radius: 4px;
+  }
+  .sidebar-toggle:hover { color: var(--text); }
+  .sidebar-toggle:focus-visible { outline: 2px solid var(--accent); outline-offset: 2px; }
+  .topbar .brand-sm {
+    font-family: var(--font-serif); color: var(--gold); font-weight: 700;
+    font-size: 1.05rem; text-decoration: none;
+  }
+  body.nav-open .nav-scrim {
+    display: block; position: fixed; inset: 0; background: rgba(0,0,0,0.5); z-index: 18;
+  }
+}
+
+@media (prefers-reduced-motion: reduce) {
+  .sidebar { transition: none; }
+  .side-nav a { transition: none; }
+}
 
 /* Alerts */
 .alert { padding: 0.75rem 1rem; border-radius: 6px; margin-bottom: 1rem; font-size: 0.9rem; }
@@ -224,8 +277,10 @@ body.dash-editing .widget-drawer { transform:translateY(0); }
 
 <div id="candle-glow"></div>
 
+{% set _p = request.path %}
+<div class="app-shell">
 {% if session.user_id is defined %}
-<nav>
+  <aside class="sidebar" id="sidebar">
     <a href="/" class="brand">
       <svg width="22" height="22" viewBox="0 0 24 24" fill="none" xmlns="http://www.w3.org/2000/svg" aria-hidden="true">
         <circle cx="12" cy="12" r="10.5" fill="none" stroke="#c8a840" stroke-width="1.5"/>
@@ -237,18 +292,48 @@ body.dash-editing .widget-drawer { transform:translateY(0); }
       </svg>
       Steward
     </a>
-    <a href="/">Dashboard</a>
-    <a href="/status">Status</a>
-    <a href="/hosts/">Hosts</a>
-    <a href="/monitors/">Monitors</a>
-    <a href="/alerts/">Alerts</a>
-    <a href="/ansible/">Ansible</a>
-    {% if session.user_role == 'admin' %}
-    <a href="/settings/">Settings</a>
-    <a href="/audit/">Audit</a>
-    {% endif %}
-    <a href="/logout" class="nav-end">{{ session.username }}</a>
-</nav>
+    <nav class="side-nav" aria-label="Primary">
+      <div class="nav-group">
+        <div class="nav-group-label">Overview</div>
+        <a href="/" {% if _p == '/' %}class="active" aria-current="page"{% endif %}>Dashboard</a>
+        <a href="/status" {% if _p.startswith('/status') %}class="active" aria-current="page"{% endif %}>Status</a>
+      </div>
+      <div class="nav-group">
+        <div class="nav-group-label">Infrastructure</div>
+        <a href="/hosts/" {% if _p.startswith('/hosts') %}class="active" aria-current="page"{% endif %}>Hosts</a>
+        {% for item in plugin_nav %}
+        <a href="{{ item.href }}" {% if _p.startswith(item.href.rstrip('/')) %}class="active" aria-current="page"{% endif %}>{{ item.label }}</a>
+        {% endfor %}
+      </div>
+      <div class="nav-group">
+        <div class="nav-group-label">Monitoring</div>
+        <a href="/monitors/" {% if _p.startswith('/monitors') %}class="active" aria-current="page"{% endif %}>Monitors</a>
+        <a href="/alerts/" {% if _p.startswith('/alerts') %}class="active" aria-current="page"{% endif %}>Alerts</a>
+      </div>
+      <div class="nav-group">
+        <div class="nav-group-label">Automation</div>
+        <a href="/ansible/" {% if _p.startswith('/ansible') %}class="active" aria-current="page"{% endif %}>Ansible</a>
+      </div>
+      {% if session.user_role == 'admin' %}
+      <div class="nav-group">
+        <div class="nav-group-label">Admin</div>
+        <a href="/settings/" {% if _p.startswith('/settings') %}class="active" aria-current="page"{% endif %}>Settings</a>
+        <a href="/audit/" {% if _p.startswith('/audit') %}class="active" aria-current="page"{% endif %}>Audit</a>
+      </div>
+      {% endif %}
+    </nav>
+    <div class="side-user">
+      <a href="/logout" title="Log out">⏻ {{ session.username }}</a>
+    </div>
+  </aside>
+{% endif %}
+  <div class="app-main">
+{% if session.user_id is defined %}
+    <div class="topbar">
+      <button class="sidebar-toggle" type="button" aria-label="Toggle navigation"
+              aria-controls="sidebar" aria-expanded="false" onclick="toggleNav(this)">☰</button>
+      <a href="/" class="brand-sm">Steward</a>
+    </div>
 {% endif %}
 
 <script>
@@ -263,6 +348,16 @@ function setTimeRange(val) {
   history.replaceState({}, '', url);
   document.body.dispatchEvent(new CustomEvent('rangeChange'));
 }
+// Mobile sidebar (off-canvas). Keep aria-expanded in sync; Escape closes it.
+function setNavOpen(open) {
+  document.body.classList.toggle('nav-open', open);
+  var btn = document.querySelector('.sidebar-toggle');
+  if (btn) btn.setAttribute('aria-expanded', open ? 'true' : 'false');
+}
+function toggleNav(btn) { setNavOpen(!document.body.classList.contains('nav-open')); }
+document.addEventListener('keydown', function (e) {
+  if (e.key === 'Escape' && document.body.classList.contains('nav-open')) setNavOpen(false);
+});
 </script>
 
     <main>
@@ -280,12 +375,45 @@ function setTimeRange(val) {
             </span>
         </div>
         {% endif %}
+        {% if undecryptable_secrets and session.user_role == 'admin' %}
+        {% set _sec_tab = {
+          'smtp.password': '/settings/notifications/',
+          'oidc.client_secret': '/settings/auth/',
+          'ldap.bind_password': '/settings/auth/',
+          'ansible.ssh_private_key': '/settings/ansible/',
+          'ansible.become_password': '/settings/ansible/',
+          'ansible.vault_password': '/settings/ansible/',
+        } %}
+        <div style="background:color-mix(in srgb,var(--red) 12%,var(--bg-elevated));
+                    border:1px solid color-mix(in srgb,var(--red) 35%,var(--border));
+                    border-radius:6px;padding:0.6rem 1rem;margin-bottom:1rem;
+                    font-size:0.84rem;display:flex;align-items:flex-start;gap:0.6rem;">
+            <span style="color:var(--red);flex-shrink:0;">⚠</span>
+            <span>
+                <strong style="color:var(--text);">
+                    {{ undecryptable_secrets | length }} stored secret{{ 's' if undecryptable_secrets | length != 1 }}
+                    can’t be decrypted
+                </strong>
+                — the app secret key changed, so {{ 'they' if undecryptable_secrets | length != 1 else 'it' }}
+                must be re-entered:
+                {% for k in undecryptable_secrets -%}
+                    {%- if _sec_tab.get(k) %}<a href="{{ _sec_tab[k] }}" style="color:var(--accent);">{{ k }}</a>
+                    {%- else %}<code>{{ k }}</code>{% endif %}{% if not loop.last %}, {% endif %}
+                {%- endfor %}.
+            </span>
+        </div>
+        {% endif %}
         {% if error %}
         <div class="alert alert-error">{{ error }}</div>
         {% endif %}
         {% block breadcrumb %}{% endblock %}
         {% block content %}{% endblock %}
     </main>
+  </div>{# .app-main #}
+</div>{# .app-shell #}
+{% if session.user_id is defined %}
+<div class="nav-scrim" onclick="setNavOpen(false)"></div>
+{% endif %}
 
 {% block extra_scripts %}{% endblock %}
 

steward/templates/hosts/detail.html

@@ -26,8 +26,17 @@
 {% set inp = "width:100%;padding:0.4rem 0.65rem;background:var(--bg);border:1px solid var(--border-mid);border-radius:4px;color:var(--text);font-size:0.88rem;" %}
 {% set lbl = "font-size:0.75rem;color:var(--text-muted);display:block;margin-bottom:0.2rem;" %}
 
+{# ── Live vitals strip (host_agent fragment) — full width on top, polled ────── #}
+<div id="hv-strip"
+     hx-get="/plugins/host_agent/vitals/{{ host.id }}"
+     hx-trigger="load, every 15s"
+     hx-swap="innerHTML"></div>
+
+{# ── Monitors + Agent, side by side on wide screens, stacked when narrow ────── #}
+<div style="display:grid;grid-template-columns:repeat(auto-fit,minmax(420px,1fr));gap:1rem;align-items:start;margin-bottom:1rem;">
+
 {# ── Monitors ─────────────────────────────────────────────────────────────── #}
-<div class="card">
+<div class="card" style="margin-bottom:0;">
   <div style="display:flex;align-items:baseline;justify-content:space-between;margin-bottom:0.5rem;">
     <h3 class="section-title" style="margin-bottom:0;">Monitors</h3>
     {% if uptime %}
@@ -114,20 +123,15 @@
     </form>
   </details>
   {% endif %}
+</div>{# Monitors card #}
+
+{# ── Agent (host_agent management panel) ──────────────────────────────────── #}
+<div hx-get="/plugins/host_agent/panel/{{ host.id }}" hx-trigger="load" hx-swap="innerHTML">
+  <div class="card" style="margin-bottom:0;"><span style="color:var(--text-muted);font-size:0.85rem;">Loading agent…</span></div>
 </div>
+</div>{# Monitors + Agent grid #}
 
-{# ── Agent (host_agent plugin fragment, embedded across the plugin boundary) ── #}
-<div hx-get="/plugins/host_agent/panel/{{ host.id }}" hx-trigger="load"
-     hx-swap="innerHTML">
-  <div class="card"><span style="color:var(--text-muted);font-size:0.85rem;">Loading agent…</span></div>
-</div>
-
-{# ── Docker (docker plugin fragment; renders nothing if the host has none) ──── #}
-{% if "docker" in enabled_plugins %}
-<div hx-get="/plugins/docker/host/{{ host.id }}" hx-trigger="load" hx-swap="innerHTML"></div>
-{% endif %}
-
-{# ── Ansible ──────────────────────────────────────────────────────────────── #}
+{# ── Ansible (full width) ─────────────────────────────────────────────────── #}
 <div class="card">
   <h3 class="section-title">Ansible</h3>
   {% if linked_target %}
@@ -157,9 +161,9 @@
       <div class="form-group">
         <label>Source</label>
         <select name="source_name" required
-                hx-get="/ansible/playbook-options" hx-trigger="change"
-                hx-target="#hp-playbook" hx-swap="innerHTML" hx-include="this">
-          <option value="">— choose source —</option>
+                hx-get="/ansible/playbook-options" hx-trigger="load, change"
+                hx-target="#hp-playbook" hx-swap="innerHTML" hx-include="this"
+                hx-on::after-settle="htmx.trigger('#hp-playbook','change')">
           {% for s in ansible_sources %}<option value="{{ s }}">{{ s }}</option>{% endfor %}
         </select>
       </div>
@@ -168,7 +172,6 @@
         <select name="playbook_path" id="hp-playbook" required
                 hx-get="/ansible/playbook-vars" hx-trigger="change"
                 hx-target="#hp-vars" hx-swap="innerHTML" hx-include="closest form">
-          <option value="">— choose a source first —</option>
         </select>
       </div>
       <div id="hp-vars"></div>
@@ -213,7 +216,17 @@
     </form>
   </div>
   {% endif %}
-</div>
+</div>{# Ansible card #}
+
+{# ── Docker (docker plugin fragment; renders nothing if the host has none) ──── #}
+{% if "docker" in enabled_plugins %}
+<div hx-get="/plugins/docker/host/{{ host.id }}" hx-trigger="load" hx-swap="innerHTML"></div>
+{% endif %}
+
+{# ── SNMP (renders nothing unless a configured SNMP device maps to this host) ── #}
+{% if "snmp" in enabled_plugins %}
+<div hx-get="/plugins/snmp/host/{{ host.id }}" hx-trigger="load" hx-swap="innerHTML"></div>
+{% endif %}
 
 {{ toggle_script() }}
 <script>var _mt=document.getElementById('mtype'); if(_mt) mtoggle(_mt.value);</script>

steward/templates/hosts/uptime.html

@@ -1,9 +1,10 @@
 {% extends "base.html" %}
+{% from "_macros.html" import crumbs %}
 {% block title %}Uptime / SLA — Steward{% endblock %}
+{% block breadcrumb %}{{ crumbs([("Hosts", "/hosts/"), ("Uptime / SLA", "")]) }}{% endblock %}
 {% block content %}
 <div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:1.5rem;">
   <h1 class="page-title" style="margin-bottom:0;">Uptime / SLA</h1>
-  <a class="btn btn-ghost" href="/hosts/">← Hosts</a>
 </div>
 
 {# ── Summary pills ──────────────────────────────────────────────────────────── #}

steward/templates/monitors/edit.html

@@ -1,10 +1,11 @@
 {% extends "base.html" %}
+{% from "_macros.html" import crumbs %}
 {% from "monitors/_fields.html" import type_fields, toggle_script %}
 {% block title %}Edit Monitor — Steward{% endblock %}
+{% block breadcrumb %}{{ crumbs([("Monitors", "/monitors/"), ("Edit monitor", "")]) }}{% endblock %}
 {% block content %}
 <div style="margin-bottom:1.25rem;">
-  <a href="/monitors/" style="color:var(--text-muted);font-size:0.85rem;">← Monitors</a>
-  <h1 class="page-title" style="margin:0.4rem 0 0;">Edit Monitor</h1>
+  <h1 class="page-title" style="margin-bottom:0;">Edit Monitor</h1>
 </div>
 
 {% set inp = "width:100%;padding:0.4rem 0.65rem;background:var(--bg);border:1px solid var(--border-mid);border-radius:4px;color:var(--text);font-size:0.88rem;" %}

steward/templates/settings/_tabs.html

@@ -1,5 +1,6 @@
-{# settings/_tabs.html — include at top of each settings section #}
-<h1 class="page-title">Settings</h1>
+{# settings/_tabs.html — include at top of each settings section.
+   No section <h1>: the breadcrumb kicker ("Settings › …") names the section and
+   the tab strip is the visual header, so we don't stack a redundant title. #}
 <div style="display:flex;gap:0;border-bottom:1px solid var(--border-mid);margin-bottom:1.5rem;">
   {% set tabs = [
     ("general",       "General",        "/settings/general/"),

steward/templates/settings/plugin_detail.html

@@ -8,7 +8,6 @@
 {% include "settings/_tabs.html" %}
 
 <div style="display:flex;align-items:center;gap:1rem;margin-bottom:1.5rem;">
-  <a href="/settings/plugins/" class="btn btn-ghost btn-sm">← Plugins</a>
   <h1 class="page-title" style="margin-bottom:0;">{{ plugin.get('name', plugin._dir) }}</h1>
   <span style="color:var(--text-muted);font-size:0.82rem;">v{{ plugin.get('version', '?') }}</span>
   {% if fail_reason %}

steward/templates/settings/thresholds.html

@@ -70,6 +70,21 @@
           "Keep container start/stop/die/health history this long.") }}
 </div>
 
+<div class="card" style="max-width:640px;margin-top:1rem;">
+  <h2 class="section-title" style="margin-bottom:0.5rem;">Host metrics retention</h2>
+  <p style="font-size:0.82rem;color:var(--text-muted);margin-bottom:1.25rem;">
+    Bounds how much host-agent metric history is stored (CPU, memory, disk, network,
+    temps, …). Raw per-sample points are kept for the raw window, then rolled up
+    into hourly averages kept for the rollup window. Host charts read raw for the
+    recent part of a range and hourly for older data. Applied by the hourly cleanup.
+  </p>
+
+  {{ days("Raw metrics", "metrics_raw_days", "metrics.retention.raw_days",
+          "Keep per-sample host metrics this long, then roll up to hourly averages.") }}
+  {{ days("Rolled-up metrics", "metrics_rollup_days", "metrics.retention.rollup_days",
+          "Keep the hourly-averaged series this long for multi-week history.") }}
+</div>
+
 <div style="margin-top:1rem;display:flex;align-items:center;gap:1rem;">
   <button type="submit" class="btn">Save</button>
   <span style="font-size:0.82rem;color:var(--text-muted);">Takes effect immediately — no restart.</span>

tests/core/test_fd_stability.py

@@ -0,0 +1,46 @@
+"""Real-fd regression canary.
+
+Drives a real socket code path (the TCP reachability probe) in a tight loop and
+asserts the process's open-fd count doesn't grow. If someone reintroduces a
+socket leak in the probe path — the class of bug behind the Errno 24 lockups —
+this fails in CI instead of in production. Uses real descriptors, not fakes, so
+the assertion has teeth.
+"""
+import asyncio
+import os
+
+import pytest
+
+from steward.monitors.ping import tcp_check
+
+_ITERATIONS = 100
+
+
+def _fd_count() -> int | None:
+    try:
+        return len(os.listdir("/proc/self/fd"))
+    except OSError:
+        return None
+
+
+async def _hammer_tcp_check() -> None:
+    # 127.0.0.1:1 has no listener → connection refused immediately. Each call must
+    # fully release its socket; a leak would add ~one fd per iteration.
+    for _ in range(_ITERATIONS):
+        await tcp_check("127.0.0.1", 1)
+
+
+def test_tcp_check_does_not_leak_fds():
+    if _fd_count() is None:
+        pytest.skip("/proc/self/fd unavailable on this platform")
+
+    asyncio.run(_hammer_tcp_check())   # warm up (lazy imports, caches)
+    before = _fd_count()
+    asyncio.run(_hammer_tcp_check())
+    after = _fd_count()
+
+    # Small slack for interpreter-internal fds; a genuine leak over 100 iterations
+    # would be far larger than this.
+    assert after - before <= 5, (
+        f"open fds grew {before}->{after} over {_ITERATIONS} tcp_check calls "
+        "— possible socket leak")

tests/core/test_metrics_retention.py

@@ -0,0 +1,17 @@
+"""Unit tests for the plugin_metrics rollup cutoff helpers (no DB)."""
+from datetime import datetime, timezone
+
+from steward.core.metrics_retention import _hour_floor, _rollup_cutoff
+
+
+def test_hour_floor_drops_sub_hour():
+    dt = datetime(2026, 6, 20, 15, 42, 9, 123456, tzinfo=timezone.utc)
+    assert _hour_floor(dt) == datetime(2026, 6, 20, 15, 0, 0, 0, tzinfo=timezone.utc)
+
+
+def test_rollup_cutoff_is_hour_aligned_and_offset():
+    now = datetime(2026, 6, 20, 15, 42, 9, tzinfo=timezone.utc)
+    cutoff = _rollup_cutoff(now, 7)
+    assert (cutoff.minute, cutoff.second, cutoff.microsecond) == (0, 0, 0)
+    # 7 whole days back, then floored to the hour.
+    assert cutoff == datetime(2026, 6, 13, 15, 0, 0, 0, tzinfo=timezone.utc)

tests/core/test_plugin_nav.py

@@ -0,0 +1,57 @@
+"""Unit tests for plugin-contributed sidebar nav collection."""
+import types
+
+from steward.core import plugin_manager as pm
+
+
+def _mod(nav):
+    """A stand-in plugin module; omit get_nav by passing nav=None."""
+    m = types.SimpleNamespace()
+    if nav is not None:
+        m.get_nav = lambda: nav
+    return m
+
+
+def test_collect_and_get(monkeypatch):
+    monkeypatch.setattr(pm, "_PLUGIN_NAV", [])
+    pm._collect_plugin_nav("docker", _mod([{"label": "Docker", "href": "/plugins/docker/"}]))
+    assert pm.get_plugin_nav() == [
+        {"plugin": "docker", "label": "Docker", "href": "/plugins/docker/", "section": "Infrastructure"}
+    ]
+
+
+def test_missing_hook_is_noop(monkeypatch):
+    monkeypatch.setattr(pm, "_PLUGIN_NAV", [])
+    pm._collect_plugin_nav("x", _mod(None))
+    assert pm.get_plugin_nav() == []
+
+
+def test_reload_replaces_same_plugin_entries(monkeypatch):
+    monkeypatch.setattr(pm, "_PLUGIN_NAV", [])
+    pm._collect_plugin_nav("docker", _mod([{"label": "Docker", "href": "/old"}]))
+    pm._collect_plugin_nav("docker", _mod([{"label": "Docker", "href": "/new"}]))
+    assert [e["href"] for e in pm.get_plugin_nav() if e["plugin"] == "docker"] == ["/new"]
+
+
+def test_malformed_item_skipped(monkeypatch):
+    monkeypatch.setattr(pm, "_PLUGIN_NAV", [])
+    pm._collect_plugin_nav("x", _mod([{"label": "NoHref"}, {"label": "Ok", "href": "/ok"}]))
+    assert [e["href"] for e in pm.get_plugin_nav()] == ["/ok"]
+
+
+def test_raising_hook_skipped(monkeypatch):
+    monkeypatch.setattr(pm, "_PLUGIN_NAV", [])
+
+    def boom():
+        raise RuntimeError("nope")
+
+    pm._collect_plugin_nav("x", types.SimpleNamespace(get_nav=boom))
+    assert pm.get_plugin_nav() == []
+
+
+def test_get_plugin_nav_sorted_by_section_then_label(monkeypatch):
+    monkeypatch.setattr(pm, "_PLUGIN_NAV", [])
+    pm._collect_plugin_nav("traefik", _mod([{"label": "Traefik", "href": "/t"}]))
+    pm._collect_plugin_nav("docker", _mod([{"label": "Docker", "href": "/d"}]))
+    # Same default section → sorted by label.
+    assert [e["label"] for e in pm.get_plugin_nav()] == ["Docker", "Traefik"]

tests/core/test_scheduler_overlap.py

@@ -0,0 +1,58 @@
+"""Unit tests for the scheduler's poll-overlap guard (_DueTracker).
+
+The tracker is pure (clock passed in) so we can assert the overlap policy
+deterministically: a task whose prior run is still in flight is never re-fired,
+and a skipped task isn't penalised — it runs on the next tick once it finishes.
+This is the rail that stops a hung poll from stacking overlapping runs (which
+would amplify any per-poll resource/fd use).
+"""
+from steward.core.scheduler import ScheduledTask, _DueTracker
+
+
+def _task(name: str, interval: int) -> ScheduledTask:
+    return ScheduledTask(name=name, coro_factory=lambda: None, interval_seconds=interval)
+
+
+def test_due_only_after_interval_elapses():
+    t = _task("a", 60)
+    tr = _DueTracker()
+    assert tr.due([t], now=59) == []      # 59 < 60
+    assert tr.due([t], now=60) == [t]     # interval elapsed
+
+
+def test_in_flight_task_is_skipped_even_when_due():
+    t = _task("a", 0)  # due every tick
+    tr = _DueTracker()
+    tr.mark_started(t, now=0)
+    assert tr.due([t], now=100) == []     # still running → skipped
+    tr.mark_done("a")
+    assert tr.due([t], now=100) == [t]    # finished → eligible again
+
+
+def test_skip_does_not_advance_last_run_so_it_retries():
+    t = _task("a", 10)
+    tr = _DueTracker()
+    tr.mark_started(t, now=0)
+    assert tr.due([t], now=100) == []     # due but in flight
+    assert tr.last_run["a"] == 0          # not advanced by the skip
+    tr.mark_done("a")
+    assert tr.due([t], now=100) == [t]    # retried promptly after completion
+
+
+def test_mark_started_advances_last_run_and_marks_in_flight():
+    t = _task("a", 10)
+    tr = _DueTracker()
+    tr.mark_started(t, now=50)
+    assert tr.last_run["a"] == 50
+    assert tr.in_flight == {"a"}
+    assert tr.due([t], now=55) == []      # 5 < 10 (not yet due)
+    assert tr.due([t], now=61) == []      # due by interval, but still in flight
+    tr.mark_done("a")
+    assert tr.due([t], now=61) == [t]
+
+
+def test_independent_tasks_do_not_block_each_other():
+    a, b = _task("a", 0), _task("b", 0)
+    tr = _DueTracker()
+    tr.mark_started(a, now=0)             # a hangs
+    assert tr.due([a, b], now=10) == [b]  # b still fires

tests/core/test_self_monitor.py

@@ -0,0 +1,115 @@
+"""Unit tests for the self-fd watchdog (no DB, no network).
+
+Exercises the fd accounting, the percentage/warning floor, and that the
+recorder degrades to a no-op where fd accounting isn't available. The DB session
+and the alert pipeline are faked — we only assert that the right metrics get
+handed to record_metric.
+"""
+import asyncio
+import logging
+
+import steward.core.alerts as alerts
+import steward.core.self_monitor as sm
+
+
+# ── fd accounting ────────────────────────────────────────────────────────────
+
+def test_count_open_fds_sane_or_none():
+    n = sm.count_open_fds()
+    # On the Linux CI image /proc exists; stdin/stdout/stderr are always open.
+    assert n is None or n >= 3
+
+
+def test_fd_soft_limit_positive_or_none():
+    soft = sm.fd_soft_limit()
+    assert soft is None or soft > 0
+
+
+# ── fake session / app plumbing ──────────────────────────────────────────────
+
+class _Ctx:
+    async def __aenter__(self):
+        return self
+
+    async def __aexit__(self, *_):
+        return False
+
+
+class _FakeSession(_Ctx):
+    def begin(self):
+        return _Ctx()
+
+
+class _FakeApp:
+    def db_sessionmaker(self):
+        return _FakeSession()
+
+
+def _patch_recorder(monkeypatch):
+    recorded: list[tuple] = []
+
+    async def fake_record_metric(session, source_module, resource_name, metric_name, value):
+        recorded.append((source_module, resource_name, metric_name, value))
+
+    monkeypatch.setattr(alerts, "record_metric", fake_record_metric)
+    return recorded
+
+
+# ── record_self_metrics ──────────────────────────────────────────────────────
+
+def test_records_both_fd_metrics(monkeypatch):
+    recorded = _patch_recorder(monkeypatch)
+    monkeypatch.setattr(sm, "count_open_fds", lambda: 50)
+    monkeypatch.setattr(sm, "fd_soft_limit", lambda: 1000)
+
+    asyncio.run(sm.record_self_metrics(_FakeApp()))
+
+    assert ("steward", "process", "open_fds", 50.0) in recorded
+    assert ("steward", "process", "open_fds_pct", 5.0) in recorded
+
+
+def test_no_pct_metric_when_limit_unknown(monkeypatch):
+    recorded = _patch_recorder(monkeypatch)
+    monkeypatch.setattr(sm, "count_open_fds", lambda: 50)
+    monkeypatch.setattr(sm, "fd_soft_limit", lambda: None)
+
+    asyncio.run(sm.record_self_metrics(_FakeApp()))
+
+    metric_names = [m for (_s, _r, m, _v) in recorded]
+    assert "open_fds" in metric_names
+    assert "open_fds_pct" not in metric_names   # meaningless without a ceiling
+
+
+def test_warns_past_floor(monkeypatch, caplog):
+    _patch_recorder(monkeypatch)
+    monkeypatch.setattr(sm, "count_open_fds", lambda: 900)
+    monkeypatch.setattr(sm, "fd_soft_limit", lambda: 1000)   # 90% ≥ FD_WARN_PCT
+    monkeypatch.setattr(sm, "_warned", False)
+
+    with caplog.at_level(logging.WARNING, logger=sm.logger.name):
+        asyncio.run(sm.record_self_metrics(_FakeApp()))
+
+    assert any("soft limit" in r.message for r in caplog.records)
+
+
+def test_no_warning_below_floor(monkeypatch, caplog):
+    _patch_recorder(monkeypatch)
+    monkeypatch.setattr(sm, "count_open_fds", lambda: 100)
+    monkeypatch.setattr(sm, "fd_soft_limit", lambda: 1000)   # 10%
+    monkeypatch.setattr(sm, "_warned", False)
+
+    with caplog.at_level(logging.WARNING, logger=sm.logger.name):
+        asyncio.run(sm.record_self_metrics(_FakeApp()))
+
+    assert not any("soft limit" in r.message for r in caplog.records)
+
+
+def test_noop_when_fd_count_unavailable(monkeypatch):
+    # /proc absent → must return before touching the DB.
+    monkeypatch.setattr(sm, "count_open_fds", lambda: None)
+
+    class _Boom:
+        def db_sessionmaker(self):
+            raise AssertionError("must not open a session when fds are unknown")
+
+    asyncio.run(sm.record_self_metrics(_Boom()))   # no raise

tests/core/test_undecryptable_secrets.py

@@ -0,0 +1,29 @@
+"""Unit tests for undecryptable-secret detection (the admin-banner source)."""
+from steward.core.crypto import encrypt_secret, init_crypto
+from steward.core.settings import _is_undecryptable, get_undecryptable_secrets
+
+
+def test_is_undecryptable_false_for_current_key():
+    init_crypto("key-A")
+    tok = encrypt_secret("hunter2")
+    assert _is_undecryptable(tok, "smtp.password") is False
+
+
+def test_is_undecryptable_true_after_key_rotation():
+    init_crypto("key-A")
+    tok = encrypt_secret("hunter2")          # sealed under key-A
+    init_crypto("key-B")                     # key changed/lost
+    assert _is_undecryptable(tok, "smtp.password") is True
+
+
+def test_is_undecryptable_ignores_plaintext_and_empty():
+    init_crypto("key-A")
+    assert _is_undecryptable("", "smtp.password") is False
+    assert _is_undecryptable("plain-value", "smtp.password") is False
+    assert _is_undecryptable(None, "smtp.password") is False
+
+
+def test_get_undecryptable_secrets_returns_sorted_list():
+    # Default (nothing scanned) is an empty list; the accessor is always safe to
+    # call from the template context processor.
+    assert isinstance(get_undecryptable_secrets(), list)

tests/integration/test_docker.py

@@ -151,6 +151,46 @@ def test_persist_scopes_containers_by_host(app):
 
 
 @_NEEDS_DB
+def test_large_memory_values_persist_as_bigint(app):
+    """A container using >2^31 bytes of RAM must persist. Regression: mem_usage_bytes
+    / mem_limit_bytes were int4 and overflowed (asyncpg 'value out of int32 range'),
+    failing the whole ingest batch for any host with a >2.1 GB container."""
+    from sqlalchemy import text
+    from steward.models.hosts import Host
+
+    persist = _persist_fn(app)
+    now = datetime.now(timezone.utc)
+    big_usage = 8_185_077_760    # ~8.18 GB — well past int4 max (2_147_483_647)
+    big_limit = 17_179_869_184   # 16 GB
+    snapshot = [(now, [{
+        "name": "bigmem", "container_id": "deadbeef", "image": "postgres",
+        "status": "running", "cpu_pct": 1.0, "mem_pct": 50.0,
+        "mem_usage_bytes": big_usage, "mem_limit_bytes": big_limit,
+        "ports": [], "started_at": None,
+    }])]
+
+    async def _go():
+        async with app.db_sessionmaker() as s:
+            async with s.begin():
+                await s.execute(text("DELETE FROM docker_metrics"))
+                await s.execute(text("DELETE FROM docker_containers"))
+                h = Host(id=str(uuid.uuid4()), name="bigmemhost", address="10.0.0.9")
+                s.add(h)
+                await s.flush()
+                await persist(s, h, snapshot)
+            row = (await s.execute(text(
+                "SELECT mem_usage_bytes, mem_limit_bytes FROM docker_containers "
+                "WHERE name = 'bigmem'"))).first()
+            metric = (await s.execute(text(
+                "SELECT mem_usage_bytes FROM docker_metrics "
+                "WHERE container_name = 'bigmem'"))).scalar()
+            return row, metric
+
+    row, metric = asyncio.run(_go())
+    assert row == (big_usage, big_limit)   # current-state row round-trips
+    assert metric == big_usage             # time-series row round-trips
+
+
 def test_lifecycle_events_derived_across_snapshots(app):
     from sqlalchemy import text
     from steward.models.hosts import Host
@@ -191,6 +231,44 @@ def test_lifecycle_events_derived_across_snapshots(app):
     assert not any(e[0] == "start" for e in events)
 
 
+@_NEEDS_DB
+def test_vanished_container_is_reaped(app):
+    """A container absent from the latest snapshot is deleted, not left behind as
+    a permanent stopped row. Regression: removed one-shot containers (CI jobs,
+    buildkit builders) accumulated forever and inflated the dashboard counts."""
+    from sqlalchemy import text
+    from steward.models.hosts import Host
+
+    persist = _persist_fn(app)
+    t1 = datetime(2026, 6, 19, 12, 0, 0, tzinfo=timezone.utc)
+    t2 = datetime(2026, 6, 19, 12, 0, 30, tzinfo=timezone.utc)
+    base = {"status": "running", "cpu_pct": 1.0, "mem_pct": 1.0,
+            "restart_count": 0, "exit_code": None, "oom_killed": False, "health": None}
+    keep = {**base, "name": "keep"}
+    gone = {**base, "name": "ci-job-123"}
+
+    async def _go():
+        async with app.db_sessionmaker() as s:
+            async with s.begin():
+                await s.execute(text("DELETE FROM docker_containers"))
+                h = Host(id=str(uuid.uuid4()), name="reap", address="10.9.9.8")
+                s.add(h)
+                await s.flush()
+                await persist(s, h, [(t1, [keep, gone])])   # both present
+                before = [r[0] for r in (await s.execute(text(
+                    "SELECT name FROM docker_containers WHERE host_id = :h ORDER BY name"),
+                    {"h": h.id})).all()]
+                await persist(s, h, [(t2, [keep])])         # ci-job-123 vanished
+                after = [r[0] for r in (await s.execute(text(
+                    "SELECT name FROM docker_containers WHERE host_id = :h ORDER BY name"),
+                    {"h": h.id})).all()]
+                return before, after
+
+    before, after = asyncio.run(_go())
+    assert before == ["ci-job-123", "keep"]
+    assert after == ["keep"]   # the vanished container was reaped
+
+
 @_NEEDS_DB
 def test_swarm_topology_persisted(app):
     from sqlalchemy import text
@@ -267,6 +345,8 @@ def test_disk_usage_persisted(app):
                 "FROM docker_disk_usage WHERE host_id=:h"), {"h": hid})).first()
             img_count = (await s.execute(text(
                 "SELECT COUNT(*) FROM docker_images WHERE host_id=:h"), {"h": hid})).scalar()
+            # End the autobegun read transaction before opening the next write one.
+            await s.rollback()
             # Re-report with one image dropped — the stale row must be pruned.
             async with s.begin():
                 await persist(s, await s.get(Host, hid), [], None, disk2)
@@ -369,3 +449,23 @@ def test_retention_rollup_and_prune(app):
     assert events_left == 1                    # only the in-window event survives
     assert counts["buckets_rolled"] == 1 and counts["raw_rows_rolled"] == 3
     assert counts["events_pruned"] == 1 and counts["rollup_pruned"] == 1
+
+
+def test_widget_dedup_collapses_cross_manager_duplicates():
+    """The same swarm task is reported by every manager (identical container_id);
+    the dashboard widget must count it once. Older agents send no container_id,
+    so those rows can't be matched and are all kept."""
+    from types import SimpleNamespace
+    from plugins.docker.dedup import dedup_by_container_id
+
+    def c(cid, name, status="running"):
+        return SimpleNamespace(container_id=cid, name=name, status=status)
+
+    rows = [c("abc", "web@m1"), c("abc", "web@m2"),  # same task, two managers
+            c("def", "db"), c("", "noid-1"), c("", "noid-2")]
+    out = dedup_by_container_id(rows)
+
+    assert len(out) == 4                                       # one "abc" dropped
+    abc = [r for r in out if r.container_id == "abc"]
+    assert len(abc) == 1 and abc[0].name == "web@m1"           # first occurrence wins
+    assert sum(1 for r in out if r.container_id == "") == 2    # id-less rows all kept

tests/integration/test_host_metrics.py

@@ -0,0 +1,161 @@
+"""Integration: host-metrics read paths (DISTINCT ON latest + SQL date_bin history).
+
+Validates the two host_agent query helpers that back the slow host views, against
+a live Postgres: the latest-per-(resource,metric) lookup and the bucket-averaged
+history (which now aggregates in SQL instead of shipping raw rows to Python).
+Requires STEWARD_DATABASE_URL.
+"""
+from __future__ import annotations
+
+import asyncio
+import os
+import uuid
+from datetime import datetime, timedelta, timezone
+
+import pytest
+
+pytestmark = pytest.mark.integration
+
+_NEEDS_DB = pytest.mark.skipif(
+    not os.environ.get("STEWARD_DATABASE_URL"),
+    reason="integration test needs a live Postgres (STEWARD_DATABASE_URL)",
+)
+
+
+@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)
+
+
+@_NEEDS_DB
+def test_latest_distinct_on_and_sql_bucketed_history(app):
+    from sqlalchemy import text
+    from steward.models.metrics import PluginMetric
+    from plugins.host_agent.metrics_query import (
+        SOURCE_MODULE, _history_for_host, _latest_metrics_for_host,
+    )
+
+    now = datetime.now(timezone.utc)
+    hostname = "metrics-host-" + uuid.uuid4().hex[:8]
+
+    async def _go():
+        async with app.db_sessionmaker() as s:
+            async with s.begin():
+                await s.execute(
+                    text("DELETE FROM plugin_metrics WHERE resource_name LIKE :p"),
+                    {"p": hostname + "%"},
+                )
+                rows = []
+                # Host-level CPU samples, oldest → newest (10, 20, 30).
+                for i, val in enumerate([10.0, 20.0, 30.0]):
+                    rows.append(PluginMetric(
+                        source_module=SOURCE_MODULE, resource_name=hostname,
+                        metric_name="cpu_pct", value=val,
+                        recorded_at=now - timedelta(minutes=30 - i * 10),
+                    ))
+                # A sub-resource (root mount) to exercise the host:% match.
+                rows.append(PluginMetric(
+                    source_module=SOURCE_MODULE, resource_name=hostname + ":/",
+                    metric_name="disk_used_pct", value=80.0,
+                    recorded_at=now - timedelta(minutes=5),
+                ))
+                s.add_all(rows)
+
+            latest = await _latest_metrics_for_host(s, hostname)
+            hist = await _history_for_host(s, hostname, now - timedelta(hours=1))
+            return latest, hist
+
+    latest, hist = asyncio.run(_go())
+
+    # DISTINCT ON returns the newest sample per (resource, metric).
+    assert latest[hostname]["cpu_pct"] == 30.0
+    assert latest[hostname + ":/"]["disk_used_pct"] == 80.0
+
+    # SQL date_bin aggregation returns cpu_pct buckets; averages stay in range.
+    cpu = hist["cpu_pct"]
+    assert cpu, "expected bucketed cpu_pct history"
+    assert all(10.0 <= v <= 30.0 for _, v in cpu)
+
+
+@_NEEDS_DB
+def test_rollup_aggregates_old_raw_and_prunes(app):
+    from sqlalchemy import text
+    from steward.core.metrics_retention import rollup_plugin_metrics
+    from steward.models.metrics import PluginMetric
+
+    now = datetime.now(timezone.utc)
+    hostname = "rollup-host-" + uuid.uuid4().hex[:8]
+
+    async def _go():
+        async with app.db_sessionmaker() as s:
+            async with s.begin():
+                for tbl in ("plugin_metrics", "plugin_metrics_hourly"):
+                    await s.execute(
+                        text(f"DELETE FROM {tbl} WHERE resource_name LIKE :p"),
+                        {"p": hostname + "%"},
+                    )
+                old = (now - timedelta(days=10)).replace(minute=5, second=0, microsecond=0)
+                s.add_all([
+                    PluginMetric(source_module="host_agent", resource_name=hostname,
+                                 metric_name="cpu_pct", value=10.0, recorded_at=old),
+                    PluginMetric(source_module="host_agent", resource_name=hostname,
+                                 metric_name="cpu_pct", value=20.0,
+                                 recorded_at=old.replace(minute=35)),  # same hour bucket
+                    PluginMetric(source_module="host_agent", resource_name=hostname,
+                                 metric_name="cpu_pct", value=99.0,
+                                 recorded_at=now - timedelta(hours=1)),  # recent → kept raw
+                ])
+            async with s.begin():
+                counts = await rollup_plugin_metrics(s, raw_days=7, rollup_days=90, now=now)
+            hrly = (await s.execute(text(
+                "SELECT value_avg, sample_count FROM plugin_metrics_hourly "
+                "WHERE resource_name = :h AND metric_name = 'cpu_pct'"), {"h": hostname})).all()
+            raw_left = (await s.execute(text(
+                "SELECT count(*) FROM plugin_metrics WHERE resource_name = :h"), {"h": hostname})).scalar()
+            return counts, hrly, raw_left
+
+    counts, hrly, raw_left = asyncio.run(_go())
+    assert counts["buckets_rolled"] == 1
+    assert counts["raw_rows_rolled"] == 2
+    assert len(hrly) == 1
+    assert abs(float(hrly[0][0]) - 15.0) < 0.001   # avg(10, 20)
+    assert hrly[0][1] == 2
+    assert raw_left == 1                            # only the recent sample remains
+
+
+@_NEEDS_DB
+def test_history_merges_hourly_and_raw(app):
+    from sqlalchemy import text
+    from steward.models.metrics import PluginMetric, PluginMetricHourly
+    from plugins.host_agent.metrics_query import _history_for_host
+
+    now = datetime.now(timezone.utc)
+    hostname = "merge-host-" + uuid.uuid4().hex[:8]
+
+    async def _go():
+        async with app.db_sessionmaker() as s:
+            async with s.begin():
+                for tbl in ("plugin_metrics", "plugin_metrics_hourly"):
+                    await s.execute(
+                        text(f"DELETE FROM {tbl} WHERE resource_name LIKE :p"),
+                        {"p": hostname + "%"},
+                    )
+                old_bucket = (now - timedelta(days=10)).replace(minute=0, second=0, microsecond=0)
+                s.add(PluginMetricHourly(
+                    source_module="host_agent", resource_name=hostname, metric_name="cpu_pct",
+                    bucket=old_bucket, value_avg=42.0, value_max=50.0, sample_count=120))
+                s.add(PluginMetric(
+                    source_module="host_agent", resource_name=hostname, metric_name="cpu_pct",
+                    value=77.0, recorded_at=now - timedelta(hours=1)))
+            # 30d range with a 7d raw window → spans the rollup boundary.
+            return await _history_for_host(s, hostname, now - timedelta(days=30), raw_days=7)
+
+    hist = asyncio.run(_go())
+    cpu = hist["cpu_pct"]
+    vals = [v for _, v in cpu]
+    assert 42.0 in vals, "expected the rolled-up hourly point"
+    assert 77.0 in vals, "expected the recent raw point"
+    assert cpu == sorted(cpu, key=lambda p: p[0]), "series must be time-ordered"

tests/plugins/docker/test_routes.py

@@ -0,0 +1,50 @@
+"""Unit tests for the docker plugin's presentation layer (no DB).
+
+CI never renders these templates through the running app (the unit-lane app is
+created with testing=True, which skips plugin loading), so a Jinja syntax error
+would otherwise ship green. These tests parse every docker template and smoke
+the routes module so a broken tag or import is caught in the unit lane.
+"""
+import pathlib
+
+import jinja2
+
+from plugins.docker import routes as r
+
+_TEMPLATES = pathlib.Path(r.__file__).parent / "templates" / "docker"
+
+
+def test_all_docker_templates_parse():
+    env = jinja2.Environment()
+    files = sorted(_TEMPLATES.glob("*.html"))
+    assert files, "no docker templates found"
+    for f in files:
+        # Raises TemplateSyntaxError on an unbalanced/typo'd tag.
+        env.parse(f.read_text())
+
+
+def test_routes_module_exposes_new_views():
+    # Import-smoke + confirms the #942 view functions are defined.
+    for name in ("container_detail", "container_history", "swarm", "disk", "index", "rows"):
+        assert callable(getattr(r, name)), name
+    assert r.docker_bp.name == "docker"
+
+
+def test_human_bytes_formats_binary_units():
+    assert r._human_bytes(None) == "—"
+    assert r._human_bytes(0) == "0 B"
+    assert r._human_bytes(512) == "512 B"
+    assert r._human_bytes(1024) == "1.0 KiB"
+    assert r._human_bytes(1536) == "1.5 KiB"
+    assert r._human_bytes(1024 ** 2) == "1.0 MiB"
+    assert r._human_bytes(int(1.5 * 1024 ** 3)) == "1.5 GiB"
+    assert r._human_bytes(1024 ** 4) == "1.0 TiB"
+
+
+def test_human_uptime_compact():
+    from datetime import datetime, timedelta, timezone
+    now = datetime.now(timezone.utc)
+    assert r._human_uptime(None) is None
+    assert r._human_uptime(now - timedelta(minutes=8)).endswith("m")
+    assert "h" in r._human_uptime(now - timedelta(hours=5, minutes=12))
+    assert "d" in r._human_uptime(now - timedelta(days=3, hours=4))

tests/plugins/docker/test_swarm_view.py

@@ -0,0 +1,79 @@
+"""Unit tests for the swarm-aware container view builder (pure, no DB/ORM)."""
+from types import SimpleNamespace
+
+from plugins.docker.swarm_view import build_swarm_services
+
+NODE_HOSTNAME = {"nA": "docker01", "nB": "docker02", "nC": "docker03"}
+HOST_NAME = {"h1": "docker01", "h2": "docker02"}  # agents run on docker01/02 only
+
+
+def _c(host_id, name, *, service=None, node=None, status="running", cpu=1.0):
+    return SimpleNamespace(host_id=host_id, name=name, service_name=service,
+                           node_id=node, status=status, cpu_pct=cpu, mem_pct=2.0,
+                           health=None, restart_count=0, container_id=name)
+
+
+def _svc(name, running, desired, placement, mode="replicated"):
+    import json
+    return SimpleNamespace(service_name=name, running=running, desired=desired,
+                           mode=mode, placement_json=json.dumps(placement))
+
+
+def test_real_replicas_plus_agentless_ghost():
+    containers = [
+        _c("h1", "web.1.aaa", service="web", node="nA"),
+        _c("h2", "web.2.bbb", service="web", node="nB"),
+        _c("h1", "db", node=None),  # standalone, non-swarm
+    ]
+    services = [_svc("web", running=3, desired=3,
+                     placement=[{"node_id": "nA", "running": 1},
+                                {"node_id": "nB", "running": 1},
+                                {"node_id": "nC", "running": 1}])]  # nC has no agent
+
+    view = build_swarm_services(containers, services, NODE_HOSTNAME, HOST_NAME)
+
+    assert len(view["services"]) == 1
+    svc = view["services"][0]
+    assert svc["name"] == "web" and svc["state"] == "healthy"
+    reals = [r for r in svc["replicas"] if not r["ghost"]]
+    ghosts = [r for r in svc["replicas"] if r["ghost"]]
+    assert sorted(r["host"] for r in reals) == ["docker01", "docker02"]
+    assert len(ghosts) == 1 and ghosts[0]["host"] == "docker03" and ghosts[0]["count"] == 1
+
+    # The non-swarm container is grouped by host, untouched.
+    assert len(view["standalone"]) == 1
+    g = view["standalone"][0]
+    assert g["host"] == "docker01" and [c.name for c in g["containers"]] == ["db"]
+
+
+def test_placement_fully_covered_has_no_ghosts():
+    containers = [_c("h1", "api.1.x", service="api", node="nA"),
+                  _c("h2", "api.2.y", service="api", node="nB")]
+    services = [_svc("api", 2, 2, [{"node_id": "nA", "running": 1},
+                                   {"node_id": "nB", "running": 1}])]
+    view = build_swarm_services(containers, services, NODE_HOSTNAME, HOST_NAME)
+    assert all(not r["ghost"] for r in view["services"][0]["replicas"])
+
+
+def test_multiple_tasks_same_node_partially_agentless():
+    # Placement says 2 running on nA, but we only have 1 local row → 1 ghost on nA.
+    containers = [_c("h1", "cache.1.x", service="cache", node="nA")]
+    services = [_svc("cache", 2, 2, [{"node_id": "nA", "running": 2}])]
+    view = build_swarm_services(containers, services, NODE_HOSTNAME, HOST_NAME)
+    ghosts = [r for r in view["services"][0]["replicas"] if r["ghost"]]
+    assert len(ghosts) == 1 and ghosts[0]["count"] == 1 and ghosts[0]["host"] == "docker01"
+
+
+def test_service_without_a_service_row_is_synthesised():
+    # Container labeled with a service, but no manager reported the service row.
+    containers = [_c("h1", "orphan.1.x", service="orphan", node="nA")]
+    view = build_swarm_services(containers, [], NODE_HOSTNAME, HOST_NAME)
+    svc = view["services"][0]
+    assert svc["name"] == "orphan" and svc["running"] == 1 and svc["desired"] == 1
+    assert all(not r["ghost"] for r in svc["replicas"])
+
+
+def test_down_service_state():
+    services = [_svc("idle", 0, 2, [])]
+    view = build_swarm_services([], services, NODE_HOSTNAME, HOST_NAME)
+    assert view["services"][0]["state"] == "down"

tests/plugins/snmp/test_host_match.py

@@ -0,0 +1,69 @@
+"""Unit tests for mapping config-defined SNMP devices onto a Steward host."""
+from plugins.snmp.routes import _devices_for_host
+
+_CFG = [
+    {"name": "core-switch", "host": "192.168.1.1"},
+    {"name": "ups", "host": "192.168.1.10"},
+    {"name": "by-name", "host": "edge-router"},
+    "not-a-dict",                     # tolerated, skipped
+    {"name": "no-host"},              # no host → never matches
+]
+
+
+def test_matches_by_address():
+    out = _devices_for_host(_CFG, "192.168.1.1", "somehost")
+    assert [d["name"] for d in out] == ["core-switch"]
+
+
+def test_matches_by_name_case_insensitive():
+    out = _devices_for_host(_CFG, "10.0.0.9", "Edge-Router")
+    assert [d["name"] for d in out] == ["by-name"]
+
+
+def test_no_match_returns_empty():
+    assert _devices_for_host(_CFG, "10.0.0.1", "nope") == []
+
+
+def test_blank_address_and_name_match_nothing():
+    # A host with no address/name must not match a device with a blank host.
+    assert _devices_for_host(_CFG, "", None) == []
+
+
+def test_implicit_match_unaffected_when_host_id_passed():
+    # Passing the new host_id arg must not break the legacy implicit match.
+    out = _devices_for_host(_CFG, "192.168.1.1", "somehost", host_id="h-1")
+    assert [d["name"] for d in out] == ["core-switch"]
+
+
+# ── Explicit binding ─────────────────────────────────────────────────────────
+
+def test_explicit_host_id_matches_exact_uuid():
+    cfg = [{"name": "sw", "host": "10.0.0.5", "host_id": "uuid-abc"}]
+    out = _devices_for_host(cfg, "10.0.0.9", "other", host_id="uuid-abc")
+    assert [d["name"] for d in out] == ["sw"]
+
+
+def test_explicit_steward_host_matches_by_name_or_address():
+    cfg = [{"name": "pdu", "host": "10.0.0.5", "steward_host": "Switch01"}]
+    # Bind hits the Host's name despite a different poll target / address.
+    by_name = _devices_for_host(cfg, "10.0.0.9", "switch01", host_id="h9")
+    by_addr = _devices_for_host(cfg, "switch01", "whatever", host_id="h9")
+    assert [d["name"] for d in by_name] == ["pdu"]
+    assert [d["name"] for d in by_addr] == ["pdu"]
+
+
+def test_explicit_binding_is_exclusive():
+    # Device polls 192.168.1.1 but is explicitly bound to host A by name.
+    cfg = [{"name": "sw", "host": "192.168.1.1", "steward_host": "hostA"}]
+    # Host B *would* match implicitly by the poll target — but must NOT,
+    # because the explicit binding takes over and points only at host A.
+    hostB = _devices_for_host(cfg, "192.168.1.1", "hostB", host_id="b")
+    hostA = _devices_for_host(cfg, "10.0.0.2", "hostA", host_id="a")
+    assert hostB == []
+    assert [d["name"] for d in hostA] == ["sw"]
+
+
+def test_explicit_host_id_wrong_uuid_does_not_match():
+    cfg = [{"name": "sw", "host": "10.0.0.5", "host_id": "uuid-abc"}]
+    out = _devices_for_host(cfg, "10.0.0.5", "sw", host_id="uuid-zzz")
+    assert out == []

tests/plugins/snmp/test_poller.py

@@ -0,0 +1,76 @@
+"""Unit tests for the SNMP poller (no pysnmp / no network).
+
+The unit lane doesn't install the `snmp` extra, so these exercise the parts that
+don't need pysnmp: the version→mpModel mapping, that the poller is async, and the
+graceful "pysnmp missing → empty result" path. Live polling against a real device
+is verified out-of-band (CI has no SNMP target). Uses asyncio.run() directly so it
+doesn't depend on the pytest-asyncio mode.
+"""
+import asyncio
+import inspect
+
+from plugins.snmp import poller
+
+
+def test_poll_device_is_coroutine():
+    # The scheduler awaits it directly (no executor) — it must be async.
+    assert inspect.iscoroutinefunction(poller.poll_device)
+
+
+def test_mp_model_maps_version_to_int():
+    assert poller._mp_model("1") == 0      # SNMP v1
+    assert poller._mp_model("2c") == 1     # SNMP v2c
+    assert poller._mp_model("2") == 1      # anything non-"1" → v2c model
+
+
+def test_poll_device_without_pysnmp_returns_empty(monkeypatch):
+    # When the optional dep is absent, polling is disabled gracefully (no raise).
+    monkeypatch.setattr(poller, "_pysnmp_available", lambda: False)
+    out = asyncio.run(
+        poller.poll_device("192.0.2.1", 161, "public", "2c", [{"oid": "1.3.6.1.2.1.1.3.0"}])
+    )
+    assert out == {}
+
+
+# --- _close_engine: the fd-leak guard ---------------------------------------
+# A fresh SnmpEngine per poll leaks its UDP socket unless closed; _close_engine
+# must release it across both pysnmp API shapes. These fakes stand in for the
+# engine since the unit lane has no pysnmp.
+
+def test_close_engine_uses_engine_level_close_7x():
+    # canonical pysnmp 7.x: close method lives directly on the engine.
+    calls = []
+
+    class Engine:
+        def close_dispatcher(self):
+            calls.append("engine")
+
+    poller._close_engine(Engine())
+    assert calls == ["engine"]
+
+
+def test_close_engine_falls_back_to_dispatcher_6x():
+    # pysnmp-lextudio 6.2.x: no engine-level close; go through the dispatcher.
+    calls = []
+
+    class Dispatcher:
+        def closeDispatcher(self):
+            calls.append("dispatcher")
+
+    class Engine:
+        transportDispatcher = Dispatcher()
+
+    poller._close_engine(Engine())
+    assert calls == ["dispatcher"]
+
+
+def test_close_engine_never_raises():
+    # A failing close must not break the poll loop.
+    class Engine:
+        def close_dispatcher(self):
+            raise OSError("boom")
+
+    poller._close_engine(Engine())  # no exception
+
+    # No close path at all (defensive) is also fine.
+    poller._close_engine(object())

tests/plugins/unifi/test_scheduler.py

@@ -0,0 +1,92 @@
+"""Unit tests for the UniFi poll scheduler's client lifecycle (no network/DB).
+
+The leak guard: the module-cached UnifiClient holds a long-lived httpx pool, so
+whenever the scheduler discards it (login failure, or a mid-poll error that
+forces re-auth) it MUST close it first — otherwise every failure tick orphans a
+connection pool and leaks file descriptors. These tests drive both discard paths
+and assert the client was closed and the cache cleared. Uses asyncio.run()
+directly so it doesn't depend on the pytest-asyncio mode.
+"""
+import asyncio
+import types
+
+import plugins.unifi.client as client_mod
+import plugins.unifi.scheduler as scheduler
+
+
+def _make_app():
+    app = types.SimpleNamespace()
+    app.config = {"PLUGINS": {"unifi": {"host": "h", "username": "u", "password": "p"}}}
+    return app
+
+
+class _FakeClient:
+    """Stand-in for UnifiClient; records close() and can fail on demand."""
+    created: list["_FakeClient"] = []
+    login_fails = False
+    health_fails = False
+
+    def __init__(self, **_kwargs):
+        self.closed = False
+        _FakeClient.created.append(self)
+
+    async def login(self):
+        if _FakeClient.login_fails:
+            raise RuntimeError("login boom")
+
+    async def get_health(self):
+        if _FakeClient.health_fails:
+            raise RuntimeError("health boom")
+        return []
+
+    async def close(self):
+        self.closed = True
+
+
+def _install(monkeypatch):
+    monkeypatch.setattr(client_mod, "UnifiClient", _FakeClient)
+    monkeypatch.setattr(scheduler, "_client", None)
+    _FakeClient.created = []
+    _FakeClient.login_fails = False
+    _FakeClient.health_fails = False
+
+
+def test_login_failure_closes_and_clears_client(monkeypatch):
+    _install(monkeypatch)
+    _FakeClient.login_fails = True
+
+    asyncio.run(scheduler._do_poll(_make_app()))
+
+    assert len(_FakeClient.created) == 1
+    assert _FakeClient.created[0].closed is True   # closed, not just dropped
+    assert scheduler._client is None               # cache cleared → re-auth next tick
+
+
+def test_poll_failure_closes_and_clears_client(monkeypatch):
+    _install(monkeypatch)
+    _FakeClient.health_fails = True   # login OK, first API call blows up
+
+    asyncio.run(scheduler._do_poll(_make_app()))
+
+    assert len(_FakeClient.created) == 1
+    assert _FakeClient.created[0].closed is True
+    assert scheduler._client is None
+
+
+def test_drop_client_is_noop_when_unset(monkeypatch):
+    _install(monkeypatch)
+    # No client cached — dropping must not raise.
+    asyncio.run(scheduler._drop_client())
+    assert scheduler._client is None
+
+
+def test_drop_client_swallows_close_errors(monkeypatch):
+    _install(monkeypatch)
+
+    class _Boom:
+        async def close(self):
+            raise OSError("close boom")
+
+    monkeypatch.setattr(scheduler, "_client", _Boom())
+    asyncio.run(scheduler._drop_client())   # must not propagate
+    assert scheduler._client is None

tests/test_config.py

@@ -1,6 +1,6 @@
 import textwrap
 import pytest
-from steward.config import load_bootstrap
+from steward.config import load_bootstrap, _resolve_secret_key
 
 
 def test_load_bootstrap_from_yaml(tmp_path):
@@ -36,3 +36,32 @@ def test_missing_database_url_raises(tmp_path):
     cfg_file.write_text("secret_key: s\n")
     with pytest.raises(ValueError, match="Database URL is required"):
         load_bootstrap(cfg_file)
+
+
+# ── secret key resolution ──────────────────────────────────────────────────────
+
+def test_secret_key_existing_file_is_reused(tmp_path, monkeypatch):
+    key_file = tmp_path / "secret.key"
+    key_file.write_text("persisted-key\n")
+    monkeypatch.setattr("steward.config._SECRET_KEY_FILE", key_file)
+    monkeypatch.delenv("STEWARD_SECRET_KEY", raising=False)
+    assert _resolve_secret_key({}) == "persisted-key"
+
+
+def test_secret_key_generated_and_persisted_when_writable(tmp_path, monkeypatch):
+    key_file = tmp_path / "sub" / "secret.key"   # parent doesn't exist yet
+    monkeypatch.setattr("steward.config._SECRET_KEY_FILE", key_file)
+    monkeypatch.delenv("STEWARD_SECRET_KEY", raising=False)
+    key = _resolve_secret_key({})
+    assert key and key_file.read_text().strip() == key
+
+
+def test_secret_key_unpersistable_raises_instead_of_ephemeral(tmp_path, monkeypatch):
+    # A file sits where a directory is needed, so mkdir/write fails → must raise
+    # (an ephemeral key would silently orphan every encrypted secret).
+    blocker = tmp_path / "blocker"
+    blocker.write_text("not a dir")
+    monkeypatch.setattr("steward.config._SECRET_KEY_FILE", blocker / "secret.key")
+    monkeypatch.delenv("STEWARD_SECRET_KEY", raising=False)
+    with pytest.raises(RuntimeError, match="could not persist"):
+        _resolve_secret_key({})

tests/test_templates_parse.py

@@ -0,0 +1,34 @@
+"""Syntax-parse every first-party template so a broken tag fails the unit lane.
+
+The app only renders a handful of templates in unit tests (e.g. the login page),
+so a Jinja syntax error elsewhere would otherwise ship green. env.parse() checks
+syntax without resolving extends/includes/imports — enough to catch an unbalanced
+or mistyped tag across the whole template tree (e.g. the base.html layout).
+"""
+import pathlib
+
+import jinja2
+
+import steward
+
+_REPO = pathlib.Path(steward.__file__).parent.parent
+_CORE_TEMPLATES = _REPO / "steward" / "templates"
+_PLUGINS = _REPO / "plugins"
+
+
+def test_all_steward_templates_parse():
+    env = jinja2.Environment()
+    files = sorted(_CORE_TEMPLATES.rglob("*.html"))
+    assert files, "no steward templates found"
+    for f in files:
+        env.parse(f.read_text())  # raises TemplateSyntaxError on a bad tag
+
+
+def test_all_plugin_templates_parse():
+    # First-party plugins ship templates too (host_agent fragments, docker, …);
+    # they aren't rendered in the unit lane, so parse them here to catch a bad tag.
+    env = jinja2.Environment()
+    files = sorted(_PLUGINS.rglob("templates/**/*.html"))
+    assert files, "no plugin templates found"
+    for f in files:
+        env.parse(f.read_text())