- Dockerfile: python:3.14-slim, the Python CI tests on (was 3.13). tini from
apt is PID 1 (ENTRYPOINT tini -- entrypoint.sh): it reaps the zombies ssh
and ansible leave and forwards SIGTERM.
- cli: Hypercorn's serve() with one worker (run, stream and scheduler state is
in-process), a 10 s graceful timeout, and SIGTERM/SIGINT as the shutdown
trigger. Access and error lines go through Steward's logging. --debug keeps
Quart's reloading dev server.
- Logging: steward/core/logging_setup.py installs one root handler before
anything logs (until now a library's basicConfig on import was the only
configuration). Its level is the `log.level` setting (DEBUG/INFO/WARNING/
ERROR, default INFO), applied at boot and on save in Settings → General.
- CI publish: build, then "The image boots with no internet" (rule 164,
snippet #3552), then push, so a broken image never reaches a tag (rule 177).
On an --internal network with its own Postgres, the image must migrate and
answer /health, run tini as PID 1, serve first-run setup with its assets
from /static and no third-party host, and exit 0 on SIGTERM. Logs and
clean-up run always.
Tests: tests/core/test_logging_setup.py; the smoke step itself.
Not in this commit: the dependency lockfile (it conflicts with rule 154 and
is waiting on the operator).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The "[queued — waiting for run X on the same host]" and run-slot lines were
broadcast to live viewers and never stored, so a run that sat queued had no
record of why once the live view was gone (CI 8439: the overlap test found
the order right and the explanation missing). _claim_hosts now returns the
line it waited on and start_run seeds the run's output with every wait.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- One run per host: start_run claims every host it touches (addresses read
from the inventory Steward wrote; a source's own inventory file is claimed
as that file) before it takes a run slot. The claim is all-or-nothing. A run
waiting on a host shows queued and names the run it waits for, honours a
cancel, and fails after HOST_WAIT_SECONDS (1 h) with which host and run.
- A due schedule whose previous run is still queued or running skips the
firing: last_run_at moves on, last_run_id still names the run going, and
last_error says "Skipped: the previous run was still going". The Schedules
page shows that beside the run (muted), and a launch error in red.
- An alert rule that fires while the action run from an earlier firing is
still going starts no new run; the event says so and names the run. An
in-process set closes the gap between the check and the run being linked.
- docs/core/ansible.md: queued/cancelled statuses and the overlap rules.
Tests: tests/core/test_ansible_overlap.py (claim keys, waiting, all-or-nothing,
cancel, deadline) and tests/integration/test_ansible_overlap.py (two real
runs on one host take turns; schedule skip then fire; alert action skip then
start).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- git ships in the image; checkouts live under /data/ansible/sources
(STEWARD_ANSIBLE_SOURCE_DIR), on the persistent volume.
- One task, ansible_git_pulls, reads the sources every 30 s and pulls each
as its interval passes: a source added in Settings is pulled without a
restart, a removed one is forgotten.
- git_pull: a 120 s timeout ending the process group, GIT_TERMINAL_PROMPT=0,
pull --ff-only, a failed clone removed; it raises GitPullError with git's
message, so the Sync button stops saying Synced after a failure (and the
message is now escaped).
- An alert action's vars and the firing's steward_alert_* vars reach
ansible-playbook as one JSON extra-vars file, not -e key=value, which
Ansible split on whitespace (a rule named 'High CPU on web').
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Output is read in 64 KiB chunks and split into lines here; a line past
64 KiB is cut and marked instead of raising out of readline's limit.
- ansible-playbook runs in its own session; whatever ends the read (finish,
an error, the task being cancelled) SIGTERMs then SIGKILLs its whole group,
so no ssh or worker it forked outlives the run.
- A run whose task is cancelled (shutdown) records 'cancelled' with
finished_at before re-raising, instead of staying 'running'.
- A cancel that lands after the process exited 0 no longer turns a
successful run into 'cancelled'.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- prune_by_age(column) registers a time-series column with the hourly cleanup;
Traefik (metrics, access summaries, global stats) and UniFi (WAN, client,
WLAN, DPI, events, alarms, speedtests) register theirs in setup().
- Alert events join monitor results and Ansible runs on data.retention_days,
read from Settings each run; Ansible run log files older than it are deleted.
- audit.retention_days (365) keeps the audit log longer, set on Settings > General.
- metric_series.points reads metrics.retention.raw_days, so the chart's
raw/rollup seam is where the rollup put it (RAW_DAYS removed).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An audit of 1,980 text/surface pairs (every steward.css rule and every inline
template style, tints composited over page, card and hovered row) found:
- Red, yellow and orange status text on its own 20% tint over a hovered row
at 3.97-4.06:1 (badges, count badge, alert-error, btn-warn). --red and
--yellow are lifted 15% toward the primary text, --orange 18%, as --blue
already was: worst case now 4.54:1.
- Inputs, selects, ghost and range buttons outlined in --border-mid at
1.4-1.9:1, on surfaces 1.1:1 apart, so the outline is what marks them.
New --border-control (45% toward secondary text): 3.13:1 worst. The
dashboard switcher and widget-drawer fields drop their inline styles for
the shared .input class.
- The breadcrumb separator painted in a border colour (1.4-1.9:1): now
--text-dim (4.64:1) and aria-hidden, since it is decoration.
tests/test_contrast.py resolves the tokens (color-mix and tints included)
and holds every drawn pair to 4.5:1, status text on its tint to 4.5:1 and
control outlines to 3:1, with a test that the guard catches the old pairs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Measured on dev (385a579), agent posts cut to five samples still dropped:
flush=8.8s for 990 rows, 6.6s for 680, 8.4s for 1000 — about 110 rows/s,
~9 ms a row, with rules and evaluation near zero. One random read per
row fits a random uuid4 primary key whose index no longer fits in memory:
every new row lands on a random page of it. (Inferred from the rate; the
table and index sizes have not been read.)
- steward.core.ids.time_ordered_id: a UUIDv7-layout string (48-bit ms
timestamp, then random), same 36-char shape as the stored uuid4 ids, so
no migration. Python's uuid7() arrives in 3.14; the image runs 3.13.
- Used for plugin_metrics and its hourly rollup, monitor_results,
audit_events, the docker metrics/hourly/events/logs tables, and the
traefik and unifi time series.
- MAX_SAMPLES_PER_INGEST 5 → 2 (~400 rows) for now: it fits the agent's
10 s even at the old rate, so hosts land data whichever way the new rate
comes out. Raise it once the new rate is measured.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The first instrumented drop on dev read "dropped after 10.0s … (completed:
parse=0.1s)": the whole stall is inside record_metrics. It now marks its
three steps (flush of the rows, the rule and state reads, the per-sample
evaluation) on the caller's phases, and the drop line carries the request's
samples and metric row count.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The dispatcher ticked once per MONITORS_POLL_INTERVAL and stamped
last_checked_at when the whole batch finished. With a 60 s tick and a 60 s
monitor the next tick saw 59.x s elapsed and skipped, so the monitor ran
every 120 s; a 30 s monitor could never run faster than the tick.
- The dispatcher ticks every 5 s (DISPATCH_TICK_SECONDS) and reads the
monitors and intervals fresh each tick, so a changed interval applies at
the next one.
- Due is measured from the scheduled start, stamped before the check runs,
with half a tick of slack so tick jitter never costs a whole tick. A
long-overdue monitor runs once.
- Each check runs in its own task and records its own result, so a slow
probe holds up nobody; a monitor still being checked is not started
again, and it is freed in a finally.
- A check past CHECK_DEADLINE_SECONDS (120) is recorded down with "Check did
not finish within 120 s" rather than holding its monitor forever.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Traefik: routers are graded from their latest scrape (router_state: crit on
any real 5xx share, warn on p95 past the latency critical cutoff or a high
4xx share). Failing and slow routers lead as cards that say why; healthy
ones fold into one table that stays open or shut across polls. Proxy
counts are stat tiles; certificates are graded by Settings -> Thresholds
(tls_days).
UniFi: WAN health is a stat strip (latency graded by the thresholds),
devices carry status dots, event categories and firewall actions are
badges, the alarm count reuses .count-badge (renamed from .nav-count).
SNMP: each device wears a status badge and its readings are stat tiles.
Agent fleet: the table comes first, with a Status column (Reporting /
Silent / Never reported, by the host page's stale_after_seconds), silent
agents first, hosts linked, last seen as |ago.
Audit: entries name their entity (one query per type, steward/audit/
service.py) and link it; something deleted since keeps the name its entry
recorded and says "(deleted)".
Tokens: charts read colours through chart-colors.js (stewardColor resolves
a token to the rgb the browser computed; Chart.js defaults take the page's
text and border colours), so host, history, Traefik and SNMP charts drop
their hex. Template hex colours reach 0 (HEX_CEILING = 0), the last rgba
pills go, and heavy weights (600/700/bold, 60 of them) become 500, the
house pair being 400/500; a guard keeps templates there.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Runs read "update.yml on web-1" (run_title over the scope label, beside
describe_scope) on the runs list, the run page's title and breadcrumb, and
the run-started panel. The list shows started |ago (full time on hover),
|duration (counting "so far" while a run is going) and a dry-run tag.
Run statuses join STATES, so the list and the run page use status_badge:
queued dim, running info (blue, so work in progress doesn't read as a
problem), success ok "Succeeded", failed crit, interrupted warn, cancelled
dim. run_started.html leaves its hard-coded colours for tokens; the run
page's facts move to the .kv list and its failures to var(--red). Template
hex colours: 54 -> 47.
Also fixes#5124's sidebar count, which CI run 8389 failed on: the stylesheet
guard allows no weight above 500, and .nav-count asked for 600.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Alerts page's active list now says each alert in one line around the
resource it watches, e.g. "cpu_pct on web-1 is 94% (> 90%), firing for 12m",
with web-1 linking to its host, monitor or container page (or the plugin's
page for Traefik, UniFi and SNMP). Values take their unit from the metric
name through formatting.metric_value. A silent series reads "No cpu_pct from
web-1 for 20m"; an up/down metric says the word.
Actions follow the routes' roles: acknowledge, edit, toggle and New rule for
operators and admins, Delete for admins only, nothing for viewers.
The sidebar shows how many alerts need attention beside Alerts, from an
async context processor that skips htmx fragments, logged-out pages and the
database-down page, and renders without the count if the query fails.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The host page's vitals strip becomes its health header: Live or Stale,
then one sentence — "Memory 93% is past its critical threshold", "Every
reading is within its thresholds", or, for a stale host, "No report since
20m ago; the readings below are from then" — and last seen as |ago. CPU,
memory, root disk and load per core follow as stat tiles graded by
Settings → Thresholds, with their sparklines.
The full-metrics page drew its bars and gauges at a hard-coded 70/90
(temperatures 70/85) while every other page used the configured cutoffs
(80/90 by default), so a 75 % CPU was amber here and fine everywhere else.
bar() now takes the metric kind and colours its .meter-fill through
status_class; gauges, temperatures and per-core bars do the same.
The host's monitor rows move onto the Monitors list's shapes: a status dot,
the name linking to the monitor's page, Ping/TCP/DNS/HTTP, pill-up/
pill-down, latency and uptime through status_class, labelled icon buttons.
HEX_CEILING drops to 54.
New formatting.worst_reading(readings, thresholds) picks the reading
furthest past its cutoff, the first of equals winning.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Monitors list now opens with a count strip (down, up with how many
degraded, pending, paused) and sorts down, degraded, pending, up, paused,
by name within each. Degraded is up but worth a look: uptime at or below
its warning floor, or the latest response past its critical cutoff, both
from Settings → Thresholds; a slow-but-tolerable reply is coloured, not
sorted.
A monitor had no page of its own: every link went to the list. Now
/monitors/<id> shows its state and since when, latency, uptime for 24h/7d/
30d, TLS days, the last checks, recent failures with their errors, and its
settings (HTTP header names, never their values). Status links to it, and
adding or editing returns to it.
Adding is its own page (/monitors/new, form.html shared with edit) behind
an "Add monitor" button; a refused add re-renders the form with the error
and what was typed instead of a bare 400 text. Row actions are labelled
icon buttons, type names read Ping/TCP/DNS/HTTP (|monitor_type), and the
heartbeat uses pill-up/pill-down, so HEX_CEILING drops to 56.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5120 "Status page: what is wrong first".
- A "Needs attention" card heads the Status board: each down monitor with
how long it has been down, and each active alert (firing, acknowledged,
pending, no data) with its condition and how long, linking to its history.
- StatusEntry.since / since_at_least: when the current up/down run began,
from the recent checks (streak_start); "for at least" when every check we
have agrees.
- Counts as stat tiles (Down, Active alerts, Pending, Up); names wrap
instead of being cut off; dots carry their label.
- Uptime and TLS are coloured by status_class against Settings → Thresholds
instead of literal cutoffs; heartbeat pills and DOWN/UP use tokens and
status_badge rather than hex and bold.
- tests/core/test_status_page.py: streak_start cases, and the rows render
with the problem card before the list, each since-time, threshold tones
and no hex colour.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5095 "Share view on the shared stylesheet; breadcrumbs, focus-visible and
wide tables".
- dashboard/share_view.html loads theme.css and steward.css; its own copy of
the old purple palette is gone. Only the read-only frame and the x/y/w/h
grid stay inline, written in tokens. HEX_CEILING 75 -> 58.
- Breadcrumbs on settings/general, settings/thresholds, the dashboard share
page and share-link-created. tests/test_breadcrumbs.py fails on any page
extending base.html without a crumbs() trail, unless it is on the named
list of section top pages, logged-out and error pages (rule 134).
- :focus-visible draws --fs-focus-ring on links, buttons, summaries, inputs
and anything with tabindex.
- Any card holding a .table scrolls sideways (card:has(table.table)), so a
wide table no longer widens the page at phone width; .card-flush scrolls
too, main gets min-width:0, and .table-wrap covers a table outside a card.
- tests/test_share_view.py renders the share view: shared sheets, no hex
colour, escaped name, widgets loading through the share token.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5094 "Status and layout macros, and form classes".
- formatting.STATES: one state -> (tone, label) map for up, down, stale,
firing, pending, acknowledged, no_data, ok, warn, crit, resolved and
inactive; status_info() reads it (an unmapped state is drawn dim under its
own name). required() lets a macro refuse a missing argument.
- _macros.html: status_badge and status_dot (a dot always carries its label,
sr-only unless shown), stat_tile, empty_state, page_header (actions via a
call block) and icon_button, which raises without a label.
- steward.css: tone classes (badge-/dot-/stat- ok, warn, crit, info, dim),
.sr-only, .page-header, .stat-tile, .empty-state, .btn-icon and the form
classes .field, .field-label, .field-hint and .input. Info text is a 40%
mix (--blue): the house --fs-info-fg falls to 4.18:1 on its tint over a
hovered row, 40% holds 5.09:1.
- First users: the Alerts list, history and maintenance pages (headers,
state badges, empty states) and the alert forms' hints; the inline inp/lbl
style sets on the monitor and host forms become .input / .field-label.
- tests/core/test_macros.py renders each macro through the app environment.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5099 "A name containing a quote runs script when an admin clicks Delete".
Inline handlers like onclick="return confirm('Delete {{ rule.name }}?')" are
HTML-decoded before the JavaScript is parsed, so autoescaping did not stop a
name such as x');fetch(…);// from running.
- Every prompt (22 handlers, plus the docker prune and token revoke ones)
is now data-confirm="…", read through dataset by one submit listener in
base.html; nothing rendered is parsed as JavaScript.
- The remaining inline handlers take integers only (|int), and the time
range buttons pass this.dataset.range.
- tests/test_inline_handlers.py fails on any {{ }} in an inline handler that
is not |int, with a self-test; tests/_templates.py is the one template
walker the template guards now share.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5093 "Shared format filters and status_class driven by Settings → Thresholds".
- steward/core/formatting.py: human_bytes (binary units), human_rate,
human_duration (two largest units), ago (a <time> titled with the absolute
UTC time) and status_class(value, kind, thresholds) -> ok/warn/crit/unknown,
which raises on a kind the thresholds don't name.
- app.py registers them as the bytes, rate, duration and ago filters and the
status_class global; threshold_style is now graded by status_class (and caps
its weight at 500).
- Replaces the copies: docker's _human_bytes/_human_uptime, core alerts'
_duration, the host-agent fmt_bps/fmt_bytes macros, the Traefik and UniFi
byte-rate arithmetic, the swarm-failures ago macro and two uptime spans.
- Tests: a table per helper, and every template compiled against the app's
filters, so an unknown filter fails CI rather than a page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5092 "Tokens". base.html's inline palette was its own: white text on the
gold button (1.81:1), a red at 3.26:1 on a card, a warning yellow at hue
42 beside the gold accent at 46, a lavender page title, serif labels at
13px and weights up to 700.
- static/theme.css is design system 6's generated sheet (FabledSword + the
Steward accent). static/steward.css is base.html's stylesheet moved out,
with :root now aliases onto those tokens (--bg, --text-muted, --green…)
so every template keeps rendering; it states no colour of its own.
- Status text colours are Steward overrides recorded in the design system
(tokens 182-184): success #5FB87A, warning #E08A3C, error #EC7A6C. The
house -fg mixes read "up" as grey hint text (dE 8.2) and warning as the
accent (dE 11.5); these are >= 4.5:1 on page, card, hover and their own
15-25% tints. --orange is now the midpoint of warning and error.
- Primary button and active range button: dark text on gold (7.82:1).
Danger button on the house destructive fill. Badges and alerts take the
-fg colour on a tint; the gold badge uses --fs-accent-fg.
- Type: Inter body, Fraunces only for page titles and h1/h2, JetBrains
Mono for code; section and widget titles become the tiny caps label in
the body face; weights capped at 500.
- Removed: the lavender title and share-view gradient, #606080, and the
undefined --card-bg, --accent-faint, --accent-bg, --bg-alt, --border-low
and --row-alt (now real tokens); the dead settings/index.html.
- Guards (tests/test_style_tokens.py): every var(--x) used is defined;
steward.css states no colour or weight over 500; hex colours in
templates may only go down (ceiling 75, from 81 before this change),
each with a self-test that it can fail.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5091 "Local assets". base.html, the dashboard and the share view loaded
htmx and Chart.js from unpkg/jsdelivr, Gridstack from jsdelivr and their
fonts from Google. On an instance with no outbound network (rule 164) every
page lost htmx, and so every button and poll; and each page load told those
hosts about the visitor (rule 155).
- steward/static/vendor: htmx 2.0.4, Chart.js 4.4.4, Gridstack 12.6.0 (the
versions the pages already used), with their licences (0BSD, MIT, MIT).
- steward/static/fonts: Inter, Fraunces and JetBrains Mono, the design
system's faces (SIL OFL, licences beside them), every unicode-range
subset, with fonts.css. The serif face is now Fraunces in place of
EB Garamond (base) and Libertinus Serif (share view); applying Inter and
the type scale is the tokens step.
- Templates use url_for('static'); /static already passes every gate.
- Guard (tests/test_no_remote_assets.py): no template or stylesheet loads a
script, stylesheet, font or image from another origin (comments stripped,
example URLs in help text allowed), with a self-test that it can fail;
every url_for('static') file and every font the sheet names exists.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5066 "Alert history page". /alerts/history lists every alert transition
in a range (1h–30d, the shared range picker), newest first, all rules or
one (?rule=). Each row: from → to, the value against the rule's condition
("no data" for a silent one), email and webhook sent or failed with the
error, "muted · <window>" (and "notified when the window ended") taken out
of the event note, other notes, and a link to the Ansible run an action
started. Range and rule changes swap only the rows (HTMX). Viewer role.
The Alerts page links each rule's history and the whole history. The query
moves to alert_svc.history, which MCP get_alert_history now uses too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5065 "Renames carry their alerts". Alert rules, maintenance-window scopes
and metric history key a thing by (source, name), so a rename used to leave
the rules watching a name nothing reported under, and split the history.
- core/renames.carry_rename moves AlertRule.resource_name,
MaintenanceWindow.scope_resource, plugin_metrics and plugin_metrics_hourly
to the new name in the caller's transaction. A host rename also carries
its parts: host_agent "<host>:…" and docker "<host>/…" (HOST_SOURCES).
Matching is exact (a prefix compare, not LIKE or a collation range).
History left under the new name by something deleted is dropped first
rather than merged in.
- Web monitor edit, MCP update_monitor and web host edit call it; MCP
answers with `carried` (it used to return a warning naming stranded rules).
- A name another monitor of the same type, or another host, already has is
refused on create and rename: two would share one history.
- The rename is audited (monitor.updated / host.renamed) with what moved.
Tests: an integration test renames a firing monitor through the web route
and finds its rule still firing and cleared by a sample on the new name,
its window and raw+hourly history moved and an orphan's history dropped; a
host rename moves agent and container parts but not a longer-named host's
or another source's; a guard that every host_agent resource is "<host>" or
"<host>:…".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The rule name goes into the Subject header, and EmailMessage refuses a
header holding a line break, so the whole email failed. The web form can't
enter one but MCP takes any string. The subject now folds whitespace.
Found by the new From-fallback test in CI run 8349.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5064 "Notifications that arrive":
- render_webhook: alert.* is escaped to sit inside a JSON string (a quote
or newline in a rule name no longer breaks the body); raw.* is the plain
value for |tojson; a missing value reads "no data"; the template runs
sandboxed and the body is checked to be JSON before it is sent.
- SMTP: 15s timeout; port 465 connects with implicit TLS, otherwise the
box means STARTTLS; a From address setting (falls back to the username).
- A webhook error no longer carries the URL, which is a secret.
- Settings → Notifications: "Send a test email" / "Send a test webhook"
send a sample alert with the form's values (unsaved) and show the result
and the body that went out; template help names alert.* and raw.*.
- Notification audits go through audit_request; the test send is audited
as settings.notification.tested, with the channel and outcome only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Notifications and alert actions now start only once the evaluating
transaction commits, and never for one that rolls back. record_metrics
queues them on the session (_after_commit), and after_commit/after_rollback
listeners on the sync session spawn or drop them. Spawned tasks are held in a
module set so they cannot be garbage-collected before they run.
A maintenance window no longer skips evaluation. Alerts in its scope still
change state, but their notifications and actions are held, and each event
says so ("notification muted: maintenance window <name>"; also on no-data
and resume events). notify_after_maintenance, run every minute with the
silence check as the new core task alert_reconcile, sends one notification
for an alert still firing after its window ends and runs its action. The
event's note records the release.
Alert actions go through runner.trigger_run like every other run path, so
the playbook and the inventory must be ones the source lists (#5034 guards).
The firing event links to the run, or its note says why the action did not
start. The maintenance page and form wording now say what a window does.
Tests:
- integration: a rolled-back evaluation notifies nobody, and dispatch waits
for the commit; maintenance mutes while the state moves and notifies once
when the window ends; an action whose playbook is outside its source is
refused, with the note set;
- the alert-action test uses a listed inventory name and waits, with a
deadline, for the run the action starts;
- unit: _muted_by.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Agent offline is an ordinary rule now. The host_agent scheduler records
host_agent/<host>/seconds_since_seen every minute (from last_seen_at, or
registration time for an agent that never reported), replacing the task
that only logged stale agents. The catalog offers the metric, and picking
it on a new rule fills "> 180".
No data: each rule's state tracks its last sample and its usual gap,
smoothed so that an outage counts as at most three gaps. A core task,
alert_silence_check, runs every 60s and marks a rule no_data_since once its
series has had no sample for three usual gaps (at least 5 minutes; an hour
while the cadence is unknown). A never-sampled rule counts from when it
started watching. The mark sits beside the state, so a firing alert that
goes silent stays firing. Each silence records one event and sends one
notification (rule option notify_no_data, default on; muted in maintenance
windows). The next sample clears it with a "data resumed" event and
notification. Creating, changing, disabling or re-enabling a rule restarts
its watch.
Migration 0029 adds the columns and seeds each existing rule's last sample
and gap from its last day of data, so an upgrade does not flag every rule.
Shown on the Alerts page (a NO DATA badge with its start time), in MCP
list_alerts, list_alert_rules and get_host (state "no_data", no_data_since),
in get_overview findings (alert_no_data, or a firing alert marked silent)
and in the weekly report. The rule form has the notify option, and its
metric reference now matches what traefik records and lists host_agent.
Emails say "no data" and carry the event's note.
The metric-catalog guard also reads positional MetricBatch.add arguments.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5061 "Metric catalog belongs to what emits: plugins register their
metrics; names match what is recorded".
- The static METRIC_CATALOG is gone. metric_catalog() is core's own sources
(the four monitor types and "steward", the app's open_fds/open_fds_pct,
which were recorded but never offered) plus each loaded plugin's
get_alert_metrics() export. A disabled plugin's sources leave the picker.
- traefik offers what it records: error_rate_4xx_pct and error_rate_5xx_pct.
The old "error_rate" was never recorded, so a rule on it could not fire.
- tests/core/test_metric_catalog.py walks every MetricBatch.add /
record_metric call (traefik's tuple loop included) and fails on a recorded
metric the catalog lacks; it shows on a synthetic writer that it can fail.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5060 "One evaluation path: record_metrics batch API; host-agent ingest
evaluates rules".
- record_metrics(session, rows) writes metric rows and evaluates the rules
that watch them, reading rules, states and maintenance windows once per
batch. Each rule sees its key's samples oldest first, so an agent catching
up after an outage counts consecutive checks as they happened.
- Host-agent ingest now goes through it. Before, it wrote plugin_metrics
directly and no host_agent rule could ever fire. The catalog gains the
per-mount disk_used_pct ("<host>:<mount>").
- MetricBatch gathers a transaction's samples; the docker ingest, traefik,
unifi, snmp, monitor and self-monitor writers record one batch each instead
of a rule query per sample.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5059 "Alert state machine: acknowledgement holds until resolve; N
consecutive OK to resolve; editing, disabling or deleting a rule resets it".
- The state machine is now next_step(), a pure function of one sample. An
acknowledged alert stays acknowledged while breached; before, the next
breaching sample re-fired it. Firing and acknowledged alerts resolve after
consecutive_ok_required OK checks in a row (new rule column, default 1;
rule form, rules list and MCP create/update_alert_rule clear_after_checks).
- Changing what a rule watches (source, resource, metric, operator,
threshold) or disabling it resets its alert to inactive; a firing or
acknowledged one records a resolved event whose new note says why, and
nobody is notified. Web and MCP edits share alerts.service.apply_rule_change.
- Migration 0028: consecutive_ok_required, consecutive_ok_count,
alert_events.note, and metric_value nullable (a reset has no sample).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5037 "Secrets at rest and in run output: webhook URL and git token
encrypted, askpass via env, secret values scrubbed from logs".
- webhook.url joins SECRET_KEYS; existing values are sealed at startup by
migrate_plaintext_secrets. Settings -> Notifications no longer echoes the
URL: it names the saved host, keeps it when left blank, and has a Remove
box.
- Git source http_token is sealed inside ansible.sources on save and at
startup, opened in get_sources, and counted by the undecryptable-secrets
scan as ansible.sources[<name>].http_token.
- The askpass script holds no token: git runs it with STEWARD_GIT_TOKEN in
its environment and printf prints it as data.
- The executor masks every secret the run is handed (become/vault
passwords, bootstrap passwords, sensitive run vars, and the per-host agent
tokens the host_agent plugin now passes as redact) in each line before it
is broadcast, logged or stored.
- Bundled "Write agent config" tasks are no_log; a test fails if any bundled
task templating steward_token lacks it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5036 "SSO and LDAP: users keep their auth source, no merging into local
accounts, is_active checked, LDAP TLS real and filter escaped".
- users gain auth_source (local/ldap/oidc) and external_id (migration 0027).
A provider account is found by (source, subject) only: the LDAP entry DN
or the OIDC sub. A first sign-in whose username or email another account
holds is refused with a message naming why; a disabled user stays out.
- Local passwords sign in local accounts only.
- LDAP: port 636 is LDAPS, otherwise StartTLS before any bind; the server
certificate is checked against the system CAs or a CA pasted in
Settings -> Auth. The login name is escaped into the filter (RFC 4515),
and an empty password is refused before the directory sees it.
- OIDC: the ID token must carry this sign-in's nonce and the userinfo sub.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5035 "Sessions: lifetime applied, role and is_active re-read per request,
logout is a POST, failed logins throttled".
- steward/core/proxy.py: trusted_value / is_https / client_ip read
X-Forwarded-Proto and -For only as far as the trusted proxy hops
(Settings → General, default 1, 0 = trust none).
- The session cookie is Secure, and HSTS is sent, only on requests that
arrived over HTTPS by that rule; plain-HTTP and LAN use are unchanged
and nothing redirects.
- A login lasts Settings → General's session lifetime from when it began
(session.permanent + a login_at stamp checked every request — the
cookie's own age slides on every response). A deleted or deactivated
user is logged out on their next request; a role change applies at once.
- Logout is a POST (a form in the sidebar); GET /logout is gone.
- Failed logins pause an address after login.max_failures within
login.lockout_minutes (Settings → General; defaults 5 in 15 min),
keyed by address so nobody can lock a user out by name. Failures are
audited as auth.login_failed with the address.
- to_access_cfg projects the new settings into app.config at startup and
on save, beside the existing projections.
- Integration tests seed real users and a login_at, since every request
now re-reads the session's user.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5034 "Playbook runs: path and inventory checks in trigger_run, connection
vars admin-only, confirm enforced server-side, web actions audited".
Ansible runs inside the Steward container, so an operator who could name
any playbook path, set ansible_connection=local or ProxyCommand ssh args,
or put {{ lookup('pipe', …) }} in a value could run commands on Steward.
- steward/ansible/guards.py: check_playbook (the source must list it),
check_inventory_scope (steward:* or an inventory file the source has),
check_vars_for_role (ansible_* beyond host/port/user/password/become,
and any Jinja in a value, are admin-only), changed_vars (an edit is
held to the role only for what it changes).
- trigger_run applies the path checks, so every run path — web, MCP,
alerts, schedules, plugins — gets them.
- Variables are checked by role on the web run form, host-page runs,
schedule create/update, inventory group/target create/update, and MCP
run_playbook / create_schedule / update_schedule.
- A steward:confirm playbook needs the confirmation ticked on the server
side for a real (non-check) web run, as MCP already required.
- Audited: runs, cancel, schedule create/update/enable/disable/delete/
run-now, playbook create/update/delete, inventory group and target
create/update/delete (variable names only, never values). One
audit_request helper serves the web routes.
- The CSRF form guard and the route guard now each test that they fail
on a synthetic broken input (rule: a guard must be able to fail).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5033 "Share links open only their own dashboard; every route without a
role is on an explicit allow-list". Any live share token in ?s= made the
caller a viewer on every viewer route — host pages, container logs, any
dashboard — and the host-agent widget routes had no role check at all.
- steward/dashboard/share.py: a share token opens, read-only, exactly the
widget fragment URLs its own dashboard loads — same path, same
configured parameters; only the time window (hours/range) may change.
- require_role marks its view; public_route marks the deliberately open
ones (login, setup, health, the share page, agent ingest, install.sh,
agent.py) with the reason. A before_request refuses any view with
neither, so a route that forgets its decorator is closed, not open.
- /plugins/host_agent/widget and /widget/history now require viewer.
- The history widget's range buttons keep the rest of the query (and so
a share link's token) instead of rebuilding it; the uptime widget's
"Full SLA" link shows to logged-in users, not to share viewers.
- tests/test_route_access.py: a source scan that fails on any unmarked
route in core or bundled plugins, the default-deny refusal, and the
share matcher's cases.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5032 "Plugin install takes its URL and checksum from the catalog, requires
the checksum, and extracts only inside the plugin folder". The install form
carried the download URL and checksum as hidden fields, so any admin POST
could make Steward fetch and import code from anywhere, unchecked, and a zip
member like ../x.py was written wherever it pointed.
- The route takes only the plugin name and reads download_url and
checksum_sha256 from the configured catalogs; a name no repository lists
is refused. Installs are audited (plugin.install / plugin.install_failed).
- The checksum is required; the name must be a plain identifier.
- extract_plugin refuses any member that resolves outside the plugin
folder, caps the download at 50 MB and the unpacked size at 200 MB.
- The zip unpacks into a staging folder and is swapped in whole: a refused
or failed install leaves the installed copy untouched, and an update no
longer leaves the old version's deleted files behind.
- The catalog shows "Can't install: the catalog lists no checksum" in place
of the button for an entry without one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5031 "Forged-request protection: SameSite cookie, per-session CSRF token
on every non-GET, HTMX header". A page on another site could make a
logged-in browser POST to Steward with the session cookie attached.
- steward/auth/csrf.py: a per-session token, checked in one before_request
for every POST/PUT/PATCH/DELETE (header X-CSRF-Token or form field
csrf_token, constant-time compare). The bearer-authenticated host-agent
ingest is exempt; /mcp never reaches Quart.
- Every POST form renders {{ csrf_field() }}; base.html carries the token
in a meta tag and adds it to every HTMX request; the dashboard's fetch()
calls send it.
- The session cookie is SameSite=Lax; login issues a fresh token.
- PUBLIC_BASE_URL is now loaded from Settings at startup (it was only set
when Settings → General was saved).
- tests/test_csrf.py: refusal, acceptance, rotation, the ingest exemption,
and a guard that every POST form in the templates carries the field.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5015 "Built-in host-agent playbooks make an MCP caller supply steward_url"
- run_playbook, create_schedule and set_alert_action fill a steward-builtin
playbook's steward_url when the caller leaves it out or empty, from
Settings → General → Public base URL, else the scheme and host the MCP
request arrived on — the same source the web actions use.
- run_playbook's reply names what it filled (filled_by_steward), and
get_playbook shows that value as the variable's default.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5014 "Ansible run scope is shown as a raw id"
- inventory_gen.describe_scope / scope_labels turn "steward:target:<uuid>"
into "target web-1", "steward:group:<uuid>" into "group Docker Swarm",
and a deleted one into "target abcdef01 (deleted)".
- MCP run and schedule rows carry scope_label beside the stored scope; the
ansible.run.started audit detail records it too.
- The web run page and schedules table show the name, with the stored
scope as the tooltip.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5012 "get_host cuts `parts` without saying so"
- The agent's sub-resources come back as mounts (fullest first),
interfaces and disk_io (busiest first) and temperatures (hottest first),
each bounded on its own with total, truncated and a `more` that names
parts_limit and query_metrics.
- Per-core CPU is a summary (count, min/avg/max, three busiest), so 38
cores no longer push every interface out of the reply.
- New argument parts_limit (default 20, max 200).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MCP-8 #4972 "Connect card, docs, and end-to-end integration test".
- AI access page: a "Connect a client" card with the /mcp URL (from
Settings → General → Public base URL, else the request host) and recipes
for Claude Code, Claude Desktop (mcp-remote) and other Streamable HTTP
clients, plus the reverse-proxy note. The token-created page shows the
same recipes with the new token filled in, ready to paste.
- docs/core/mcp.md: connecting, the reverse-proxy requirements, the
security model (token ceiling, default deny, check-mode default, secrets,
audit), tools by role, response conventions, adding tools.
docs/architecture.md describes the /mcp dispatch outside the blueprint
pipeline; docs/reference/code-map.md gains an MCP section.
- tests/integration/test_mcp.py: over HTTP, a viewer token lists no write
tools and is refused one by name; an operator token reads get_overview
and its maintenance-window create/end land in the audit log as
"<user> via MCP" with the token id. (Revoked → 401 was already covered.)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MCP-7 #4971 "Monitoring-config write tools: monitors, alert rules, acks,
maintenance".
The validation the forms did inline (or not at all) moves to the services,
and both the routes and the tools call it, so bad input is refused the same
way with the same reason:
- monitors/service: monitor_fields, build_config, normalise_target. Port
1-65535, status 100-599, timeout 1-60, method GET/HEAD/POST, headers an
object of strings, a key the type does not take is named. The forms now
answer 400 with the reason instead of redirecting silently or a 500.
- alerts/service: METRIC_CATALOG and SOURCE_MODULES, resources_for,
metrics_for, rule_fields (source, resource and metric from the same
choices the rule form offers), add_rule, acknowledge, window_fields.
Tools: create_monitor, update_monitor (config merged over the stored one;
a rename warns about rules still watching the old name),
set_monitor_enabled, create_alert_rule, update_alert_rule,
acknowledge_alert, create_maintenance_window (duration or end),
end_maintenance_window (ends one in progress, cancels an upcoming one) —
operator. delete_monitor, delete_alert_rule — admin. get_alert_rule_options
— viewer. Every write is audited as "<user> via MCP".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MCP-6 #4970 "Automation tools: playbooks, runs, schedules, alert actions".
Viewer: list_playbooks, get_playbook (secret defaults masked), list_inventory
(variable names only), list_ansible_runs, get_ansible_run (log tail + grep),
list_schedules, wait_for_run (polls, 120s cap).
Operator: run_playbook (check mode unless check=false; a steward:confirm
playbook applies only with confirmed=true; secret variables reach the run
but are never stored, audited or returned), cancel_ansible_run, and
create/update/delete_schedule (secrets refused, check by default).
Admin: set_alert_action, and the docker plugin's prune_docker, which
previews from the disk report unless check=false (Ansible's check mode
skips the prune commands, so a check run would show nothing).
Every write is audited as "<user> via MCP" with the token id; the audit
tool's user filter includes those rows. prune_extra_vars moves to the
docker service so the page and the tool share it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MCP-5 of milestone 461 "MCP server: Steward as an LLM-inspectable,
LLM-operable hub". Contributed through each plugin's register_mcp_tools
hook, so they exist only while the plugin is enabled.
docker (viewer, read-only):
- list_containers: filter by host, status, compose project, or problem
(restarting, crash_looping, unhealthy, oom_killed, exited_nonzero);
problems sort first and each row names what is wrong.
- get_container: state, CPU/memory history across the raw/hourly seam,
lifecycle events. A unique part of a swarm task name is enough.
- list_swarm_services: replicas per node (agent-less placements as ghosts)
and the swarm's nodes.
- search_container_logs: text (% and _ literal), stream, range, context
lines, paging by a full-precision `before` cursor; over-long lines keep
both ends and say how much of the middle was cut.
- get_docker_disk_usage: bytes per host, reclaimable first.
- mcp_overview: failing containers and under-replicated swarm services
join get_overview.
host_agent: list_agents (reporting / stale / pending, version, OS), and an
overview finding for agents registered over 30 minutes ago that never
checked in.
Settings > Thresholds gains the TLS-expiry warn/crit days (were 14/3 in
code) and the crash-loop deaths-per-hour count (3), which get_overview
and the docker tools now read.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MCP-4 of milestone 461 "MCP server: Steward as an LLM-inspectable,
LLM-operable hub". Twelve read-only viewer tools:
- get_overview: what is wrong now, ranked critical-then-oldest. Failing
monitors with their error and when the run of failures began, firing,
pending and still-breaching acknowledged alerts, TLS certificates expiring
within 14 days, agents that stopped reporting, hosts over the configured
CPU/memory/disk thresholds (Settings > Thresholds), failed Ansible runs in
the last 24h, and findings from enabled plugins through a new optional
`mcp_overview(db)` plugin hook. Each finding names the call that shows
its evidence.
- list_hosts, get_host, get_host_metrics
- list_monitors, get_monitor (HTTP header values and URL passwords are
redacted)
- list_alerts, get_alert_history, list_alert_rules, list_maintenance_windows
- query_metrics: the whole metric store, narrowed one argument at a time
(sources > resources > latest values > a series summary)
- list_audit_events
Shared shaping in steward/mcp/tools/_shape.py: UTC ISO times with age_s,
bounded lists that say when and how they were cut, series as
min/avg/max/p95/last plus a ≤60-point series, lookup by id or
case-insensitive name with a did-you-mean, and a range argument that
refuses an unknown key instead of reading it as 24h.
New core reads in steward/core/metric_series.py (sources and resources
via loose index scans, latest values via DISTINCT ON, series across the
raw/hourly-rollup seam). monitors/service.py gains monitor_states,
down_since and recent_failures; agent_overview_by_host now carries the
fullest disk and the last sample time.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MCP-3 of milestone 461 "MCP server: Steward as an LLM-inspectable,
LLM-operable hub". The /mcp tools run outside Quart's request pipeline, so
the queries they share with the pages now live in plain functions that take
an open session:
- steward/hosts/service.py: uptime, status, monitors, agent metrics,
agent overview (stale threshold passed in, not read from current_app),
CPU sparks.
- plugins/docker/service.py: bucket, container_history, host_map,
list_containers, container_events, query_logs.
- steward/alerts/service.py: active_alerts (also used by core/reports),
list_rules, list_maintenance_windows.
- steward/monitors/service.py: monitor_rows (was routes._rows_data).
Routes are repointed and render the same data. One behavior fix rides
along: the container-log search now escapes % and _ so they match
literally instead of acting as LIKE wildcards.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Second step of milestone 461. An MCP client (Claude Code, Claude Desktop
via mcp-remote, Open WebUI) connects to /mcp with a personal API token.
- mcp>=2.2 (SDK v2 MCPServer). Mounted by wrapping app.asgi_app, so /mcp
never enters Quart's pipeline; stateless streamable HTTP so a redeploy
never strands a client on a session id the new process didn't issue.
- Auth: Authorization: Bearer stw_… resolves to a Caller (user +
effective role) in a ContextVar; 401 with WWW-Authenticate otherwise.
The web pipeline's database gate runs first, so a DB restart answers
503 instead of failing inside a tool.
- Registry, default-deny: register_tool(fn, role=, read_only=) is the only
door; role and read_only are required. Tools above the caller's role,
or from a disabled plugin, are neither listed nor callable. Undeclared
arguments are refused (not silently dropped); a tool's ValueError
reaches the agent as its reason; a DB failure mid-tool says so.
- Plugins contribute tools via an optional register_mcp_tools(register)
export, all-or-nothing, replaced on hot-reload. Documented in
docs/plugins/writing-a-plugin.md.
- First tool: whoami.
- Tests drive app.asgi_app directly (Quart's test client never reaches
/mcp); the integration test mints a real token against Postgres and
runs initialize → tools/list → whoami → revoke → 401.
Task #4966 "MCP-2 — /mcp transport".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
First step of milestone 461 (MCP server). A token is the credential an
MCP client presents as `Authorization: Bearer stw_…` to act as its user.
- api_tokens table (0026): sha256 hash only, display prefix, role
ceiling, expiry, revocation, throttled last_used_at.
- steward/core/api_tokens.py: mint, list, revoke, resolve. Resolution
re-reads the user's current role and acts with the lower of that and
the token's ceiling, so demoting or disabling a user weakens or kills
their tokens with nothing to revoke by hand.
- /ai-access/ (sidebar → Automation → AI access): every user manages
their own tokens, can't grant above their own role, sees the token
once. Session-only: a share-link viewer is refused. Create and revoke
are audited.
- Integration tests drive the routes against real Postgres so their
transactions run for real (#884 escaped CI behind a mocked session).
Task #4965 "MCP-1 — Personal API tokens".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The admin banner counts every plugin load failure, but settings/plugins.html
only rendered failure reasons inside the plugin card macro -- which iterates
discovered plugins. A plugin that failed BECAUSE it has no directory was
therefore counted in the banner and shown nowhere on the page the banner links
to. That dead end is what the operator hit: "1 plugin failed to load" on the
dashboard, nothing wrong on the details page, and no way to clean it up.
Folds the reason into the "Configured but not installed" rows rather than
adding a competing section, and drives that list from the UNION of stored
plugin.* settings and recorded load failures instead of stored settings alone.
A failure with no stored row -- a plugin enabled by a DEFAULTS entry, which is
exactly what plugin.http was -- would otherwise still be counted and still be
invisible.
Rows carry `removable`: only a plugin with a real stored row can be cleaned up,
so a default-declared one shows the reason and an explanation instead of a
Remove button that would delete nothing while the default reasserts it. Rows
with a reason read as an error (red, "Failed to load"); rows that merely have
leftover config stay a warning. A discovered plugin that failed is excluded --
its own card already shows the reason.
Replaces find_orphaned_plugin_names outright rather than adding a second
overlapping helper; it had one caller.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
