Operator chose a single canonical summary. Repurpose the status_overview
widget into 'Overview' (host count + monitor up/down/pending + Alerts link;
key kept so existing dashboards upgrade in place) and remove the redundant
fixed top strip from the dashboard view. Drops _get_summary_stats and its
now-unused PingResult/DnsResult imports (rule 22).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Hosts-Overview: move the CPU sparkline inline next to the cpu % value it
represents (was floated far right on the name line, reading as unrelated).
- Host Agent history widget: caption the chart with the host it represents
(the panel title is generic) — links to the host hub.
- History widget: snap the y-axis to a stable 0..next-10%-band ceiling instead
of auto-scaling to the exact peak, so the 'zoom' no longer jumps each poll.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Restore the host_agent fleet-glance widget to one horizontal row per host
(name left, metric cells + sparklines right) instead of the multi-column
block grid, and zebra-stripe odd rows so adjacent hosts read as distinct.
Names still wrap rather than hard-truncate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The agent reports every few seconds, so multi-hour history series were hundreds–
thousands of points — a dense, noisy line (esp. CPU). Bucket-average server-side
to ~120 points (keeps the shape, drops the noise) for both the history-graph
widget and the host-detail charts. Also fix the chart tooltip title showing the
raw epoch-ms (e.g. 1,781,720,471,459) — format it as HH:MM.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address graphical issues raised from the dashboard screenshot:
- No more truncated host names. Hosts-Overview and Host-Agent-Resources put the
host name on its own line (full, wraps if needed) with its data grouped beneath
it; ping/dns (.ping-name) and uptime widget names wrap instead of ellipsis.
Prefer vertical overflow over cramming/truncating a row.
- History graph fills the panel: drop the fixed 0–100 y-axis ceiling (beginAtZero
+ 8% grace) so the lines use the vertical space instead of hugging the bottom;
axis labels still show real %.
- Container-query breakpoints: the widget body is now a query container, so
fragments restyle to their OWN panel width. Host-list widgets flow into 2 cols
≥520px and 3 cols ≥900px (.host-blocks) — use the width, remove vertical
deadspace — and collapse to one column when narrow. Mirrored into the share view.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Operator ask: show the graphs next to their fields in the fleet widget, like the
host page's AGENT panel. Each row now renders cpu/mem/disk/load as a value with a
trend sparkline beneath it (1h window), instead of bare numbers.
- _fleet_rows fetches a per-host recent series (cpu/mem/load host-level, disk from
the root mount) in one 1h query and attaches a sparkline per metric to each row.
- widget_table.html lays out a metric cell (label + threshold-coloured value +
sparkline) per field, mirroring panel.html. Threshold colour is computed in the
loop and passed into the cell macro (keeps Jinja macro scope clean).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Milestone 72 phase D — the dashboard view IS the edit surface (operator ask):
- /d/<id> renders the grid via Gridstack in STATIC mode — positioned exactly like
before, live HTMX widget bodies keep polling. An "Edit" button flips the same
grid interactive (grid.setStatic(false)) in place, so you drag/resize over real
data. "Done" flips it back. Layout autosaves on change.
- The widget picker is now a bottom drawer (position:fixed overlay) revealed by
body.dash-editing — so the dashboard width is identical entering/leaving edit.
- Add: POST returns a single grid item; JS inserts it + grid.makeWidget +
htmx.process so it loads live data. Remove: POST 204 + grid.removeWidget.
Per-panel drag handle + remove ✕ are in the DOM for editors, shown only while
editing.
- Gridstack loads for everyone (viewers get the static grid); edit wiring is
gated on can_edit. Mobile collapses to one column.
- Removed the separate /d/<id>/edit route + edit.html + _edit_panels.html
(rule 22). Dashboard-list Edit and new-dashboard create now deep-link
/d/<id>?edit=1 which opens edit mode on load.
Browser-only behaviour — CI can't exercise it; needs an operator visual check.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Milestone 72 phase C — bring the host-view graphs onto the dashboard:
- host_resource_history widget reworked into a real host-view chart: epoch-ms
linear axis (no Chart.js date adapter), themed like the host-detail charts,
maintainAspectRatio:false so it fills the resized panel, unique canvas per
widget instance (wid), and empty states ("pick a host" / "no metrics yet").
Was previously unusable — it had a broken time axis and no way to choose a host.
- Add a "host" param type: the edit form renders a live dropdown of hosts
(dashboard routes now pass the host list to the editor); the chosen host_id is
stored in config and fed to the widget.
- Hosts-overview widget gains a per-row CPU sparkline (last hour) via the shared
sparkline_svg helper — the host-view at-a-glance trend, on the main widget.
Charts/sparklines render only in the browser, so CI can't exercise them — needs
an operator visual check.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Milestone 72 phase B — Grafana-style drag-resize grid for dashboard widgets:
- DashboardWidget: replace `position` with a 12-col grid placement
(grid_x/grid_y/grid_w/grid_h). Migration 0021 backfills the old position
order into a 3-up grid (4 cols x 4 cells each) and drops position.
- View + share render a static CSS grid from x/y/w/h: fixed cell height
(h * 70px) with the body scrolling, so the arranged layout is what's shown;
collapses to a single column under 820px.
- Edit view: Gridstack.js 12.6.0 (vanilla, CDN, pinned) — drag the title bar
to move, drag a corner/edge to resize; every change autosaves to a new
/d/<id>/edit/layout endpoint. Replaces the SortableJS position reorder.
- add_widget appends at the bottom-left; remove no longer renumbers.
_get_widgets now orders by grid position (drives DOM + mobile fallback order).
Note: Gridstack drag-resize is browser-only, so CI can't exercise it — needs an
operator visual check of the edit experience.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Milestone 72 phase A (clarity wins, no schema change):
- Add shared metric_style() macro in _macros.html: colours a numeric metric
amber (>=warn) / red (>=crit) on the value ITSELF, not just the status dot.
Defaults 80/90 for percentage gauges; load /core uses warn 80 / crit 100.
Applied to CPU/mem/disk across host_agent panel + fleet widget + the unified
hosts-overview widget.
- Link every host reference to the host hub (/hosts/<id>) in ping, dns,
hosts-overview, host_agent fleet, and uptime widgets; status widget entries
link to their own detail_url. All guarded on session.user_id so the public
share view degrades to plain text (the bare href carries no share token).
- Fix truncation: title= tooltips on all names that ellipsis, plus a hover
underline affordance on the now-clickable ping-name links.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
New maintenance/system_update.yml: cross-distro OS package upgrade (apt +
dnf/yum) with two run-form flags:
- restart_services: restart every service that needs it after the upgrade
(Debian via needrestart -r a, installed if missing; RHEL via
needs-restarting -s → systemctl try-restart).
- reboot_if_required: reboot the host only when a reboot is actually pending
(Debian /var/run/reboot-required; RHEL needs-restarting -r). Never reboots
otherwise.
Tagged steward:category=maintenance and steward:confirm=true (significant /
reboot-capable), so it shows in the run UI with the confirm gate and both flags
as fill-in fields. No code changes — auto-discovered from the builtin source.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Host agent panel (reporting state) gains a "Re-provision" collapsible (admin +
linked target + managed key): bootstrap user/password → provision.yml, which
reinstalls the steward account + managed key + agent. This is the missing path
after regenerating the managed key — Update alone can't fix a broken key.
- Remove the Ansible sections (run-playbook + target link) from the host EDIT
page — they were the stale free-text version and the wrong place. They live on
the host detail hub (dropdown + discovered variables). Edit page now links to
the host page; edit_host route simplified (no ansible fetch).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
When a session has expired, require_role now bounces to /login?next=<path> and
sends the user back there after re-login.
- middleware: safe_next_url() (same-site relative path/query only; rejects
off-site, protocol-relative, javascript:, and the auth pages — no open
redirect). _login_redirect() builds the next param; for HTMX requests it uses
HX-Current-URL (the page, not the fragment) and HX-Redirect so the whole
browser navigates instead of swapping the login page into a fragment.
- login GET/POST carry `next` (hidden field), validated, used on success for
local + LDAP; OIDC stashes it in session across the IdP round-trip.
- login.html: hidden next field + next on the SSO link.
- tests for safe_next_url.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
When ansible.ssh_private_key can't be decrypted (app secret key changed),
decrypt_secret returns the ciphertext unchanged; the executor was writing that
enc:v1: blob as the SSH key file → cryptic "Load key ...: error in libcrypto"
→ Permission denied. Now:
- build_credentials skips a still-encrypted key value (never writes ciphertext
as a key file).
- start_run broadcasts a plain-language run error: "The managed SSH key could
not be decrypted (the app secret key changed). Regenerate it in Settings →
Ansible and re-provision the host(s)."
The run still fails (no usable key), but the reason is now obvious instead of a
libcrypto error. Operator remedy: regenerate the managed key + re-provision.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Run UX: when start_run throws before/around launch (e.g. ENOSPC creating the
temp dir — the box is out of disk), the run was marked "failed" with empty
output. Now the exception is broadcast + written to the run output/results so
the run view shows e.g. "[run error] OSError: [Errno 28] No space left on
device" instead of a blank failure.
Widget audit follow-ups (no broken links were found; these are consistency):
- host_resource_history widget now charts root (/) disk, consistent with the
host panel (was the opaque "disk worst").
- host_resources widget: tooltip on the health dot explaining it warns on the
worst mount while the number shows root.
- status_overview widget detail_url /status → /status/ (avoid redirect).
- Normalize ad-hoc widget empty-states to the shared .empty style (wording,
which distinguishes "configured" vs "data yet", preserved).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bring the dashboard in line with the unified host IA + improved displays.
- New core "Hosts — Overview" widget (/hosts/overview/widget): one row per host
combining monitor status (ping dot + latency, uptime 24h) with the agent
glance (CPU / memory / disk root + stale flag), each row linking to the host
hub. Reads agent data from the generic PluginMetric table via a core-safe
_agent_overview_by_host helper (no host_agent import); freshness vs the
plugin's stale window. The granular Ping/DNS/Uptime/Agent widgets stay.
- Group the add-widget picker into Core monitors / Monitoring capabilities /
Integrations (a `group` field on every WIDGET_REGISTRY entry + section
headings in _edit_panels.html), matching the Settings → Plugins taxonomy.
- Fix the agent fleet widget rows to link to the host hub (/hosts/<id>) instead
of the old /plugins/host_agent/<id>/ page.
Scribe #903.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extend the playbook metadata convention with a namespaced `# steward:<key>:`
comment block:
- steward:category — free-text grouping label, shown as a badge in the browse
list and on the run form.
- steward:confirm — true/yes/1/on marks a playbook destructive; the run form
then requires a confirmation tick (required checkbox in the shared vars
fragment) before it can launch.
sources.discover_playbook_meta() parses description + category + confirm (first
match per key; `# description:` still primary, `# steward:description:` alias).
discover_playbook_description() now delegates to it. The browse list reads
per-playbook meta to show category badges + descriptions; the run-form and
playbook-vars fragments render the badge + confirm gate.
Bundled playbooks tagged: docker_prune → category maintenance + confirm true;
provision/install/update → category host-agent.
Docs: docs/reference/playbook-authoring.md updated (keys now implemented) and a
quick reference added next to the code at steward/ansible/PLAYBOOK_CONVENTIONS.md.
Tests added for category/confirm/alias parsing.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A self-contained guide (docs/reference/playbook-authoring.md) to the contract
between a playbook and Steward's run UI — the `# description:` comment,
vars/vars_prompt → fill-in fields, secret naming, hosts: all targeting,
managed credentials, and idempotency. Includes a "what Steward reads" summary
table, the metadata/extensibility note (only `# description:` today; reserved
`# steward:<key>:` namespace for future keys), and an annotated example.
Meant to be fed to another session authoring Steward-friendly playbooks.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Playbooks can ship a human description Steward reads and shows when one is
selected. Convention: a `# description: <text>` magic comment (Ansible rejects
unknown play keys, so a comment is the portable place — works for third-party
playbooks too); falls back to the first play's name:. sources
.discover_playbook_description().
Surfaced at the top of the shared _playbook_vars.html partial, which loads on
playbook selection in the host run form, schedules form, and browse run form.
All four bundled playbooks (provision/install/update/docker_prune) now carry a
description line. Unit tests added.
Scribe #900.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Each at-a-glance metric (CPU, Memory, Disk /, Load) now shows a 6h sparkline
(reused core.status.sparkline_svg + a recent-series query; disk uses the root
mount sub-resource) so trend/consistency is visible, not just the instant.
- Load is now normalized: "Load /core" = 1m load ÷ CPU cores as % (100% = run
queue matches capacity), comparable across different hardware. Cores derived
from the per-core CPU metrics already collected — no agent change. Raw load
in the tooltip.
- Added a "Pressure 10s" line: PSI cpu/mem/io (some, avg10), the
hardware-independent saturation signal already collected by the agent.
Scribe #898.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The "Disk (worst)" number was opaque at a glance. Show the root filesystem (/)
usage instead — what people actually care about — on the host panel and the
dashboard widget. "Worst" is kept only for the widget's health dot (so a full
/var or /data still warns) and on the full-metrics page (per-mount + worst
trend). Added hover tooltips defining CPU / Memory / Disk / Load.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A fresh install had no prompt for general.public_base_url, so first-run agent
installs/share links silently used the request Host header. Add a "Steward URL"
field to the first-run /setup page (create-admin), pre-filled with the current
address, with help text. setup_post saves general.public_base_url and applies
it to app.config immediately (no restart). Editable later in Settings → General.
Scribe #896.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The assertion failed ("Pass steward_url/token/pubkey") because those were
injected as inventory HOST vars, but the playbooks declared them in the play
`vars:` block — and play vars OUTRANK inventory host vars, so the empty
defaults won and the injected values never reached the play.
- Pass globals (steward_url, steward_pubkey, steward_user, agent_interval) as
extra-vars via the JSON -e @file (highest precedence, space-safe). Keep only
the per-host steward_token as an inventory host var.
- provision.yml / install.yml: drop steward_token from `vars:` so the host var
isn't shadowed; assertions use `| default('')` for the un-defaulted token.
This was the next layer under the inventory-format fix — first the inventory
wouldn't parse, now the injected vars actually apply.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Stop making operators type playbook paths and guess extra-var names.
- Reusable infra: shared ansible/_playbook_vars.html (the discovered-variable
fields) + ansible/_playbook_options.html; two HTMX endpoints —
/ansible/playbook-options (a source's playbooks, optional ?selected for edit)
and /ansible/playbook-vars (a playbook's vars:/vars_prompt: as fill-in
fields). browse _run_form.html refactored to include the shared partial.
- Host "Run a playbook against this host": source dropdown → playbook dropdown
→ variable fields, all chained via HTMX. Handler reuses _parse_run_params so
var__/secret__ fields flow through extra_vars_map + the unpersisted
secret_vars channel.
- Schedules: playbook free-text+datalist → source-dependent dropdown; fixed the
extra-vars edit pre-fill to read extra_vars_map (stale list key after the
earlier JSON-extra-vars change).
Scribe #895.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A failed provision looked successful in two ways:
1. The host panel showed the agent as deployed (metrics + Update/Rotate/Remove)
because provision/deploy mint the registration row BEFORE the playbook runs
and the panel keyed "installed" on that row. Now gated on the agent actually
checking in (reg.last_seen_at). Three states: reporting (metrics + lifecycle),
pending (token minted but no check-in → "no metrics yet, deploy may be
running/failed" banner + retry + Clear pending registration), and none
(install path).
2. The run reported success though nothing ran — ansible-playbook exits 0 on
"no hosts matched"/empty inventory. The executor now treats an empty PLAY
RECAP (returncode 0 but no hosts executed) as failed, with a clear failure
note. Non-zero exits and recap failed/unreachable were already caught.
Scribe issue #887.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
generate_inventory() emits Ansible's dynamic --list shape (all.hosts is a
LIST, vars under _meta) — valid only as an executable inventory script's
stdout. We were writing it to a static file and passing -i, so Ansible's yaml
plugin rejected it ("Invalid 'hosts' entry for 'all' group, requires a
dictionary, found ...list...") and fell back to implicit localhost → "no hosts
matched". Affected every steward:* scope run; surfaced on the first real
provision.
- New inventory_to_yaml(inv): convert the --list dict → a valid static YAML
inventory (all.hosts dict keyed by host, groups under all.children, group
vars preserved, injected per-host vars like steward_token retained).
- Wire it into runner.trigger_run, host_agent deploy + provision.
- executor writes the file as inventory.yml so the yaml plugin's extension
check reliably claims it.
- generate_inventory unchanged (still the --list dict); conversion happens at
write time. Unit tests added.
Scribe issue #885.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The "could not decrypt a stored secret" warning was generic, so an operator
couldn't tell which of the six secret settings was encrypted under an old key.
Thread the setting key through _decode → decrypt_secret(context=...) so the log
now reads e.g. "Could not decrypt stored secret smtp.password (wrong/rotated
key — re-enter it)". Pure diagnostic; decrypt behaviour unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
fetch_scope_targets() runs a SELECT that autobegins the session transaction
(SQLAlchemy 2.0), so the following `async with db.begin()` raised
"A transaction is already begun on this Session" → 500 on both
/plugins/host_agent/deploy and /provision.
Mint registrations directly into the autobegun transaction, build the
inventory while the ORM objects are still live, then await db.commit().
Escaped CI because unit tests use a mock db_sessionmaker; no integration
test exercises these routes (follow-up noted in Scribe issue #884).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Dev-only instance, no bookmarks — per family rule 22, fully remove old paths
instead of shimming them.
- Delete the /plugins/host_agent/ (index) and /plugins/host_agent/settings/
redirect routes; delete the now-dead host_list.html fleet template.
- Move the remaining management POST routes off /settings/ to /fleet/
(add-host, rotate-token, delete) — single canonical prefix.
- Repoint real callers to canonical URLs: dashboard widgets (host resources →
/hosts/, history → /plugins/host_agent/fleet/), the full-metrics page
breadcrumb + back link (→ the host hub), Settings→Ansible link, and the
agent panel's curl-install link.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Final phase of the host-IA unification (milestone 70).
- Settings → Plugins split into two tiers: "Monitoring capabilities"
(host_agent, http, snmp, docker — built-in host facets, surfaced via
Hosts/Status, on by default) and "Integrations" (traefik, unifi — external
systems, off until configured). Presentation only: a CAPABILITY_PLUGINS set
(overridable by plugin.yaml `kind:`) tags each plugin; module loading,
optional deps, and migrations are untouched.
- Drop the "default-enable a plugin" framing in the UI copy — capabilities are
described as built-in, not optional add-ons.
- Nav: remove the standalone "Uptime" item (folded into Hosts; still reachable
via the SLA button on the Hosts list).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Hosts list: new Agent column (latest CPU/mem read from the generic
PluginMetric table — no host_agent import) + an admin "Agent fleet" button.
- /plugins/host_agent/ (old fleet page) now redirects to /hosts/ (folded into
the hub; kept as a redirect so widgets/links don't 404).
- Agent management moved off the "settings" URL: the management page is now
/plugins/host_agent/fleet/ ("Agent fleet" — bulk provision/install/update +
registrations + curl install), reachable from the Hosts list. Old
/plugins/host_agent/settings/ redirects there. Per-host management lives on
the host detail page; this page is now explicitly the bulk/fleet view.
Milestone 70 phases 2-3. Phase 4 (plugins capability/integration split + nav
cleanup) next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Make Hosts the front-and-center hub. A host now has a real detail page at
/hosts/<id> that pulls its facets into one view, instead of management being
scattered across a nav-less Host-Agents area and the edit form.
- hosts: new GET /hosts/<id> detail route + hosts/detail.html. Shows the
monitors summary (ping/DNS status + latency + uptime 24h/7d/30d), an Ansible
section (linked target, link/create, run-playbook), and an embedded Agent
panel. Hosts list name links here; ansible-link redirects here.
- host_agent: GET /plugins/host_agent/panel/<host_id> — a self-contained HTMX
fragment embedded into the core hub across the plugin boundary (core never
imports plugin models). Shows live agent metrics + Update/Rotate/Remove when
installed, or the provisioning path when not: inline "generate managed key"
warning, a prompt to link an Ansible target first, then Provision (bootstrap
password) / Install (managed key) tied to the host's target scope.
Part of milestone 70 (Hosts hub). Phase 2+ will enrich the list, redirect the
old fleet/settings pages, and re-taxonomize plugins into capabilities vs
integrations.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Provisioning review corrections + the matching frontend, plus breadcrumb
header integration.
- provision.yml: authorize the managed pubkey with a regexp match on the
' steward-managed' comment so rotating the key REPLACES the host's steward
key in place instead of stacking a second authorized entry. Hand-added keys
(other comments) are untouched.
- update.yml (new): refresh agent.py + restart only. Does NOT rotate the token
or rewrite /etc/steward-agent.conf — the host keeps its identity. Asserts the
agent is already installed and fails clearly otherwise.
- host_agent /update route: runs update.yml as the managed steward user (no
token minting). Token rotation stays a deliberate action.
- settings/ansible/generate-key honors a safe relative `next` redirect, so an
inline trigger elsewhere returns to its page.
- Host Agents settings: reworked into a clear lifecycle — an intro card that
explains it runs Ansible to deploy the agent (+ inline "generate managed key"
warning/trigger when none exists), then three labelled cards: 1 Provision,
2 Install/enroll, 3 Update. Each explains what it does.
- base.html: breadcrumb now renders as a kicker line directly above the page
title (moved below alerts, tightened margin) so nested and top-level views
share one consistent header.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
When you click Run, Steward now parses the playbook and renders a field
for each declared variable instead of a blank extra-vars textarea. The
form loads on demand via HTMX (/ansible/run-form/<source>/<playbook>).
- sources.discover_playbook_variables: parse vars: defaults + vars_prompt:
(vars_prompt wins on name collision; non-scalar vars skipped; role/include
vars not traversed). Flags secret-looking names + vars_prompt private.
- Run-time values flow through a JSON extra-vars file (-e @file), which is
space/quote-safe — fixes a latent shlex-split bug in the old -e key=value
textarea path. executor.build_extra_vars_file (pure) + start_run merge.
- Secret-flagged fields are masked AND routed through an unpersisted
secret_vars channel (runner.trigger_run → start_run), so passwords entered
at run time never land in the DB / run history.
- Defaults shown as placeholders (not prefilled): an untouched field falls
through to the inventory/play default instead of overriding it.
- routes: run_form HTMX endpoint; _parse_run_params now returns
(params, secret_vars, err) and reads var__/secret__ fields. Schedules drop
secret vars (can't prompt unattended).
- templates: ansible/_run_form.html fragment; browse.html rewired to HTMX,
static JS run-form removed. Advanced section keeps limit/tags/check + a
free-form extra-vars escape hatch.
- tests: test_playbook_variables.py (discovery + extra-vars file).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Turn the agent-install playbook into a full provisioning + maintenance
path. Solves the bootstrap chicken-and-egg: first contact uses an
operator-supplied password (one run, never stored), which creates a
dedicated `steward` login account with NOPASSWD sudo + Steward's managed
public key. Every run thereafter connects as `steward` with the managed
key — fully unattended (scheduled prune, agent updates).
- core/crypto: generate_ssh_keypair() — ed25519, OpenSSH formats.
- settings: ansible.ssh_public_key (non-secret, displayed) + ansible.ssh_user
(default steward); to_ansible_cfg extended.
- settings UI + route: "Generate managed key" (private encrypted, public
shown to copy) + SSH-user field.
- executor: build_bootstrap() writes a 0600 vars file (-e @file) for the
per-run user/password — never argv, never DB, never logged; drops the
managed key when a bootstrap password is given; --user floor from the
global ssh_user when no override.
- runner.trigger_run: pass-through `connection` kwarg, deliberately NOT
persisted on AnsibleRun.params (password stays out of the DB).
- bundled/host_agent/provision.yml: create steward user + authorized_keys
+ /etc/sudoers.d/steward (visudo-validated) + agent install.
- host_agent: /provision route + "Provision a fresh host" card (bootstrap
user/password; injects pubkey + user + token as space-safe JSON hostvars).
- Dockerfile: add sshpass (Ansible shells out to it for password SSH).
- tests: keypair generation + build_bootstrap (secret stays off argv).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Closes#550 (all four):
- Cancellation: track live subprocesses; POST /ansible/runs/<id>/cancel
(operator) SIGTERMs then SIGKILLs after a grace; new 'cancelled' status
(+ migration 0019, ALTER TYPE in autocommit). Queued runs cancel cleanly
before launch. Cancel button on run detail.
- Concurrency: global semaphore (ansible.max_concurrent_runs, default 3,
Settings→Ansible) caps simultaneous runs; excess show 'queued' (new status)
until a slot frees. Semaphore bound lazily per running loop.
- Structured results: parse PLAY RECAP into per-host ok/changed/unreachable/
failed/skipped + capture failed-task lines, stored in new results JSON
column (migration 0020); rendered as a host-summary table on run detail.
Keeps live streaming (no json-callback swap).
- Retention: full output written to a persistent log artifact
(/data/ansible/runs/<id>.log, env-overridable) beyond the 1 MB DB cap and
across restarts; in-memory replay buffer bounded + GC'd after completion;
Download-log route. Boot reconciliation now also sweeps stale 'queued'.
Unit tests for recap parsing + cancel flagging. Status colors updated across
run list / detail / schedules.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds a breadcrumb trail to nested views for orientation. Mechanism: a
{% block breadcrumb %} slot in base.html (+ styling) and a shared crumbs()
macro in templates/_macros.html; each nested page fills the block with its
trail (root→current, last item is the current page). Pages without the block
render no bar, so top-level nav roots stay clean.
Applied to: Ansible (browse, schedules, playbook editor, run detail) +
inventory (targets/groups + detail), host_agent (fleet, host detail,
settings), hosts form, settings tabs (ansible/auth/notifications/plugins/
reports + plugin detail), dashboard (list, edit), and alerts (rule form,
maintenance + new). Dynamic labels (host/target/run/dashboard names) come
from the page context — no route changes. All 60 templates Jinja-compile.
Task #873.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The inventory CRUD UI (/ansible/inventory/targets + /groups) existed but was
unreachable from the main Ansible pages — only a buried text hint pointed to
it. Add an "Inventory" button to the Runs, Browse, and Schedules headers, and
"← Ansible" back-links on the inventory target/group pages, so the targeting
features (manual runs, schedules, Deploy-via-Ansible) are findable.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds create/edit/delete of playbooks from the UI (admin only), so a homelab
user without a git workflow can author automation in-app. A new always-present
writable local source "steward-local" (/data/ansible/playbooks, env-overridable,
created on first save) is editable alongside operator local-dir sources; the
bundled and git sources stay read-only (git is GitOps, clobbered on pull).
sources.py: write_playbook / delete_playbook (traversal-guarded, .yml/.yaml
only) + validate_playbook_yaml (YAML + play-list check) + is_editable_source.
routes.py: /playbooks/new, /edit, /save, /delete (admin). Browse gains a
"New playbook" button and per-playbook + view-page Edit/Delete for editable
sources. Plain textarea editor with save-time YAML validation. Unit tests for
write/delete/guard/validate.
Task #579 — completes milestone #37 (Ansible automation).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Implements #253's framework: a small core capability registry
(steward/core/capabilities.py) where a module/plugin publishes a named,
role-gated action and a consumer discovers it via has_capability() and runs it
via invoke_capability() — no hard import, graceful degradation, permission
propagation (actor role checked against the capability's required_role).
Core publishes "ansible.run_playbook" (operator) wrapping ansible.runner.
trigger_run (extended to accept a caller-built inventory). First consumer: the
host_agent plugin gains "Deploy via Ansible" on its settings page — pick an
inventory target/group and it installs/updates the agent via the bundled
host_agent/install.yml, minting a fresh token per host and injecting it as an
inventory hostvar (turning per-host curl|sh into one run). Exposed role
ordering as middleware.role_meets. Unit tests for the registry + role checks.
Task #253 (milestone #37).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ships a read-only "steward-builtin" source (steward/ansible/bundled/) that
always appears alongside operator-configured sources, with two playbooks:
- maintenance/docker_prune.yml — docker system prune for swarm/standalone
nodes (safe by default; prune_all_images / prune_volumes extra-vars to
widen). Schedule it against a swarm-node group for recurring cleanup (#869).
- host_agent/install.yml — installs/updates the host agent (mirrors
install.sh: user, agent.py, config, hardened systemd unit), parameterised
with steward_url + steward_token extra-vars.
get_sources() now prepends the builtin source. Tests updated to find the
configured git source by name; added coverage that the bundled playbooks are
discoverable.
Tasks #869 + agent-install (milestone #37).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds cron-like recurring runs (the engine the maintenance-automation work
needs). New AnsibleSchedule model + migration 0018; a core ScheduledTask
(ansible_scheduled_runs, 60s) fires due schedules, each creating a
system-triggered AnsibleRun (triggered_by=None). Centralises the
resolve-inventory → create-run → launch flow in ansible/runner.trigger_run,
shared by the manual route (refactored to use it) and the scheduler.
Schedules UI under /ansible/schedules: create/edit/pause/delete/run-now,
with interval presets, scope targeting (all / group / target), extra-vars /
limit / tags / dry-run, and last-run status (resolved via last_run_id) +
next-run. Unit test for the due-check.
Task #549 (milestone #37).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Surfaces the metrics the agent now collects (task #868). Adds a viewer-facing
fleet page (/plugins/host_agent/) with per-host cards (CPU/mem/disk/load/temp,
stale flag) and a per-host detail page (/plugins/host_agent/<id>/) — the link
the fleet widget already pointed at but had no route for.
Detail page shows current gauges (CPU, memory incl. swap/cache, load, network
rx/tx, disk I/O, temp max, memory/cpu/io PSI), per-core CPU bars, per-mount
filesystem bars, and per-interface/disk/sensor breakdowns, plus history charts
(utilization %, throughput B/s, load & pressure) over a selectable range.
Charts use a linear epoch-ms x-axis so no Chart.js date adapter is needed.
Stale state uses the plugin's stale_after_seconds threshold. Repoints the
host_resources widget detail link from admin settings to the new fleet page.
Completes milestone #68 (task #867).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
collect_cpu now returns (aggregate, [per_core]) and build_sample takes a
rate-state dict, so the pre-existing tests that asserted the old float
return / no-state signature needed updating. Mocks the new collectors
(temps/psi/net/diskio) for determinism.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extends the push agent (v1.2.0) with the Netdata-style signals the operator
wants: per-core CPU, per-interface network throughput and per-disk I/O
(rates derived in-agent from monotonic /proc counter deltas, with
counter-reset clamping), hardware temperatures (/sys/class/hwmon),
memory-pressure PSI (/proc/pressure), and cached/buffers memory breakdown.
All stdlib-only.
Server side needs no migration — these land as additional rows in the shared
PluginMetric table via _expand_sample_to_metrics. Per-resource series use a
':' in resource_name (host:net:eth0, host:core0, host:temp:Package) so the
existing fleet widget's ':' filter ignores them; host-level totals/max are
emitted at the bare host resource. New sample keys are optional, so older
agents keep ingesting unchanged.
Unit tests for the new parsers, rate/reset logic, and the expander contract.
Data foundation for the Netdata-style host view (task #867).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds a Kuma-style "is everything up?" surface that aggregates heterogeneous
monitor types via a status-source registry (steward/core/status.py): each
type registers an async source(db) -> [StatusEntry]. Core registers ping/DNS;
the http plugin registers its own from setup() so core never imports plugin
tables. Per entry: current up/down, last-30 heartbeat bar, uptime %
(24h/7d/30d), latest latency + response sparkline, and TLS expiry countdown
(HTTP). New /status page (live htmx refresh) + a status_overview dashboard
widget + nav link. Pure-function unit tests for registry + sparkline.
First deliverable of milestone #68 (task #866).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
create_app is synchronous (returns Quart) and runs its own internal
asyncio.run for settings/migrations. _make_app wrapped it in asyncio.run,
which nested event loops and raised "a coroutine is required" — the other
integration tests already call create_app() directly. This test was added
in an unpushed commit, so CI never exercised it until now.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A fresh install enabled zero plugins — settings.py DEFAULTS had no plugin.*
keys, so to_plugins_cfg returned {} and every plugin had to be flipped on
by hand. Seed docker, host_agent, http, snmp as default-on (generic, non-
vendor-specific); traefik and unifi stay opt-in. Stored choices override
the default, so disabling persists.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The app auto-generates a key on first boot and persists it to /data/secret.key,
which the app_data volume keeps across image updates. No reason to surface this
as a user-facing env var in the deploy compose. Removed from .env.example too.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Plain ${POSTGRES_PASSWORD} in compose; default 'steward' lives in .env.example
where it's visible and editable rather than buried in compose YAML.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add the rule-46 image-publish lane: a `publish` job gated on
[lint, unit, integration] that builds the runtime Dockerfile in ci-builder
and pushes git.fabledsword.com/bvandeusen/steward:<sha> always, :dev on dev
and :latest on main. Authenticates with the REGISTRY_TOKEN repo secret
(the injected GITHUB_TOKEN lacks write:package). Steward is the first family
app to implement the publish lane; mechanics mirror CI-runner's
build-ci-python.yml.
Add compose.deploy.yml for a persistent remote instance that pulls :dev
(no source bind-mounts, persistent app_data+pgdata volumes, bundled Postgres),
plus a README "Remote dev instance" runbook and the POSTGRES_PASSWORD env key.
Fix .env.example's database URL var to the canonical STEWARD_DATABASE_URL
(single underscore) and document the publish lane in ci-requirements.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
From a host's edit page, an operator can run a playbook against just that host
via an ephemeral one-host inventory — no need for the host to exist in a source
inventory.
- ansible/sources.py: pure host_inventory_content(host) -> '<name> ansible_host=<addr>'
- executor.start_run: optional inventory_content written to the temp dir and used
as -i (overrides inventory_path); cleaned up with the creds dir
- hosts/routes.py: POST /hosts/<id>/run-playbook (operator) — validates source +
playbook, builds the ephemeral inventory, starts a user-triggered run, redirects
to the live run detail; edit page gets the ansible source list
- hosts/form.html: 'Run Ansible playbook against this host' panel (shown when
editing an existing host and sources exist) — source + playbook + extra-vars/
tags/dry-run (no --limit; single host)
- tests: unit host_inventory_content; integration runs a hosts:all/connection:local
playbook against the ephemeral inventory and asserts inventory_hostname
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Global Ansible credentials, applied to every run (manual + alert-triggered):
- core/settings.py: ansible.ssh_private_key / become_password / vault_password
(plaintext at rest, masked in UI — encryption tracked in #580) + host_key_checking
(default off); surfaced via to_ansible_cfg into app.config[ANSIBLE]
- executor.py: pure build_credentials() materializes creds into a 0600 temp dir
(--private-key, --vault-password-file, become via -e @vars-file so the password
never hits argv) cleaned up in finally; pure ansible_env() sets
ANSIBLE_HOST_KEY_CHECKING. build_ansible_command stays param-only
- settings/routes.py + ansible.html: admin-only Credentials section, masked-update
(blank keeps current, explicit Clear checkbox), reload app config on save
- tests: unit (build_credentials per cred + none; ansible_env toggle); integration
vault round-trip (ansible-vault encrypt a vars file, run via executor with the
vault password, assert it decrypted)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Dev convenience so you can author playbooks on the host and have Steward's
local-source discovery pick them up. ./ansible-dev/ is gitignored (operator
content); a safe demo playbook lives there locally.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The alertoperator Postgres enum (migration 0002) was created with the operator
VALUES as labels ('>', '<', …), but the model's Enum(AlertOperator) defaulted to
persisting enum-member NAMES ('gt', 'lt', …) — so inserting any AlertRule against
real Postgres raised 'invalid input value for enum alertoperator: "gt"'. Never
caught because unit tests mock the DB and no integration test inserted a rule
until task 250's. Add values_callable so the ORM round-trips the values.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rebuilds the deleted NUT/UPS automation as a general alert action: any metric
can drive a playbook run on transition-to-firing.
- models/alerts.py + migration 0014: AlertRule.ansible_action (JSON, admin-only,
reuses the #546 param shape); AlertEvent.ansible_run_id links a firing event to
the run it triggered
- core/alerts.py: pure alert_extra_vars() injects steward_alert_* context; on
('firing', event) with an action set, schedule _run_ansible_action (deferred
after commit, same pattern as notifications) — fires once per transition,
consecutive_failures_required is the debounce; system-triggered AnsibleRun
- alerts/routes.py: admin-only parse/validate of the action (source must be
configured, playbook must exist); operators keep editing rules, action preserved
- rules_form.html: admin-only 'On firing -> run a playbook' section
- tests: unit for alert_extra_vars; integration drives record_metric to firing
and asserts a system AnsibleRun ran the playbook with the injected var and the
event linked to it
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Run form + executor now support optional params, passed as argv (never
shell-interpolated):
- extra_vars: one key=value per line -> repeated -e
- limit -> --limit; tags -> --tags
- dry-run checkbox -> --check --diff
- executor.py: pure build_ansible_command(playbook, inventory, params); start_run
gains an optional params arg (backward-compatible)
- models/ansible.py + migration 0013: nullable params JSON column
- routes.py: create_run parses + validates (extra-var lines need '='), stores
params on the run, passes to the executor
- browse.html run form: Extra vars / Limit / Tags / Dry-run fields
- run_detail.html: shows the params used
- tests/test_ansible_command.py: unit coverage of the builder; integration test
runs a real playbook with extra-var + limit + check and asserts substitution
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The id 0012_ansible_run_triggered_by_nullable was 38 chars; alembic_version.
version_num is VARCHAR(32), so stamping it raised StringDataRightTruncationError
and the boot/migrate integration lane failed. Renamed to 0012_ansible_run_nullable.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Foundation for the Ansible automation milestone (#37) — makes the existing
manual playbook runner actually executable and the schema automation-ready.
- pyproject: [ansible] extra (full ansible package, batteries-included, pinned)
- Dockerfile: pip install .[ansible]; add openssh-client for remote runs
- models/ansible.py + migration 0012: AnsibleRun.triggered_by now nullable so
automated (alert/schedule) runs need no human actor
- ansible/routes.py + run_detail.html: show 'Triggered by' (username or 'system')
- CI: integration lane installs .[dev,ansible]; new tests/integration/
test_ansible_foundation.py runs a real connection:local playbook end-to-end,
asserts success+output, and round-trips a NULL-triggered run
No extra-vars/limit/credentials/scheduling here — those are their own #37 tasks.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
.dockerignore excluded plugins/ (leftover from the separate-repo era), so the
Dockerfile's COPY plugins/ found nothing and the build failed. Track it now that
first-party plugins ship in the image.
docker-compose.yml reworked for local dev: live-mounted ./steward + ./plugins
with PYTHONPATH=/app, --debug auto-reload, fixed dev SECRET_KEY, Postgres port
exposed, and the host-specific traefik/docker.sock mounts commented out.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The on-disk-version test hardcoded '1.0.0' and broke on the 1.1.0 bump. Compare
against agent.AGENT_VERSION so future bumps don't require touching the test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The agent now detects its own primary non-loopback IP (UDP-socket trick, no
packets sent) and sends it in the metadata bag; bumps AGENT_VERSION 1.0.0->1.1.0
and re-detects on SIGHUP. The server persists it on HostAgentRegistration.host_ip
and mirrors it into Host.address ONLY when that field is blank — an admin-typed
address is never overwritten. The reported IP shows in the settings table so
admins can see drift regardless.
- agent.py: detect_primary_ip(); host_ip in collect_metadata(); refresh on reload
- routes.py: pure pick_host_address() helper; ingest persists host_ip + mirrors
- models.py + migration host_agent_002_host_ip: new String(45) column
- settings_list.html: Reported IP column
- plugin.yaml: 1.0.0 -> 1.1.0
- tests: pick_host_address (fill-when-blank / never-clobber / no-op),
detect_primary_ip (never raises, non-loopback), metadata carries host_ip
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The fabledscryer->steward rename had only ever reached host_agent. The other
five bundled plugins (http, snmp, traefik, unifi, docker) still imported
`from fabledscryer.*` (package no longer exists) and read FABLEDSCRYER_* env
vars — so every one of them was broken at import since the original rebrand.
CI stayed green only because none are enabled by default and migrations don't
import plugin modules. Now that they version in-tree, complete the rename:
- fabledscryer.* -> steward.* imports across all five plugins
- FABLEDSCRYER_* -> STEWARD_* in plugin migration env.py files
- author/repository/homepage + user-facing 'Fabled Scryer' strings -> Steward
- snmp/scheduler.py: also drop dead `now`/datetime; record_metric from steward
Adds tests/test_no_legacy_names.py — fails if 'scryer'/'roundtable' ever
reappear in shipped code (the drift bit twice; this stops a third time).
Also clears pre-existing ruff lint debt (unused imports, semicolon statements,
mid-file import) surfaced by the new lint lane.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Single-repo CI now that plugins are bundled in-tree. Three lanes on push to
dev/main, modeled on FabledCurator's canonical ci.yml (minus frontend/Redis):
- lint: ruff check steward/ plugins/ tests/ (no dep install)
- unit: pytest -m 'not integration' — whole current suite (testing=True mocks DB)
- integration: postgres:16-alpine service via socket-discovered bridge IP; a
boot-and-migrate test creates the real app, running core + all bundled-plugin
migrations, then round-trips the DB — guards the folded-in migration graph.
Registers the 'integration' pytest marker; adds tests/integration/test_boot.py.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First-party plugins (host_agent, http, snmp, traefik, unifi, docker) are now
tracked under plugins/ and baked into the image, so they version atomically
with core — ending the cross-repo import drift the roundtable->steward rename
exposed. History for these files is preserved in the archived Roundtable-plugins
repo.
Plugin discovery becomes multi-root: PLUGIN_DIR (single) -> PLUGIN_DIRS
(bundled first, then external) + PLUGIN_INSTALL_DIR. Bundled ships in the image;
third-party plugins still mount at runtime into the external root
(STEWARD_PLUGIN_DIR, default /data/plugins) and downloads/installs land there.
Bundled shadows external on a name collision.
- config.py: load_bootstrap returns plugin_dirs + plugin_install_dir
- app.py: iterate PLUGIN_DIRS at the migration + load sites
- migration_runner.py: discover_all_in() unions every plugin root
- plugin_manager.py: resolve_plugin_path() (pure, first-root-wins); load /
install / hot-reload span all roots; installs target the external root
- settings/routes.py: _discover_plugins scans all roots, dedup bundled-first
- Dockerfile: COPY plugins/ ; docker-compose: drop host bind, document external
- tests/test_plugin_dirs.py: resolution, multi-root discovery, bootstrap split
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Renames the Python package directory, CLI command, env var prefix,
docker-compose service/container/image, Postgres role/db, and all
visible branding. Marketing form is "Fabled Steward".
Clean break from the previous rebrand: drops the fabledscryer→roundtable
import shim in __init__.py and the FABLEDSCRYER_* env var fallback in
config.py and migrations/env.py. Env vars are now STEWARD_* only.
Heads-up for existing deployments:
- Postgres user/db renamed fabledscryer → steward in docker-compose.yml.
Existing volumes need the role/db renamed inside Postgres, or override
POSTGRES_USER/POSTGRES_DB to keep the old names.
- Host-agent systemd unit is now steward-agent.service. Existing agents
keep running under the old name; reinstall to switch.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add read_config/ConfigError unit tests covering flat key-value parsing,
blank lines/comments, list fields, defaults, missing required keys, and
malformed lines. Also add tests/__init__.py to fix pytest namespace
collision with tests/plugins/ shadowing the root plugins/ package.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Records the mid-execution pivot: test harness uses testing=True which
mocks the DB, so plugin tests stick to pure-Python unit tests (agent +
metric expander). Server routes are verified manually.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
