test(login): stop the backoff-guard test racing its own one-second lock

A single recordFailure() locks for BASE_MS * 2 ** 0 — exactly one second — and
the test then does a real HTTP round trip against it. On CI that round trip took
1,456 ms and the guard correctly answered 200, failing the run for a reason that
has nothing to do with what the test is about.

Five failures lock for sixteen seconds. The subject is the guard's answer while
locked out, which is unchanged.

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

.env.example

@@ -113,6 +113,20 @@ BOT_INTERNAL_KEY=change-me-to-a-long-random-string
 # knows nothing about any of them. A module may read its own env vars, and they
 # belong here because Compose passes this file to the container.
 #
+# MODULES declares the set this deployment runs, and the container arrives at it
+# on its own — no admin panel, no `tar -xf` on the host. One entry per module,
+# `<id>@<version>=<install manifest URL>`, whitespace- or comma-separated:
+#
+#   MODULES=uo@0.3.0=https://gitea.whitlocktech.com/RunicGateway/Module-uo/releases/download/v0.3.0/module-uo-0.3.0.json
+#
+# A module already unpacked at the declared version is a no-op that never touches
+# the network, so a restart with the internet down brings the site up exactly as
+# it was; only a missing or different version is fetched, verified against the
+# sha256 its manifest declares, and unpacked. A failure is logged and shown in
+# Admin → Modules, and the site starts anyway. The variable owns what is on the
+# volume, not what runs — a module disabled from the admin panel stays disabled.
+# Leave it unset to install from the admin panel instead.
+#
 # RunicGateway/Module-uo, for example, reads UOLINK_BASE_URL / UOLINK_WS_URL /
 # UOLINK_PROTOCOL as the defaults for its connection to a uo-link sidecar, and
 # TOWNCRIER_DURATION_SEC for its news leg. Its README documents them; they are

README.md

@@ -489,6 +489,32 @@ modules/
 supported install. The directory is tracked in git (via its README) on purpose: Docker recreates a
 *missing* bind-mount source as `root:root`, and the container is uid 1000.
 
+### Three ways in, and none of them is a build
+
+| | How | Where it fits |
+|---|---|---|
+| **Admin panel** | Admin → Modules, paste the URL of a release's install manifest | The click path. Installs, upgrades, disables, uninstalls and purges, with a restart button — no shell on the box |
+| **`MODULES`** | Declare the set in the environment; the container resolves it at every start | The compose-managed host. The running set is a line in a file you version-control, not the residue of past clicks |
+| **By hand** | `tar -xf module-uo-0.3.0.tar.gz -C ./modules && mv modules/module-uo-0.3.0 modules/uo`, then restart | Development, and any host where the other two do not fit |
+
+`MODULES` takes one entry per module, whitespace- or comma-separated:
+
+```
+MODULES=uo@0.3.0=https://gitea.whitlocktech.com/RunicGateway/Module-uo/releases/download/v0.3.0/module-uo-0.3.0.json
+```
+
+The id and the version are written out rather than discovered inside the manifest so that **the
+no-op case needs no network**: a module already unpacked at the declared version is answered by
+reading its own `module.json`, so a restart with the internet down brings the site up exactly as it
+was. Only a missing or different version is fetched, and it goes through the same
+verify-and-unpack path — allowlisted `https` host, sha256 from the manifest, whole-archive
+inspection before anything is written — that the admin panel uses. A version that cannot be
+resolved is logged and shown on the admin screen; **it never stops the site from starting**.
+
+The declaration owns what is *on the volume*, never what runs. A module disabled from the admin
+panel gets its files back at the next start and stays disabled, because the row and the variable are
+answering different questions.
+
 ### What a module gets, and what it may not do
 
 At boot, `app.js` scans the volume synchronously, validates each `module.json`, and calls the
@@ -534,6 +560,8 @@ Copy `.env.example` (Compose) or `server/.env.example` (local) and fill in. **`.
 | `PORT` | `3000` | server listens on `0.0.0.0:PORT` |
 | `UPLOAD_DIR` | `<server>/uploads` | where post images are written (`/app/uploads`, volume-mounted, in Compose) |
 | `MODULES_DIR` | `<repo>/modules` | where installed modules are scanned from (`/app/modules`, bind-mounted, in Compose) |
+| `MODULES` | — | the module set this deployment runs, resolved at every start: `<id>@<version>=<install manifest URL>`, whitespace/comma separated. Already at the declared version = no network. A failure is logged and shown in Admin → Modules, never fatal. See [Modules](#modules) |
+| `MODULE_SOURCE_HOSTS` | `gitea.whitlocktech.com` | **bootstrap only** — seeds the `module_source_hosts` setting on first boot; after that the setting is authoritative and is edited in Admin → Modules |
 | `DB_HOST` / `DB_PORT` | `db` / `3306` | `db` in Compose; `127.0.0.1` for local dev |
 | `DB_NAME` / `DB_USER` / `DB_PASSWORD` | `runic_gateway` / `runic` / — | app database credentials |
 | `DB_ROOT_PASSWORD` | — | MariaDB root (Compose only) |

client/src/App.jsx

@@ -41,6 +41,7 @@ import AuthProvidersAdmin from './routes/admin/views/AuthProvidersAdmin.jsx'
 import UsersAdmin from './routes/admin/views/UsersAdmin.jsx'
 import UserDetail from './routes/admin/views/UserDetail.jsx'
 import InvitesAdmin from './routes/admin/views/InvitesAdmin.jsx'
+import ModulesAdmin from './routes/admin/views/ModulesAdmin.jsx'
 import AccountAdmin from './routes/admin/views/AccountAdmin.jsx'
 import Moderation from './routes/admin/views/Moderation.jsx'
 import ModerationUser from './routes/admin/views/ModerationUser.jsx'
@@ -169,6 +170,10 @@ export default function App() {
               <Route path="users" element={<UsersAdmin />} />
               <Route path="users/:id" element={<UserDetail />} />
               <Route path="invites" element={<InvitesAdmin />} />
+              {/* Core's own screen, and it has to be: it is how a module reaches
+                  the volume in the first place. Declared here with the rest of
+                  core's routes, above the module-supplied ones below. */}
+              <Route path="modules" element={<ModulesAdmin />} />
               <Route path="account" element={<AccountAdmin />} />
               {/* Installed modules' admin pages, at /admin/<id>/…, already inside
                   RequireAuth + AdminLayout. A module cannot supply its own auth

client/src/api/client.js

@@ -233,6 +233,20 @@ export const api = {
       req('/admin/invites', { method: 'POST', body: { email, role, sendEmail } }),
     revokeInvite: (id) => req(`/admin/invites/${id}`, { method: 'DELETE' }),
 
+    // Installed modules (MODULE_SYSTEM.md §2.7.2). `uninstallModule`'s purge flag
+    // is a query parameter rather than a body because it hangs off a DELETE, and
+    // it is spelled out at the call site rather than defaulted, so the
+    // destructive branch is never the one you get by forgetting an argument.
+    listModules: () => req('/admin/modules'),
+    installModule: (url) => req('/admin/modules', { method: 'POST', body: { url } }),
+    enableModule: (id) => req(`/admin/modules/${encodeURIComponent(id)}/enable`, { method: 'POST' }),
+    disableModule: (id) => req(`/admin/modules/${encodeURIComponent(id)}/disable`, { method: 'POST' }),
+    uninstallModule: (id, { purge } = {}) =>
+      req(`/admin/modules/${encodeURIComponent(id)}${purge ? '?purge=true' : ''}`, { method: 'DELETE' }),
+    purgeModule: (id) => req(`/admin/modules/${encodeURIComponent(id)}/purge`, { method: 'POST' }),
+    setModuleSources: (hosts) => req('/admin/modules/sources', { method: 'PUT', body: { hosts } }),
+    restartServer: () => req('/admin/modules/restart', { method: 'POST' }),
+
     // ----- moderation dashboard (admin + moderator) -----
     modSummary: () => req('/admin/moderation/stats/summary'),
     modRecent: (params = {}) => {

client/src/lib/moduleAdmin.js

@@ -0,0 +1,252 @@
+// What an admin should be told about one installed module, and what they may do
+// to it — derived, not spelled out at each button.
+//
+// Phase 4, slice 2 of docs/website/MODULE_SYSTEM.md §2.7.2. Plain JS rather than
+// a hook or a chunk of JSX, for the same reason `lib/adminNav.js` is: the test
+// runner here has no DOM, and this is the part of the Modules screen that is
+// actually worth testing.
+//
+// **The screen has four sources of truth and they are allowed to disagree**
+// (MODULE_SYSTEM.md §2.4, and slice 3 for the fourth):
+//
+//   state      what the DATABASE row records — what the operator decided, and
+//              what the last boot ended up doing
+//   liveState  what the LOADER has mounted in this process and is answering with
+//   onVolume   whether there is still a directory there at all
+//   declared   what this container's MODULES variable asks for — the only one of
+//              the four that no button on this screen can change
+//
+// Picking one and rendering it would be simpler and would lie. The case that
+// makes this concrete is the one decision 3 creates on purpose: an operator
+// disables a module (its onShutdown runs, its routes 404) and then enables it
+// again. The row says `enabled`; the loader still says `disabled`, because
+// there is no `onBoot` re-dispatch and nothing can start it before a restart.
+// It is neither running nor off, and the honest thing to show is "enabled —
+// restart to start it".
+
+/**
+ * The one-line status of a module, and whether that status is waiting on a
+ * restart.
+ *
+ * Ordering matters here. The checks run most-alarming first, so a module whose
+ * directory has been deleted is described that way rather than by whatever its
+ * row happens to still say.
+ *
+ * @param {object} m a row from GET /admin/modules
+ * @returns {{ label: string, tone: 'ok'|'warn'|'bad'|'idle', pending: boolean, detail: string }}
+ */
+export function statusOf(m) {
+  // Declared by the environment and not there at all: no row, no directory,
+  // nothing mounted. Every other branch below reads one of those three, so
+  // without this the screen would describe a module it has never had as though
+  // a row had gone stale — and the one thing the operator needs, the reason
+  // resolution failed, would be nowhere.
+  if (m.declared && !m.onVolume && m.state === null) {
+    return {
+      label: 'Declared, not installed',
+      tone: 'bad',
+      pending: false,
+      detail: m.declaredError
+        ? `MODULES asks for v${m.declaredVersion}; the last start could not install it: ${m.declaredError}`
+        : `MODULES asks for v${m.declaredVersion}. It will be installed when the server next starts.`,
+    }
+  }
+
+  // Gone from the volume, but still known. Either a hand-deleted directory (the
+  // boot reconcile marks that `startup_failed`) or an uninstall waiting for its
+  // restart. Both are "there is nothing to run here".
+  if (!m.onVolume) {
+    return {
+      label: m.state === 'disabled' ? 'Uninstalled' : 'Missing from the volume',
+      tone: m.state === 'disabled' ? 'idle' : 'bad',
+      pending: m.liveState !== null,
+      detail: m.state === 'disabled'
+        ? 'The files are gone. Its data was kept, and reinstalling brings it back.'
+        : 'A row exists but there is no module directory. Reinstall it, or uninstall to clear the row.',
+    }
+  }
+
+  // **Installed since this process booted**, and this check has to come before
+  // the failure one. `liveState` is the loader's record, and the loader scans
+  // the volume once at require time — so a module that is on the volume NOW and
+  // has no live record was put there after the scan. Anything the row still says
+  // about it therefore predates the install and is stale by definition.
+  //
+  // Found by the §7.7 browser smoke, and no unit test here had modelled it:
+  // installing over a row left `startup_failed` by the previous boot rendered
+  // "Failed at the require stage: module directory not present on the volume"
+  // one second after the file had been written to the volume — and, because that
+  // branch is not pending, suppressed the restart banner the install had just
+  // told the operator to use.
+  if (m.liveState === null) {
+    return {
+      label: 'Restart to start',
+      tone: 'warn',
+      pending: true,
+      detail: 'Installed. It mounts when the server next starts.',
+    }
+  }
+
+  if (m.state === 'startup_failed' || m.liveState === 'startup_failed') {
+    return {
+      label: 'Failed to start',
+      tone: 'bad',
+      pending: false,
+      detail: m.failureReason
+        ? `Failed at the ${m.failureStage || 'unknown'} stage: ${m.failureReason}`
+        : 'It failed to start and recorded no reason.',
+    }
+  }
+
+  if (m.state === 'disabled') {
+    return {
+      label: 'Disabled',
+      tone: 'idle',
+      pending: false,
+      detail: 'Stopped and switched off. Its routes answer 404 and it stays off across restarts.',
+    }
+  }
+
+  // The row has been switched on but the loader has not started it — the
+  // decision-3 case: disable ran its onShutdown, and nothing can start it again
+  // before a restart.
+  if (m.liveState !== 'started') {
+    return {
+      label: 'Restart to start',
+      tone: 'warn',
+      pending: true,
+      detail: m.liveState === 'disabled'
+        ? 'Enabled, but still stopped in the running server — it cannot be restarted in place.'
+        : 'Enabled. It mounts when the server next starts.',
+    }
+  }
+
+  // Running, but not the version that is installed. An upgrade writes new files
+  // and a new row while the old code stays loaded, so the row's `version` is a
+  // promise about the next boot rather than a description of this one — and
+  // "Running v2.0.0" beside a process serving v1.0.0 is the same lie as the
+  // stale-failure one above, in a different place.
+  if (m.liveVersion && m.liveVersion !== m.version) {
+    return {
+      label: 'Restart to finish upgrading',
+      tone: 'warn',
+      pending: true,
+      detail: `v${m.version} is installed; v${m.liveVersion} is still running.`,
+    }
+  }
+
+  return {
+    label: 'Running',
+    tone: 'ok',
+    pending: false,
+    detail: 'Mounted and serving.',
+  }
+}
+
+/**
+ * What the environment's declaration means for this module, as one sentence — or
+ * null if nothing declares it.
+ *
+ * Kept out of `statusOf` on purpose. A module can be running perfectly while its
+ * declared upgrade is failing, and collapsing both into one label would have to
+ * pick which of the two is "the" status. This is a second line, beside the first.
+ *
+ * The sentence an operator most needs is the uninstall one: MODULES owns what is
+ * on the volume and the row owns whether it runs, so uninstalling a declared
+ * module puts its files back at the next start and leaves it switched off. Files
+ * reappearing unexplained is exactly the kind of thing that gets debugged for an
+ * afternoon.
+ *
+ * @param {object} m a row from GET /admin/modules
+ * @returns {{ text: string, tone: 'warn'|'idle' }|null}
+ */
+export function declarationNoteFor(m) {
+  if (!m.declared) return null
+
+  if (m.declaredError) {
+    return {
+      text: `MODULES asks for v${m.declaredVersion} and the last start could not install it: ${m.declaredError}`,
+      tone: 'warn',
+    }
+  }
+  if (!m.onVolume) {
+    return {
+      text:
+        `MODULES declares v${m.declaredVersion}, so its files come back when the server next starts`
+        + (m.state === 'disabled' ? ' — switched off, until you enable it.' : '.'),
+      tone: 'warn',
+    }
+  }
+  return { text: `Declared by this deployment's MODULES variable at v${m.declaredVersion}.`, tone: 'idle' }
+}
+
+/**
+ * Which actions are offered for a module, and why the others are not.
+ *
+ * Returned as a map of `{ shown, reason }` rather than a list of shown actions,
+ * so a disabled button can say what would make it available. Every rule here
+ * mirrors one the server enforces — this is presentation, never the boundary.
+ *
+ * @param {object} m a row from GET /admin/modules
+ */
+export function actionsFor(m) {
+  const running = m.liveState === 'started'
+  const disabled = m.state === 'disabled'
+
+  return {
+    // Only offered while something is actually running: disabling a module that
+    // is already stopped has nothing to stop and no guard to flip.
+    disable: {
+      shown: !disabled && m.onVolume,
+      reason: disabled ? 'Already disabled.' : 'Nothing is running to stop.',
+    },
+    enable: {
+      shown: disabled && m.onVolume,
+      reason: 'Only a disabled module can be enabled.',
+    },
+    uninstall: {
+      shown: m.onVolume,
+      reason: 'There are no files left to remove.',
+    },
+    // The server refuses a standalone purge unless the module is disabled, so
+    // the button says so rather than offering a click that 409s.
+    purge: {
+      shown: m.onVolume && m.canPurge,
+      enabled: disabled,
+      reason: !m.canPurge
+        ? 'This module ships no purge.sql, so its data cannot be deleted.'
+        : 'Disable it first, so nothing is serving out of the tables being dropped.',
+    },
+    // A row with no directory is the one thing an uninstall cannot tidy through
+    // the normal path — offer clearing it instead.
+    forget: {
+      shown: !m.onVolume && m.state !== null,
+      reason: 'The module is still installed.',
+    },
+    running,
+  }
+}
+
+/**
+ * Does anything on this list need a restart before it matches what is running?
+ *
+ * Drives the one banner at the top of the screen rather than a badge per row:
+ * the restart is a property of the SERVER, not of a module, and offering it
+ * five times would suggest otherwise.
+ */
+export const needsRestart = (modules) => modules.some((m) => statusOf(m).pending)
+
+/**
+ * Split a hosts string the way the server will.
+ *
+ * Duplicated from `install.parseHosts` deliberately — it is four lines, and the
+ * alternative is an API round trip to preview what the field is going to mean.
+ * The server remains the one that decides; this only shows the operator how
+ * their typing will be read.
+ */
+export function parseHosts(value) {
+  return String(value || '')
+    .split(/[,\s]+/)
+    .map((h) => h.trim().toLowerCase())
+    .filter(Boolean)
+}

client/src/routes/admin/AdminLayout.jsx

@@ -45,6 +45,7 @@ const IconPulse = () => <Icon><path d="M3 12h3l2 6 4-14 2 8h7" /></Icon>
 const IconUser = () => <Icon><circle cx="12" cy="8" r="4" /><path d="M4 21a8 8 0 0 1 16 0" /></Icon>
 const IconNav = () => <Icon><path d="M4 6h16M4 12h16M4 18h10" /><circle cx="18" cy="18" r="2.5" /></Icon>
 const IconPalette = () => <Icon><path d="M12 3a9 9 0 1 0 0 18 2 2 0 0 0 1.6-3.2 2 2 0 0 1 1.6-3.2H18a3 3 0 0 0 3-3 9 9 0 0 0-9-8.6z" /><circle cx="7.5" cy="11.5" r="1" /><circle cx="10.5" cy="7.5" r="1" /><circle cx="15" cy="8.5" r="1" /></Icon>
+const IconModules = () => <Icon><path d="M12 3l8 4.5-8 4.5-8-4.5z" /><path d="M4 12l8 4.5 8-4.5" /><path d="M4 16.5L12 21l8-4.5" /></Icon>
 
 // Nav is grouped into collapsible categories. A group with no `title` renders
 // its items ungrouped (Dashboard at top, Account at bottom). Each item's `roles`
@@ -83,6 +84,10 @@ export const NAV = [
       { to: '/admin/users', label: 'Users', icon: IconUsers, roles: ['admin'] },
       { to: '/admin/invites', label: 'Invites', icon: IconUsers, roles: ['admin'] },
       { to: '/admin/settings', label: 'Settings', icon: IconGear, roles: ['admin'] },
+      // Admin-only, matching the server: every route under /admin/modules
+      // re-gates to `admin` on top of the group's staff gate, because installing
+      // a module runs its code in this process.
+      { to: '/admin/modules', label: 'Modules', icon: IconModules, roles: ['admin'] },
       { to: '/admin/appearance', label: 'Appearance', icon: IconPalette, roles: ['admin'] },
       { to: '/admin/navigation', label: 'Navigation', icon: IconNav, roles: ['admin'] },
       { to: '/admin/hero', label: 'Hero Editor', icon: IconHero, roles: ['admin'] },

client/src/routes/admin/views/ModulesAdmin.jsx

@@ -0,0 +1,424 @@
+import { useCallback, useEffect, useState } from 'react'
+import { Loading, ErrorState } from '../../../components/PageState.jsx'
+import { dateTime } from '../../../lib/format.js'
+import { statusOf, actionsFor, declarationNoteFor, needsRestart, parseHosts } from '../../../lib/moduleAdmin.js'
+import { api } from '../../../api/client.js'
+
+// Installed modules: install from a release URL, enable, disable, uninstall,
+// purge, and restart the server so the changes take effect.
+//
+// Phase 4, slice 2 of docs/website/MODULE_SYSTEM.md §2.7.2. Everything that
+// decides what a row SAYS and which buttons it offers lives in
+// lib/moduleAdmin.js, which is plain JS and has tests; this file renders it.
+//
+// Two things about this screen are unlike the rest of the admin panel and are
+// deliberate:
+//
+//   1. **Restart is a banner, not a per-row button.** A restart is a property of
+//      the server, not of a module. Offering it on five rows would suggest
+//      otherwise, and an operator who installed three modules should restart
+//      once.
+//   2. **Disable is the only action that takes effect immediately.** Everything
+//      else is "true after the next boot", because the loader reads the volume
+//      at require time (§1.12). The buttons say which they are.
+
+const TONE = {
+  ok: '#7fd0a4',
+  warn: 'var(--accent)',
+  bad: '#d98b84',
+  idle: 'var(--muted)',
+}
+
+const DANGER = { color: '#d98b84', borderColor: '#5b2020' }
+
+function Pill({ tone, children }) {
+  return (
+    <span
+      className="badge"
+      style={{ color: TONE[tone] || 'var(--muted)', borderColor: 'var(--line)', background: 'var(--panel-flat)' }}
+    >
+      {children}
+    </span>
+  )
+}
+
+// ── Install ────────────────────────────────────────────────────────────────
+
+function InstallForm({ sourceHosts, onInstalled }) {
+  const [url, setUrl] = useState('')
+  const [busy, setBusy] = useState(false)
+  const [error, setError] = useState('')
+  const [result, setResult] = useState(null)
+
+  async function submit(e) {
+    e.preventDefault()
+    setError('')
+    setResult(null)
+    if (!url.trim()) return setError('Paste the URL of a release install manifest.')
+    setBusy(true)
+    try {
+      const res = await api.admin.installModule(url.trim())
+      setResult(res)
+      setUrl('')
+      await onInstalled()
+    } catch (err) {
+      // The server's message is written to be read by whoever pasted the URL —
+      // which host was refused, which hash did not match, what the archive
+      // contained. Replacing it with something friendlier would throw away the
+      // only part that helps.
+      setError(err.message || 'Could not install that module.')
+    } finally {
+      setBusy(false)
+    }
+  }
+
+  return (
+    <div className="panel" style={{ padding: 22, marginBottom: 22 }}>
+      <div className="field-label" style={{ marginBottom: 10 }}>Install a module</div>
+      <form onSubmit={submit} style={{ display: 'flex', gap: 12, alignItems: 'flex-end', flexWrap: 'wrap' }}>
+        <label style={{ flex: '1 1 380px' }}>
+          <span className="field-label">Release install-manifest URL</span>
+          <input
+            type="url"
+            value={url}
+            onChange={(e) => setUrl(e.target.value)}
+            className="input"
+            placeholder="https://gitea.example.com/org/Module-uo/releases/download/v0.3.0/module-uo-0.3.0.json"
+          />
+        </label>
+        <button type="submit" disabled={busy} className="btn btn-primary btn-sq">
+          {busy ? 'Installing…' : 'Install'}
+        </button>
+      </form>
+
+      <p className="sans" style={{ margin: '12px 0 0', fontSize: '0.8rem', color: 'var(--muted)' }}>
+        The bundle is downloaded, checked against the <code>sha256</code> its release published, and
+        unpacked onto the modules volume. It starts serving after a restart.{' '}
+        {sourceHosts.length === 0
+          ? 'No source hosts are allowed yet — add one below before installing.'
+          : `Allowed hosts: ${sourceHosts.join(', ')}.`}
+      </p>
+
+      {error && <p className="sans" style={{ margin: '12px 0 0', color: TONE.bad, fontSize: '0.85rem' }}>{error}</p>}
+      {result && (
+        <p className="sans" style={{ margin: '12px 0 0', color: TONE.ok, fontSize: '0.85rem' }}>
+          {result.replaced ? 'Upgraded' : 'Installed'} {result.module?.name} v{result.module?.version}. Restart to load it.
+        </p>
+      )}
+    </div>
+  )
+}
+
+// ── The restart banner ─────────────────────────────────────────────────────
+
+function RestartBanner({ onDone }) {
+  const [busy, setBusy] = useState(false)
+  const [sent, setSent] = useState(false)
+
+  async function restart() {
+    // Said plainly, because it is true and because the failure mode is bad: a
+    // deployment with no supervisor does not come back on its own.
+    const ok = window.confirm(
+      'Restart the server now?\n\n'
+        + 'The site will be briefly unavailable. It comes back on its own only if something is '
+        + 'supervising the process — the shipped Docker Compose file does. If you are running '
+        + '`npm start` by hand, you will have to start it again yourself.',
+    )
+    if (!ok) return
+    setBusy(true)
+    try {
+      await api.admin.restartServer()
+      setSent(true)
+      // Nothing is coming back on this connection: the process is exiting. Give
+      // the supervisor a moment and then reload, which is what the operator was
+      // about to do anyway.
+      setTimeout(() => { if (onDone) onDone() }, 6000)
+    } catch {
+      // A failed request here is expected as often as not — the process can win
+      // the race and drop the socket before the response lands.
+      setSent(true)
+      setTimeout(() => { if (onDone) onDone() }, 6000)
+    } finally {
+      setBusy(false)
+    }
+  }
+
+  return (
+    <div className="panel" style={{ padding: 18, marginBottom: 22, borderColor: 'var(--accent)' }}>
+      <div style={{ display: 'flex', gap: 14, alignItems: 'center', flexWrap: 'wrap' }}>
+        <div style={{ flex: '1 1 320px' }}>
+          <div className="field-label" style={{ marginBottom: 4 }}>Restart needed</div>
+          <p className="sans" style={{ margin: 0, fontSize: '0.84rem', color: 'var(--muted)' }}>
+            {sent
+              ? 'Restarting. This page will reload once the server is back.'
+              : 'Modules are read from disk when the server starts, so an install, an uninstall or a re-enable only takes effect after a restart.'}
+          </p>
+        </div>
+        <button type="button" className="btn btn-primary btn-sq" disabled={busy || sent} onClick={restart}>
+          {sent ? 'Restarting…' : 'Restart the server'}
+        </button>
+      </div>
+    </div>
+  )
+}
+
+// ── The source allowlist ───────────────────────────────────────────────────
+
+function SourceHosts({ hosts, onSaved }) {
+  const [value, setValue] = useState(hosts.join(', '))
+  const [busy, setBusy] = useState(false)
+  const [error, setError] = useState('')
+  const [saved, setSaved] = useState(false)
+
+  useEffect(() => { setValue(hosts.join(', ')) }, [hosts])
+
+  async function save(e) {
+    e.preventDefault()
+    setError('')
+    setSaved(false)
+    setBusy(true)
+    try {
+      await api.admin.setModuleSources(value)
+      setSaved(true)
+      await onSaved()
+    } catch (err) {
+      setError(err.message || 'Could not save the allowlist.')
+    } finally {
+      setBusy(false)
+    }
+  }
+
+  const parsed = parseHosts(value)
+
+  return (
+    <div className="panel" style={{ padding: 22, marginTop: 22 }}>
+      <div className="field-label" style={{ marginBottom: 10 }}>Where modules may be installed from</div>
+      <form onSubmit={save} style={{ display: 'flex', gap: 12, alignItems: 'flex-end', flexWrap: 'wrap' }}>
+        <label style={{ flex: '1 1 380px' }}>
+          <span className="field-label">Allowed hosts</span>
+          <input
+            type="text"
+            value={value}
+            onChange={(e) => setValue(e.target.value)}
+            className="input"
+            placeholder="gitea.example.com, releases.example.org"
+          />
+        </label>
+        <button type="submit" disabled={busy} className="btn btn-sq">{busy ? 'Saving…' : 'Save'}</button>
+      </form>
+
+      <p className="sans" style={{ margin: '12px 0 0', fontSize: '0.8rem', color: 'var(--muted)' }}>
+        Installing a module runs its code inside this server, so only hosts listed here may be
+        installed from — over HTTPS, and re-checked on every redirect. An empty list blocks all
+        installs.{' '}
+        {parsed.length > 0 && <>Will be saved as: <code>{parsed.join(', ')}</code>.</>}
+      </p>
+
+      {error && <p className="sans" style={{ margin: '10px 0 0', color: TONE.bad, fontSize: '0.85rem' }}>{error}</p>}
+      {saved && !error && <p className="sans" style={{ margin: '10px 0 0', color: TONE.ok, fontSize: '0.85rem' }}>Saved.</p>}
+    </div>
+  )
+}
+
+// ── One module ─────────────────────────────────────────────────────────────
+
+function ModuleRow({ m, onChanged, onError }) {
+  const [busy, setBusy] = useState('')
+  const status = statusOf(m)
+  const actions = actionsFor(m)
+  const note = declarationNoteFor(m)
+
+  async function run(name, fn) {
+    setBusy(name)
+    try {
+      await fn()
+      await onChanged()
+    } catch (err) {
+      onError(err.message || `Could not ${name} ${m.id}.`)
+    } finally {
+      setBusy('')
+    }
+  }
+
+  const disable = () => run('disable', () => api.admin.disableModule(m.id))
+  const enable = () => run('enable', () => api.admin.enableModule(m.id))
+
+  function uninstall() {
+    // The purge choice is made HERE and only here, because purge.sql lives
+    // inside the directory the uninstall is about to delete — there is no
+    // "purge it later" (§2.7.2 decision 5). Two prompts rather than one, so
+    // "delete the data too" is never something you agree to by reflex.
+    if (!window.confirm(`Uninstall ${m.name}?\n\nIts files are removed. Its data is kept unless you ask otherwise next.`)) return
+    let purge = false
+    if (m.canPurge) {
+      purge = window.confirm(
+        `Also permanently delete ${m.name}'s data?\n\n`
+          + 'This drops its tables and cannot be undone. This is the only moment it can be offered — '
+          + 'the script that does it is part of the files being removed.\n\n'
+          + 'OK deletes the data. Cancel keeps it.',
+      )
+    }
+    return run('uninstall', () => api.admin.uninstallModule(m.id, { purge }))
+  }
+
+  function purge() {
+    if (!window.confirm(`Permanently delete ${m.name}'s data?\n\nThis drops its tables and cannot be undone.`)) return
+    return run('purge', () => api.admin.purgeModule(m.id))
+  }
+
+  const forget = () => run('forget', () => api.admin.uninstallModule(m.id))
+
+  return (
+    <tr>
+      <td className="adm-td" style={{ color: 'var(--text)' }}>
+        <div style={{ fontWeight: 600 }}>{m.name}</div>
+        <div className="dim" style={{ fontSize: '0.76rem' }}>
+          {/* A declared module that has never installed has no version to show —
+              only the one MODULES asks for, which the status column carries. */}
+          {m.id}{m.version ? ` · v${m.version}` : ''}
+        </div>
+        {m.capabilities?.length > 0 && (
+          <div className="dim" style={{ fontSize: '0.72rem', marginTop: 2 }}>{m.capabilities.join(' · ')}</div>
+        )}
+      </td>
+
+      <td className="adm-td">
+        <Pill tone={status.tone}>{status.label}</Pill>
+        <div className="dim" style={{ fontSize: '0.74rem', marginTop: 4, maxWidth: 380 }}>{status.detail}</div>
+        {/* The environment's declaration, on its own line: a module can be
+            running fine while its declared upgrade is failing, and the status
+            above can only be one of those two things. */}
+        {note && (
+          <div
+            style={{
+              fontSize: '0.74rem',
+              marginTop: 4,
+              maxWidth: 380,
+              color: note.tone === 'warn' ? TONE.warn : 'var(--muted)',
+            }}
+          >
+            {note.text}
+          </div>
+        )}
+      </td>
+
+      <td className="adm-td dim" style={{ fontSize: '0.74rem' }}>
+        {/* Provenance. Null for a directory placed on the volume by hand, which
+            stays a supported install — so it is shown as that, not as missing.
+            A declared module can also reach a boot with no provenance: the
+            no-op path never fetches, so it has no sha256 to record and no
+            reason to write a row. Saying "by hand" there would be the one
+            wrong answer. */}
+        {m.source ? (
+          <>
+            <div style={{ wordBreak: 'break-all', maxWidth: 260 }}>{m.source}</div>
+            {m.sha256 && <div style={{ marginTop: 2 }}>sha256 {m.sha256.slice(0, 12)}…</div>}
+          </>
+        ) : (
+          <span>{m.declared ? 'From the declared module set' : 'Placed on the volume by hand'}</span>
+        )}
+        {m.installedAt && <div style={{ marginTop: 2 }}>{dateTime(m.installedAt)}</div>}
+      </td>
+
+      <td className="adm-td" style={{ textAlign: 'right', whiteSpace: 'nowrap' }}>
+        <div style={{ display: 'inline-flex', gap: 6, flexWrap: 'wrap', justifyContent: 'flex-end' }}>
+          {actions.disable.shown && (
+            <button type="button" className="pill" style={{ fontSize: '0.72rem' }} disabled={Boolean(busy)} onClick={disable}>
+              {busy === 'disable' ? 'Stopping…' : 'Disable'}
+            </button>
+          )}
+          {actions.enable.shown && (
+            <button type="button" className="pill" style={{ fontSize: '0.72rem' }} disabled={Boolean(busy)} onClick={enable}>
+              {busy === 'enable' ? 'Enabling…' : 'Enable'}
+            </button>
+          )}
+          {actions.purge.shown && (
+            <button
+              type="button"
+              className="pill"
+              style={{ fontSize: '0.72rem', ...DANGER, opacity: actions.purge.enabled ? 1 : 0.45 }}
+              disabled={Boolean(busy) || !actions.purge.enabled}
+              title={actions.purge.enabled ? undefined : actions.purge.reason}
+              onClick={purge}
+            >
+              {busy === 'purge' ? 'Purging…' : 'Purge data'}
+            </button>
+          )}
+          {actions.uninstall.shown && (
+            <button type="button" className="pill" style={{ fontSize: '0.72rem', ...DANGER }} disabled={Boolean(busy)} onClick={uninstall}>
+              {busy === 'uninstall' ? 'Removing…' : 'Uninstall'}
+            </button>
+          )}
+          {actions.forget.shown && (
+            <button type="button" className="pill" style={{ fontSize: '0.72rem' }} disabled={Boolean(busy)} onClick={forget}>
+              {busy === 'forget' ? 'Clearing…' : 'Clear the row'}
+            </button>
+          )}
+        </div>
+      </td>
+    </tr>
+  )
+}
+
+// ── The screen ─────────────────────────────────────────────────────────────
+
+export default function ModulesAdmin() {
+  const [data, setData] = useState(null)
+  const [error, setError] = useState('')
+  const [actionError, setActionError] = useState('')
+
+  const load = useCallback(async () => {
+    setError('')
+    try {
+      setData(await api.admin.listModules())
+    } catch {
+      setError('Could not load installed modules.')
+    }
+  }, [])
+  useEffect(() => { load() }, [load])
+
+  if (error) return <ErrorState message={error} />
+  if (!data) return <Loading />
+
+  const modules = data.modules || []
+  const sourceHosts = data.sourceHosts || []
+
+  return (
+    <section>
+      {needsRestart(modules) && <RestartBanner onDone={() => window.location.reload()} />}
+
+      <InstallForm sourceHosts={sourceHosts} onInstalled={load} />
+
+      {actionError && (
+        <p className="sans" style={{ margin: '0 0 14px', color: TONE.bad, fontSize: '0.85rem' }}>{actionError}</p>
+      )}
+
+      <div className="panel-flat">
+        <table className="adm-table">
+          <thead>
+            <tr>
+              <th className="adm-th">Module</th>
+              <th className="adm-th">Status</th>
+              <th className="adm-th">Installed from</th>
+              <th className="adm-th" />
+            </tr>
+          </thead>
+          <tbody>
+            {modules.length === 0 && (
+              <tr>
+                <td className="adm-td" colSpan={4} style={{ color: 'var(--muted)' }}>
+                  No modules installed. Paste a release install-manifest URL above to add one.
+                </td>
+              </tr>
+            )}
+            {modules.map((m) => (
+              <ModuleRow key={m.id} m={m} onChanged={load} onError={setActionError} />
+            ))}
+          </tbody>
+        </table>
+      </div>
+
+      <SourceHosts hosts={sourceHosts} onSaved={load} />
+    </section>
+  )
+}

client/test/apiClient.test.js

@@ -140,3 +140,48 @@ test('DELETE self-service session revoke encodes the id and uses the DELETE meth
   assert.equal(calls[0].opts.method, 'DELETE')
   assert.match(calls[0].url, /\/auth\/me\/sessions\/a%20b%2Fc$/)
 })
+
+// ── admin: installed modules (MODULE_SYSTEM.md §2.7.2) ──────────────────
+//
+// These pin the URLs, because the destructive one differs from the harmless one
+// by a query parameter and nothing else.
+
+test('module actions hit the right paths and methods', async () => {
+  const cases = [
+    [() => api.admin.listModules(), 'GET', '/api/v1/admin/modules'],
+    [() => api.admin.installModule('https://x/y.json'), 'POST', '/api/v1/admin/modules'],
+    [() => api.admin.enableModule('uo'), 'POST', '/api/v1/admin/modules/uo/enable'],
+    [() => api.admin.disableModule('uo'), 'POST', '/api/v1/admin/modules/uo/disable'],
+    [() => api.admin.purgeModule('uo'), 'POST', '/api/v1/admin/modules/uo/purge'],
+    [() => api.admin.setModuleSources('a.com'), 'PUT', '/api/v1/admin/modules/sources'],
+    [() => api.admin.restartServer(), 'POST', '/api/v1/admin/modules/restart'],
+  ]
+  for (const [call, method, url] of cases) {
+    calls = []
+    willReply({ body: {} })
+    await call()
+    assert.equal(calls[0].url, url)
+    assert.equal(calls[0].opts.method || 'GET', method)
+  }
+})
+
+test('uninstall only asks for a purge when it is told to', async () => {
+  // The difference between "remove the module" and "remove the module and drop
+  // every table it owns" is this query parameter, so a default that leaned the
+  // wrong way would be irreversible.
+  willReply({ body: {} })
+  await api.admin.uninstallModule('uo')
+  assert.equal(calls[0].url, '/api/v1/admin/modules/uo')
+  assert.equal(calls[0].opts.method, 'DELETE')
+
+  calls = []
+  willReply({ body: {} })
+  await api.admin.uninstallModule('uo', { purge: true })
+  assert.equal(calls[0].url, '/api/v1/admin/modules/uo?purge=true')
+})
+
+test('a module id is URL-encoded on the way into the path', async () => {
+  willReply({ body: {} })
+  await api.admin.disableModule('a b/c')
+  assert.equal(calls[0].url, '/api/v1/admin/modules/a%20b%2Fc/disable')
+})

client/test/moduleAdmin.test.js

@@ -0,0 +1,281 @@
+import { test } from 'node:test'
+import assert from 'node:assert/strict'
+import { statusOf, actionsFor, declarationNoteFor, needsRestart, parseHosts } from '../src/lib/moduleAdmin.js'
+
+// lib/moduleAdmin.js — what the Modules screen says about a module and what it
+// lets you do to it. Phase 4, slice 2 of MODULE_SYSTEM.md §2.7.2.
+//
+// This is the part of the screen worth testing, and it is plain JS so this
+// runner can reach it (there is no DOM here). What it encodes is §2.4's rule
+// that the row, the loader and the volume are three sources of truth which are
+// ALLOWED to disagree — so most of these cases are combinations that a screen
+// picking one source would render as a lie.
+
+/** A module as GET /admin/modules returns it, with the running case as default. */
+const mod = (over = {}) => ({
+  id: 'uo',
+  name: 'Ultima Online',
+  version: '1.0.0',
+  state: 'started',
+  failureStage: null,
+  failureReason: null,
+  source: 'https://gitea.example.com/x/uo.json',
+  sha256: 'a'.repeat(64),
+  installedAt: null,
+  startedAt: null,
+  liveState: 'started',
+  liveVersion: '1.0.0',
+  capabilities: [],
+  onVolume: true,
+  canPurge: true,
+  declared: false,
+  declaredVersion: null,
+  declaredError: null,
+  ...over,
+})
+
+// ── statusOf ───────────────────────────────────────────────────────────────
+
+test('a mounted, started module is Running and needs nothing', () => {
+  const s = statusOf(mod())
+  assert.equal(s.label, 'Running')
+  assert.equal(s.tone, 'ok')
+  assert.equal(s.pending, false)
+})
+
+test('enabled in the row but disabled in the loader is "Restart to start"', () => {
+  // THE case decision 3 creates on purpose: disable ran the module's onShutdown,
+  // then the operator enabled it again. The row says enabled; nothing can start
+  // it before a restart. Showing either "Running" or "Disabled" would be false.
+  const s = statusOf(mod({ state: 'enabled', liveState: 'disabled' }))
+  assert.equal(s.label, 'Restart to start')
+  assert.equal(s.tone, 'warn')
+  assert.equal(s.pending, true)
+  assert.match(s.detail, /cannot be restarted in place/)
+})
+
+test('freshly installed and never booted into is also "Restart to start"', () => {
+  const s = statusOf(mod({ state: 'installed', liveState: null }))
+  assert.equal(s.label, 'Restart to start')
+  assert.equal(s.pending, true)
+  assert.match(s.detail, /mounts when the server next starts/)
+})
+
+test('a disabled module is Disabled, and that is not pending anything', () => {
+  // Disable takes effect immediately — it is the one action that does — so there
+  // is nothing for a restart banner to be about.
+  const s = statusOf(mod({ state: 'disabled', liveState: 'disabled' }))
+  assert.equal(s.label, 'Disabled')
+  assert.equal(s.pending, false)
+})
+
+test('a fresh install over a failed row is pending, not failed', () => {
+  // THE defect the §7.7 browser smoke found, and one no test here had modelled.
+  // Installing over a row the previous boot left `startup_failed` rendered
+  // "Failed at the require stage: module directory not present on the volume" a
+  // second after the files had been written — and suppressed the restart banner
+  // the install had just told the operator to use.
+  //
+  // `liveState === null` with the module on the volume means the loader's scan
+  // never saw it, so it arrived after boot and everything the row says predates
+  // it.
+  const s = statusOf(mod({
+    state: 'startup_failed',
+    liveState: null,
+    failureStage: 'require',
+    failureReason: 'module directory not present on the volume',
+  }))
+  assert.equal(s.label, 'Restart to start')
+  assert.equal(s.pending, true)
+  assert.doesNotMatch(s.detail, /not present on the volume/, 'the stale reason must not survive the install')
+})
+
+test('the restart banner appears for that install', () => {
+  // The second half of the same defect: the banner is driven by `pending`, so a
+  // row wrongly classified as failed silently removed the only way to act on it.
+  assert.equal(needsRestart([mod({ state: 'startup_failed', liveState: null })]), true)
+})
+
+test('an upgrade that has not been restarted into says so', () => {
+  // Same class as the stale-failure defect: the row is a promise about the next
+  // boot, not a description of this one. Reporting "Running v2.0.0" while the
+  // process is serving v1.0.0 would hide the only action that fixes it.
+  const s = statusOf(mod({ version: '2.0.0', liveVersion: '1.0.0' }))
+  assert.equal(s.label, 'Restart to finish upgrading')
+  assert.equal(s.pending, true)
+  assert.match(s.detail, /v2\.0\.0 is installed; v1\.0\.0 is still running/)
+})
+
+test('reinstalling the SAME version is not an upgrade in progress', () => {
+  assert.equal(statusOf(mod({ version: '1.0.0', liveVersion: '1.0.0' })).label, 'Running')
+})
+
+test('a failed module reports the stage and the reason it recorded', () => {
+  const s = statusOf(mod({
+    state: 'startup_failed',
+    liveState: 'startup_failed',
+    failureStage: 'schema',
+    failureReason: "Unknown column 'x' in 'field list'",
+  }))
+  assert.equal(s.label, 'Failed to start')
+  assert.equal(s.tone, 'bad')
+  assert.match(s.detail, /schema stage/)
+  assert.match(s.detail, /Unknown column/)
+})
+
+test('a failure with no recorded reason says so rather than showing a blank', () => {
+  const s = statusOf(mod({ state: 'startup_failed', liveState: 'startup_failed' }))
+  assert.match(s.detail, /recorded no reason/)
+})
+
+test('a row whose directory is gone by hand is bad, not merely disabled', () => {
+  // The boot reconcile marks this `startup_failed` because a row claiming to be
+  // enabled for a module that is not on the volume is simply untrue.
+  const s = statusOf(mod({ state: 'startup_failed', liveState: null, onVolume: false, failureStage: 'require', failureReason: 'module directory not present on the volume' }))
+  assert.equal(s.label, 'Missing from the volume')
+  assert.equal(s.tone, 'bad')
+})
+
+test('an uninstalled module reads as uninstalled, and says the data was kept', () => {
+  // Uninstall leaves the row `disabled` and the data alone — which is the whole
+  // point of keeping the row, so the screen has to say it.
+  const s = statusOf(mod({ state: 'disabled', liveState: 'disabled', onVolume: false }))
+  assert.equal(s.label, 'Uninstalled')
+  assert.equal(s.tone, 'idle')
+  assert.match(s.detail, /data was kept/i)
+})
+
+test('missing-from-the-volume beats every other status', () => {
+  // Ordering: a module with no files is described that way whatever its row
+  // still claims, because there is nothing there to be running.
+  for (const state of ['started', 'enabled', 'installed', 'startup_failed']) {
+    assert.match(statusOf(mod({ state, onVolume: false })).label, /Missing from the volume/)
+  }
+})
+
+// ── actionsFor ─────────────────────────────────────────────────────────────
+
+test('a running module offers disable, uninstall and a blocked purge', () => {
+  const a = actionsFor(mod())
+  assert.equal(a.disable.shown, true)
+  assert.equal(a.enable.shown, false)
+  assert.equal(a.uninstall.shown, true)
+  assert.equal(a.purge.shown, true)
+  // Shown but not clickable: the server refuses a standalone purge on anything
+  // that is not disabled, so offering the click would only produce a 409.
+  assert.equal(a.purge.enabled, false)
+  assert.match(a.purge.reason, /Disable it first/)
+})
+
+test('a disabled module offers enable, and purge is now live', () => {
+  const a = actionsFor(mod({ state: 'disabled', liveState: 'disabled' }))
+  assert.equal(a.enable.shown, true)
+  assert.equal(a.disable.shown, false)
+  assert.equal(a.purge.enabled, true)
+})
+
+test('a module with no purge.sql never offers purge, and says why', () => {
+  const a = actionsFor(mod({ state: 'disabled', liveState: 'disabled', canPurge: false }))
+  assert.equal(a.purge.shown, false)
+  assert.match(a.purge.reason, /ships no purge.sql/)
+})
+
+test('a module with no files offers only clearing the row', () => {
+  const a = actionsFor(mod({ state: 'disabled', liveState: null, onVolume: false }))
+  assert.equal(a.uninstall.shown, false)
+  assert.equal(a.disable.shown, false)
+  assert.equal(a.enable.shown, false)
+  assert.equal(a.purge.shown, false, 'there is no purge.sql left to run')
+  assert.equal(a.forget.shown, true)
+})
+
+test('a directory with no row yet is actionable, and offers nothing to forget', () => {
+  // A hand-placed install before its first boot: it has no row, so `state` is
+  // null. Its routes are already being served, so it must be disableable.
+  const a = actionsFor(mod({ state: null, liveState: 'started' }))
+  assert.equal(a.disable.shown, true)
+  assert.equal(a.uninstall.shown, true)
+  assert.equal(a.forget.shown, false)
+})
+
+// ── needsRestart ───────────────────────────────────────────────────────────
+
+test('the restart banner is driven by the list, not by any one module', () => {
+  // A restart is a property of the SERVER. One pending module is enough, and
+  // three do not mean three restarts.
+  assert.equal(needsRestart([mod(), mod({ id: 'b' })]), false)
+  assert.equal(needsRestart([mod(), mod({ id: 'b', state: 'installed', liveState: null })]), true)
+  assert.equal(needsRestart([]), false)
+})
+
+test('a disabled module does not ask for a restart', () => {
+  // Disable is immediate; a banner here would be asking for a restart that
+  // would change nothing.
+  assert.equal(needsRestart([mod({ state: 'disabled', liveState: 'disabled' })]), false)
+})
+
+test('a failed module does not ask for a restart either', () => {
+  // It is retried on every boot anyway, and the operator has to fix the cause
+  // first — a banner would suggest restarting is the remedy.
+  assert.equal(needsRestart([mod({ state: 'startup_failed', liveState: 'startup_failed' })]), false)
+})
+
+// ── parseHosts ─────────────────────────────────────────────────────────────
+
+test('parseHosts previews exactly what the server will store', () => {
+  assert.deepEqual(parseHosts('A.com, b.com\n c.com'), ['a.com', 'b.com', 'c.com'])
+  assert.deepEqual(parseHosts('  '), [])
+  assert.deepEqual(parseHosts(undefined), [])
+})
+
+// ── the declaration (slice 3) ──────────────────────────────────────────────
+
+test('a declared module that has never installed says so, with the reason', () => {
+  // No row, no directory, nothing mounted — invisible to the other three
+  // sources, so without this branch the screen would describe a module it has
+  // never had as a row gone stale.
+  const s = statusOf(mod({
+    state: null,
+    liveState: null,
+    liveVersion: null,
+    version: null,
+    onVolume: false,
+    declared: true,
+    declaredVersion: '0.3.0',
+    declaredError: 'could not reach releases.example.com',
+  }))
+  assert.equal(s.label, 'Declared, not installed')
+  assert.equal(s.tone, 'bad')
+  assert.equal(s.pending, false, 'a restart will not fix an unreachable host')
+  assert.match(s.detail, /could not reach releases.example.com/)
+})
+
+test('a declared module waiting for its first resolution is not reported as failed', () => {
+  const s = statusOf(mod({ state: null, liveState: null, version: null, onVolume: false, declared: true, declaredVersion: '0.3.0' }))
+  assert.match(s.detail, /installed when the server next starts/)
+})
+
+test('a running module whose declared upgrade is failing is still Running', () => {
+  // Both facts are true at once. The status is one label, so the declaration
+  // gets its own line rather than overwriting it.
+  const m = mod({ declared: true, declaredVersion: '2.0.0', declaredError: 'sha256 did not match' })
+  assert.equal(statusOf(m).label, 'Running')
+  const note = declarationNoteFor(m)
+  assert.equal(note.tone, 'warn')
+  assert.match(note.text, /sha256 did not match/)
+})
+
+test('uninstalling a declared module is told that its files come back', () => {
+  // The sentence that saves an afternoon: MODULES owns what is on the volume,
+  // the row owns whether it runs.
+  const note = declarationNoteFor(mod({ state: 'disabled', liveState: null, onVolume: false, declared: true, declaredVersion: '1.0.0' }))
+  assert.match(note.text, /come back when the server next starts/)
+  assert.match(note.text, /switched off/)
+})
+
+test('an ordinary declared module gets a quiet note, and an undeclared one none', () => {
+  assert.equal(declarationNoteFor(mod()), null)
+  const note = declarationNoteFor(mod({ declared: true, declaredVersion: '1.0.0' }))
+  assert.equal(note.tone, 'idle')
+  assert.match(note.text, /MODULES/)
+})

docker-compose.yml

@@ -39,6 +39,24 @@ services:
       # defaults to (<repo>/modules, and the repo is /app in the image), set
       # explicitly because the bind mount below is what makes it meaningful.
       MODULES_DIR: /app/modules
+      # WHICH modules this deployment runs (MODULE_SYSTEM.md §2.7.2 decision 4).
+      # One entry per module, `<id>@<version>=<install manifest URL>`, whitespace-
+      # or comma-separated. The container resolves this set for itself at every
+      # start: a module already unpacked at the declared version is left alone
+      # without a single network call — so a restart with the internet down comes
+      # up unchanged — and only a missing or different version is fetched,
+      # verified against the sha256 its release manifest declares, and unpacked.
+      # A failure is logged and surfaced in Admin → Modules; it never stops the
+      # site from starting.
+      #
+      # Uncomment to declare a set here, in the file this host version-controls,
+      # or leave it out and set MODULES in .env (env_file above) — or leave it
+      # unset entirely and install from the admin panel. What it declares is what
+      # is ON the volume, never whether a module runs: a module disabled from the
+      # admin panel gets its files back and stays disabled.
+      #
+      # MODULES: >-
+      #   uo@0.3.0=https://gitea.whitlocktech.com/RunicGateway/Module-uo/releases/download/v0.3.0/module-uo-0.3.0.json
     depends_on:
       db:
         condition: service_healthy
@@ -56,10 +74,12 @@ services:
       # pull-only deployment without building anything. A bind mount rather than
       # a named volume because placing a module directory by hand is a supported
       # install — `tar -xf uo-1.0.0.tgz -C ./modules` then restart — and that has
-      # to be doable from the host, not through `docker cp`.
+      # to be doable from the host, not through `docker cp`. It is no longer the
+      # usual way in: declare MODULES above, or install from Admin → Modules.
       #
-      # Read-WRITE: the admin panel's install/uninstall unpacks and removes
-      # directories here from inside the container.
+      # Read-WRITE: MODULES resolution at start, and the admin panel's
+      # install/uninstall, both unpack and remove directories here from inside
+      # the container.
       #
       # `modules/` is tracked (it ships a README) so the directory exists in the
       # checkout with the operator's own ownership. Do not delete it — Docker

server/.env.example

@@ -143,5 +143,24 @@ ANNOUNCE_POLL_MS=15000
 #                        deployment does nothing, deliberately, so a redeploy cannot
 #                        silently undo an operator's choice. Installs are https-only
 #                        and an empty list forbids all of them.
+#   MODULES              The module set this deployment RUNS, resolved at every
+#                        start (§2.7.2 decision 4). One entry per module, separated
+#                        by whitespace or commas:
+#
+#                            <id>@<version>=<install manifest URL>
+#
+#                        A module already unpacked at the declared version is left
+#                        alone WITHOUT touching the network, so a restart with no
+#                        route to the internet comes up unchanged; only a missing or
+#                        different version is fetched, through the same verify-and-
+#                        unpack path (and the same host allowlist) the admin panel
+#                        uses. A version that cannot be fetched is logged and shown
+#                        in Admin → Modules — it never stops the site from starting.
+#
+#                        This variable owns what is ON the volume, not what runs: a
+#                        module disabled from the admin panel stays disabled even
+#                        though its files are put back. Leave it unset to manage
+#                        modules entirely from the admin panel.
 # MODULES_DIR=/app/modules
 # MODULE_SOURCE_HOSTS=gitea.whitlocktech.com
+# MODULES=uo@0.3.0=https://gitea.whitlocktech.com/RunicGateway/Module-uo/releases/download/v0.3.0/module-uo-0.3.0.json

server/src/modules/declared.js

@@ -0,0 +1,220 @@
+// ── The declared module set ────────────────────────────────────────────────
+//
+// Phase 4, slice 3 of docs/website/MODULE_SYSTEM.md §2.7.2 — decision 4. A
+// compose-managed host is not driven by clicking: it declares which modules it
+// runs, in the file it already edits and version-controls, and the container
+// arrives at that set by itself.
+//
+//   MODULES: uo@0.3.0=https://<host>/…/module-uo-0.3.0.json
+//
+// One entry per module, `<id>@<version>=<install manifest URL>`, separated by
+// whitespace or commas. The id and the version are written out rather than left
+// to be discovered inside the manifest for one reason: **the no-op case must not
+// need the network.** A module already unpacked at the declared version is
+// answered by reading its own `module.json` off the volume, so a restart with
+// the network down brings the site up exactly as it was. Only a module that is
+// missing, or unpacked at some other version, reaches out — and it reaches out
+// through modules/install.js, the same fetch-verify-unpack path the admin panel
+// uses, under the same host allowlist.
+//
+// Three things this file deliberately does not do:
+//
+//   - **It does not decide whether a module RUNS.** Resolution owns what is on
+//     the volume; `installed_modules` owns whether a mounted module answers. An
+//     admin who uninstalls a declared module gets its directory back at the next
+//     boot with the row still `disabled`, so it stays off until they enable it.
+//     The two never fight because they are not answering the same question.
+//   - **It does not fail a boot.** A module publisher's host being unreachable
+//     must not take a shard's website down with it; core is built to serve with
+//     a module absent (§1.6). Every failure is logged loudly and kept for the
+//     admin screen, and the site comes up.
+//   - **It does not mount anything.** §1.12 makes the volume the mounting source
+//     of truth, read once at require time — which is why this runs before
+//     `require('./app')` in server.js and not from inside it.
+//
+// It runs on every boot, not only in Docker: a bare `npm start` with MODULES set
+// resolves the same way. The Docker path is the reason it exists, not a special
+// case in it.
+
+const fs = require('fs')
+const path = require('path')
+
+const install = require('./install')
+
+const log = require('../utils/logger')('modules')
+
+// The variable an operator sets. Named next to MODULES_DIR, which is the other
+// half of the same story: one says where modules live, the other says which.
+const VAR = 'MODULES'
+
+// Same id rule the loader enforces when scanning and install.js enforces when
+// placing, restated here so a declaration cannot name something neither would
+// accept.
+const ID = /^[a-z][a-z0-9-]{1,31}$/
+
+// The outcome of the last resolution, in memory, for the admin screen. Not a
+// database row: a declaration is a fact about this process's environment, and
+// writing it down would put it in front of the boot reconcile, which resets
+// every non-disabled row (§2.4). The screen merges it as a fourth source
+// alongside the row, the loader and the volume.
+let results = []
+
+/**
+ * Parse the declaration into entries.
+ *
+ * Malformed entries are collected rather than thrown: one operator typo should
+ * cost that module, not every module on the host. A duplicate id keeps the
+ * first — there is no sensible way to run two versions of one module, and
+ * silently preferring the last would make the outcome depend on the order of a
+ * list nobody reads as ordered.
+ *
+ * @param {string} value the raw variable
+ * @returns {{entries: Array<{id,version,url}>, errors: string[]}}
+ */
+function parse(value) {
+  const entries = []
+  const errors = []
+  const seen = new Set()
+
+  for (const token of String(value || '').split(/[,\s]+/).filter(Boolean)) {
+    const at = token.indexOf('@')
+    const eq = token.indexOf('=')
+    if (at < 1 || eq < at + 2) {
+      errors.push(`"${token}" is not <id>@<version>=<url>`)
+      continue
+    }
+    const id = token.slice(0, at)
+    const version = token.slice(at + 1, eq)
+    const url = token.slice(eq + 1)
+
+    if (!ID.test(id)) {
+      errors.push(`"${token}" names an invalid module id "${id}"`)
+      continue
+    }
+    if (!url) {
+      errors.push(`"${token}" has no install manifest URL`)
+      continue
+    }
+    if (seen.has(id)) {
+      errors.push(`"${id}" is declared more than once — keeping the first`)
+      continue
+    }
+    seen.add(id)
+    entries.push({ id, version, url })
+  }
+
+  return { entries, errors }
+}
+
+/**
+ * The version currently unpacked on the volume, or null.
+ *
+ * Read straight out of the module's own `module.json`, which is the same file
+ * the loader trusts for the same fact — and never from `installed_modules`,
+ * because the row records what was installed and this has to answer what is
+ * actually there. An unreadable manifest counts as absent: whatever is in that
+ * directory, it is not a module at the declared version.
+ */
+function installedVersion(id) {
+  try {
+    const manifest = JSON.parse(
+      fs.readFileSync(path.join(install.moduleDir(id), 'module.json'), 'utf8'),
+    )
+    return manifest && manifest.version ? String(manifest.version) : null
+  } catch {
+    return null
+  }
+}
+
+/**
+ * Bring the volume in line with the declaration.
+ *
+ * Never throws and never rejects. Returns one outcome per declared entry, and
+ * remembers them for `state()`.
+ *
+ * @param {object} args
+ * @param {string} [args.value] the raw variable (defaults to the environment)
+ * @param {string[]} args.hosts the install allowlist, already parsed
+ * @param {object} args.model  modules.model, for recording provenance
+ * @param {object} [args.installImpl] injection seam, as everywhere else here
+ * @returns {Promise<Array<{id,version,url,action,message}>>}
+ */
+async function resolve({ value = process.env[VAR], hosts = [], model, installImpl = install } = {}) {
+  const { entries, errors } = parse(value)
+  results = []
+
+  for (const message of errors) log.error(`${VAR}: ${message}`)
+
+  if (!entries.length) return results
+
+  log.info(`${VAR} declares ${entries.length} module(s)`, {
+    modules: entries.map((e) => `${e.id}@${e.version}`).join(' '),
+  })
+
+  for (const entry of entries) {
+    const present = installedVersion(entry.id)
+    if (present === entry.version) {
+      // The offline path, and the common one: nothing is fetched, nothing is
+      // written, and a host with no route to the internet boots unchanged.
+      log.info(`module "${entry.id}" is already at the declared version ${entry.version}`)
+      results.push({ ...entry, action: 'noop', message: null })
+      continue
+    }
+
+    try {
+      // `expect` is the declaration itself, handed down so install.js can refuse
+      // a URL that resolves to another module or another version while it is
+      // still only a manifest — a check made after the unpack would be made with
+      // the undeclared module already on the volume.
+      const result = await installImpl.install({
+        url: entry.url,
+        hosts,
+        expect: { id: entry.id, version: entry.version },
+      })
+
+      // Provenance, written exactly as the admin route writes it — the whole
+      // point of resolving in-process rather than from a script that cannot
+      // reach the database. A module installed by the compose file and one
+      // installed by an admin are then indistinguishable on the screen, which
+      // is what makes this one feature and not two.
+      if (model) {
+        await model.recordInstalled({
+          id: result.id,
+          name: result.name,
+          version: result.version,
+          source: entry.url,
+          sha256: result.sha256,
+        })
+      }
+
+      log.warn(`installed declared module "${entry.id}" v${entry.version}`, {
+        from: present || 'nothing',
+        source: entry.url,
+        sha256: result.sha256,
+      })
+      results.push({ ...entry, action: 'installed', message: null })
+    } catch (err) {
+      // Loud, and then onward. The site serves; this module does not, or serves
+      // the version that was already there.
+      log.error(
+        `could not resolve declared module "${entry.id}@${entry.version}": ${err.message}` +
+          (present ? ` — leaving version ${present} in place` : ''),
+      )
+      results.push({ ...entry, action: 'failed', message: err.message })
+    }
+  }
+
+  return results
+}
+
+/** What the last resolution decided, for the admin screen. */
+function state() {
+  return results.map((r) => ({ ...r }))
+}
+
+/** Test seam: forget the last resolution. */
+function reset() {
+  results = []
+}
+
+module.exports = { VAR, parse, installedVersion, resolve, state, reset }

server/src/modules/install.js

@@ -79,6 +79,13 @@ class InstallError extends Error {
 
 // ── The allowlist ──────────────────────────────────────────────────────────
 
+// The settings row the allowlist lives in (decision 6): seeded from
+// MODULE_SOURCE_HOSTS on a fresh install and admin-managed from then on. The KEY
+// lives here rather than in the admin controller because it is now read from two
+// places — the controller, and the boot-time resolution of the declared module
+// set (modules/declared.js), which has no route and no request.
+const HOSTS_SETTING = 'module_source_hosts'
+
 /**
  * Parse the stored allowlist setting into hostnames.
  *
@@ -313,13 +320,34 @@ function purgeFile(id) {
  * scratch directory that is removed on any failure, and the move into place is
  * the last step.
  *
+ * `expect` is what the CALLER was promised, as opposed to what the manifest
+ * promises about itself — the declared module set (modules/declared.js) pins an
+ * id and a version in the environment, and a URL that resolves to something else
+ * has to be refused rather than installed. Checked against the manifest, before
+ * a byte is downloaded: catching it after the unpack would mean the undeclared
+ * module is already on the volume when the objection is raised. The admin panel
+ * passes nothing, because there a URL is the whole of what was asked for.
+ *
  * @param {object} args
  * @param {string} args.url    the install manifest URL the admin pasted
  * @param {string[]} args.hosts the allowlist, already parsed
+ * @param {{id?: string, version?: string}} [args.expect] what the caller pinned
  * @returns {Promise<{id,name,version,sha256,source,bytes,replaced}>}
  */
-async function install({ url, hosts, fetchImpl = fetch }) {
+async function install({ url, hosts, expect = null, fetchImpl = fetch }) {
   const manifest = await fetchManifest(url, hosts, fetchImpl)
+
+  if (expect && expect.id && manifest.id !== expect.id) {
+    throw new InstallError(
+      `that URL installs the module "${manifest.id}", but "${expect.id}" was asked for`,
+    )
+  }
+  if (expect && expect.version && manifest.version !== expect.version) {
+    throw new InstallError(
+      `that URL installs ${manifest.id} v${manifest.version}, but v${expect.version} was asked for`,
+    )
+  }
+
   const target = moduleDir(manifest.id)
   const scratch = await fsp.mkdtemp(path.join(loader.dir(), `.install-${manifest.id}-`))
   const tarball = path.join(scratch, 'bundle.tar.gz')
@@ -407,6 +435,7 @@ async function removeDir(id) {
 
 module.exports = {
   InstallError,
+  HOSTS_SETTING,
   parseHosts,
   checkUrl,
   get,

server/src/router/v1/admin/modules.controller.js

@@ -30,6 +30,7 @@ const settings = require('../../../model/settings/settings.model')
 const loader = require('../../../modules/loader')
 const lifecycle = require('../../../modules/lifecycle')
 const install = require('../../../modules/install')
+const declared = require('../../../modules/declared')
 // A namespace import, like every other require in this file, and not
 // `const { runPurge } = …`: destructuring at require time captures the function
 // rather than the module, which makes it the one dependency here that cannot be
@@ -42,8 +43,9 @@ const log = require('../../../utils/logger')('admin-modules')
 // The allowlist setting. Seeded from MODULE_SOURCE_HOSTS on first boot and
 // admin-managed from then on (decision 6) — db/seed.js writes it once and never
 // overwrites it, so changing the variable later does not silently reach in and
-// undo an operator's choice.
-const HOSTS_KEY = 'module_source_hosts'
+// undo an operator's choice. The key itself lives in install.js, which is now
+// read by boot-time declared-set resolution as well as by this controller.
+const HOSTS_KEY = install.HOSTS_SETTING
 
 // A hostname, not a URL: no scheme, no path, no port, no wildcard. Deliberately
 // strict — every character allowed here is a character that can appear in the
@@ -57,24 +59,29 @@ async function allowedHosts() {
 /**
  * One module, as the admin screen needs it.
  *
- * Three sources have to be reconciled, and which one answers which question is
+ * Four sources have to be reconciled, and which one answers which question is
  * the whole of §2.4:
  *
  *   - the ROW says what the operator decided and what the last boot recorded;
  *   - the LOADER says what is mounted and answering right now;
- *   - the VOLUME says whether there is still a directory there at all.
+ *   - the VOLUME says whether there is still a directory there at all;
+ *   - the DECLARATION (slice 3) says what this container's environment asks for,
+ *     which is the only one of the four an admin cannot change from this screen.
  *
  * They can legitimately disagree, and the screen has to show that rather than
  * pick a winner. A row `enabled` with a loader state of `disabled` is a module
  * the operator has just switched back on and which is waiting for a restart —
  * exactly the case decision 3 creates, and it would be a lie to render it as
- * either "running" or "off".
+ * either "running" or "off". A declared module with no row and no directory is
+ * the newest of those disagreements: MODULES asked for it and resolution could
+ * not get it, so the screen carries the reason rather than showing nothing.
  */
-function present(row, live, onVolume) {
+function present(row, live, onVolume, declaration = null) {
+  const id = row ? row.id : live ? live.id : declaration.id
   return {
-    id: row ? row.id : live.id,
-    name: row ? row.name : live.name,
-    version: row ? row.version : live.version,
+    id,
+    name: row ? row.name : live ? live.name : id,
+    version: row ? row.version : live ? live.version : null,
     // What the database records.
     state: row ? row.state : null,
     failureStage: row ? row.failureStage : (live && live.stage) || null,
@@ -93,7 +100,16 @@ function present(row, live, onVolume) {
     capabilities: live ? live.capabilities : [],
     // What is on the volume.
     onVolume,
-    canPurge: onVolume && Boolean(install.purgeFile(row ? row.id : live.id)),
+    canPurge: onVolume && Boolean(install.purgeFile(id)),
+    // What the environment declares. `declaredVersion` is what MODULES pins, not
+    // what is installed — they differ exactly while a resolution is failing, and
+    // `declaredError` says why. Uninstalling a declared module from this screen
+    // removes its directory and disables its row; the next boot puts the
+    // directory back and leaves the row disabled, so the screen says so rather
+    // than letting the files reappear unexplained.
+    declared: Boolean(declaration),
+    declaredVersion: declaration ? declaration.version : null,
+    declaredError: declaration && declaration.action === 'failed' ? declaration.message : null,
   }
 }
 
@@ -107,18 +123,31 @@ async function list(req, res) {
     // `npm run seed` never gets here, but a test harness might.
     const live = loader.isLoaded() ? loader.list() : []
     const byId = new Map(live.map((m) => [m.id, m]))
+    const declaredById = new Map(declared.state().map((d) => [d.id, d]))
 
     const seen = new Set()
     const out = []
     for (const row of rows) {
       seen.add(row.id)
-      out.push(present(row, byId.get(row.id) || null, install.isInstalled(row.id)))
+      out.push(
+        present(row, byId.get(row.id) || null, install.isInstalled(row.id), declaredById.get(row.id)),
+      )
     }
     // A directory on the volume that has no row yet — a hand-placed install
     // before its first boot. It has to be listed, or the screen would show
     // nothing for a module whose routes are already being served.
     for (const m of live) {
-      if (!seen.has(m.id)) out.push(present(null, m, true))
+      if (!seen.has(m.id)) {
+        seen.add(m.id)
+        out.push(present(null, m, true, declaredById.get(m.id)))
+      }
+    }
+    // A module MODULES declares that has neither. Resolution failed and left
+    // nothing behind — the case an operator most needs told, because from the
+    // screen's other three sources it is indistinguishable from never having
+    // asked for it.
+    for (const d of declaredById.values()) {
+      if (!seen.has(d.id)) out.push(present(null, null, install.isInstalled(d.id), d))
     }
 
     return res.json({ modules: out, sourceHosts: await allowedHosts() })

server/src/server.js

@@ -1,8 +1,12 @@
 require('dotenv').config()
 const http = require('http')
 
-const app = require('./app')
-const internalApp = require('./internalApp')
+// NOTE: `./app` and `./internalApp` are deliberately NOT required here. Requiring
+// app.js runs `modules.load()`, which scans the volume and mounts whatever is on
+// it (MODULE_API.md §4.1) — so the declared module set has to be resolved before
+// that require, not before the listener. They are required inside start(), after
+// resolveDeclaredModules(); everything else this file needs is safe to pull in
+// now because none of it reaches the loader's scan.
 const botScore = require('./middleware/botScore')
 const announceWorker = require('./utils/announceWorker')
 const { ensureSchema, close } = require('./utils/db')
@@ -11,6 +15,9 @@ const settings = require('./model/settings/settings.model')
 const revokedSessions = require('./model/revokedSessions/revokedSessions.model')
 const mobileAuthBridge = require('./model/mobileAuthBridge/mobileAuthBridge.model')
 const moduleLifecycle = require('./modules/lifecycle')
+const declaredModules = require('./modules/declared')
+const moduleInstall = require('./modules/install')
+const moduleModel = require('./model/modules/modules.model')
 const createLogger = require('./utils/logger')
 const { evaluateBotInternalKey } = require('./utils/botInternalKey')
 const brand = require('./config/brand')
@@ -51,7 +58,9 @@ async function start() {
   }
 
   log.info('ensuring database schema...')
-  await ensureSchema()
+  // Core's schema only. Each installed module's fragment is replayed further
+  // down, after the volume has been scanned — see the require of ./app below.
+  await ensureSchema({ replayModules: false })
   log.info('seeding defaults...')
   await seedDefaults()
   await createInitialAdminFromEnv()
@@ -77,6 +86,36 @@ async function start() {
   const mode = await settings.get('site_mode')
   log.info(`site mode: ${String(mode || 'live').toUpperCase()}`)
 
+  // Bring the modules volume in line with what MODULES declares (§2.7.2
+  // decision 4), and only then require the app — the loader scans and mounts at
+  // require time, so this is the last moment at which a module can be put on the
+  // volume and still be part of this process.
+  //
+  // After the schema and the seed, because the host allowlist it installs under
+  // is a settings row that the seed creates on a fresh instance. Never throws:
+  // an unreachable release host leaves the site serving without that module
+  // rather than taking the site down with it.
+  await declaredModules.resolve({
+    hosts: moduleInstall.parseHosts(await settings.get(moduleInstall.HOSTS_SETTING)),
+    model: moduleModel,
+  })
+
+  // Requiring app.js is what scans the volume and mounts what is on it. Every
+  // line above this one runs against a core that has no modules in it yet.
+  // eslint-disable-next-line global-require
+  const app = require('./app')
+  // eslint-disable-next-line global-require
+  const internalApp = require('./internalApp')
+
+  // Now that the scan has happened, replay each module's schema fragment
+  // (MODULE_API.md §2.6). This used to ride inside ensureSchema() and could,
+  // because app.js was required at the top of this file; resolving the declared
+  // set first moved the scan after it, and a booting server quietly getting no
+  // module tables is precisely what §7.6 warns about. Caught by the browser
+  // smoke rather than by a test: every suite here stubs one side or the other.
+  // eslint-disable-next-line global-require
+  await require('./modules/schema').replayFragments()
+
   // Reconcile installed_modules with what the loader found on the volume at
   // require time, then run each module's onBoot (MODULE_API.md §2.5).
   //

server/src/utils/db.js

@@ -55,11 +55,18 @@ const SCHEMA_PATH = path.join(__dirname, '..', '..', 'db', 'schema.sql')
  * below — the discovery, splitting and per-module failure handling all live in
  * modules/schema.js, required lazily so that requiring the pool never drags the
  * loader in with it.
+ *
+ * `replayModules: false` is for a caller that has not scanned the volume YET and
+ * intends to. server.js is the one: since slice 3 it resolves the declared
+ * module set before requiring app.js, which puts core's schema *before* the scan
+ * — so it replays the fragments itself, in the one place that knows the scan has
+ * happened. Left true everywhere else, so the ordinary caller cannot get module
+ * tables by accident and lose them by refactor.
  */
-async function ensureSchema({ retries = 10, delayMs = 2000 } = {}) {
+async function ensureSchema({ retries = 10, delayMs = 2000, replayModules = true } = {}) {
   await ensureCoreSchema({ retries, delayMs })
   // eslint-disable-next-line global-require
-  await require('../modules/schema').replayFragments()
+  if (replayModules) await require('../modules/schema').replayFragments()
 }
 
 /** Core's own schema.sql, with the wait-for-the-database retry. */

server/test/adminModules.test.js

@@ -29,6 +29,7 @@ const settings = require('../src/model/settings/settings.model')
 const loader = require('../src/modules/loader')
 const lifecycle = require('../src/modules/lifecycle')
 const install = require('../src/modules/install')
+const declared = require('../src/modules/declared')
 const schema = require('../src/modules/schema')
 const db = require('../src/utils/db')
 
@@ -67,6 +68,7 @@ const originals = {
     removeDir: install.removeDir,
   },
   schema: { runPurge: schema.runPurge },
+  declared: { state: declared.state },
 }
 
 let logged
@@ -79,6 +81,7 @@ beforeEach(() => {
   Object.assign(lifecycle, originals.lifecycle)
   Object.assign(install, originals.install)
   Object.assign(schema, originals.schema)
+  Object.assign(declared, originals.declared)
 
   logged = []
   activity.log = async (entry) => { logged.push(entry) }
@@ -133,6 +136,55 @@ test('list includes a module on the volume that has no row yet', async () => {
   assert.equal(res.body.modules[0].liveState, 'started')
 })
 
+test('list carries what MODULES declares, including a module it could not install', async () => {
+  // Slice 3's fourth source. A declared module that failed to resolve has no
+  // row, no directory and nothing mounted — so it is invisible to the other
+  // three, and the reason it is missing is the one thing the operator needs.
+  modules.list = async () => [{
+    id: 'uo', name: 'UO', version: '0.2.0', state: 'enabled',
+    failureStage: null, failureReason: null, source: null, sha256: null,
+    installedAt: null, startedAt: null,
+  }]
+  loader.list = () => [{ id: 'uo', name: 'UO', version: '0.2.0', state: 'started', stage: null, reason: null, capabilities: [] }]
+  install.isInstalled = (id) => id === 'uo'
+  install.purgeFile = () => null
+  declared.state = () => [
+    { id: 'uo', version: '0.3.0', url: 'https://x/uo.json', action: 'failed', message: 'the host is down' },
+    { id: 'market', version: '1.0.0', url: 'https://x/market.json', action: 'failed', message: 'not found' },
+  ]
+
+  const res = mockRes()
+  await ctrl.list(req(), res)
+
+  const byId = Object.fromEntries(res.body.modules.map((m) => [m.id, m]))
+  // Running fine at 0.2.0 while the declared upgrade to 0.3.0 is failing: both
+  // facts survive, because collapsing them would have to discard one.
+  assert.equal(byId.uo.liveState, 'started')
+  assert.equal(byId.uo.declared, true)
+  assert.equal(byId.uo.declaredVersion, '0.3.0')
+  assert.equal(byId.uo.declaredError, 'the host is down')
+  // Declared and nowhere: listed anyway, with no version invented for it.
+  assert.equal(byId.market.declared, true)
+  assert.equal(byId.market.version, null)
+  assert.equal(byId.market.state, null)
+  assert.equal(byId.market.declaredError, 'not found')
+})
+
+test('a successfully resolved module is marked declared, with no error', async () => {
+  modules.list = async () => []
+  loader.list = () => [{ id: 'uo', name: 'UO', version: '0.3.0', state: 'started', stage: null, reason: null, capabilities: [] }]
+  install.isInstalled = () => true
+  install.purgeFile = () => null
+  declared.state = () => [{ id: 'uo', version: '0.3.0', url: 'https://x/uo.json', action: 'noop', message: null }]
+
+  const res = mockRes()
+  await ctrl.list(req(), res)
+
+  assert.equal(res.body.modules.length, 1, 'declared and present is ONE module, not two')
+  assert.equal(res.body.modules[0].declared, true)
+  assert.equal(res.body.modules[0].declaredError, null)
+})
+
 test('list survives a process where the loader never scanned', async () => {
   loader.isLoaded = () => false
   loader.list = () => { throw new Error('modules.list() before modules.load()') }

server/test/bootOrder.test.js

@@ -0,0 +1,82 @@
+// server.js's boot ORDER, which is a contract and not a style choice.
+//
+// Phase 4, slice 3 of MODULE_SYSTEM.md §2.7.2. Five steps have to happen in one
+// order, and each arrow is a dependency that is invisible at the call site:
+//
+//   core schema  →  the declared set (§2.7.2 decision 4) needs the settings row
+//                   its seed writes, to know which hosts it may install from
+//   declared set →  requiring app.js SCANS the volume (§1.12), so anything put
+//                   there afterwards is not in this process
+//   require app  →  module schema fragments (§2.6) can only be replayed once the
+//                   loader knows which modules there are
+//   fragments    →  onBoot runs against tables that exist
+//
+// This is a source-structure test, and it is worth being plain about what that
+// does and does not prove: it cannot tell you the server boots, only that nobody
+// has quietly moved one of these five lines past another. It exists because the
+// defect it guards against has already happened once and was invisible to every
+// other kind of test here. Deferring the app require — which slice 3 had to do —
+// moved core's schema ahead of the scan, and `ensureSchema` then skipped the
+// module fragments entirely. On this machine's dev database the tables already
+// existed, so the module started; on a FRESH database it would have started
+// against no tables at all. Every suite in this directory stubs either the
+// loader or the pool, so none of them could see it. The browser smoke did, from
+// one log line.
+process.env.DB_HOST = '127.0.0.1'
+process.env.DB_PORT = '59999'
+
+const fs = require('fs')
+const path = require('path')
+
+const { test } = require('node:test')
+const assert = require('node:assert/strict')
+
+const SERVER = fs.readFileSync(path.join(__dirname, '..', 'src', 'server.js'), 'utf8')
+
+/** Where a marker appears, asserted to appear exactly once. */
+function at(marker) {
+  const first = SERVER.indexOf(marker)
+  assert.notEqual(first, -1, `server.js no longer contains ${JSON.stringify(marker)}`)
+  assert.equal(
+    SERVER.indexOf(marker, first + 1),
+    -1,
+    `${JSON.stringify(marker)} appears more than once in server.js — this test cannot tell which is the boot step`,
+  )
+  return first
+}
+
+test('boot runs core schema, then the declared set, then the scan, then fragments, then onBoot', () => {
+  const steps = [
+    ['core schema', at('await ensureSchema({ replayModules: false })')],
+    ['declared module set', at('await declaredModules.resolve(')],
+    ['the volume scan', at("require('./app')")],
+    ['module schema fragments', at('replayFragments()')],
+    ['module onBoot', at('await moduleLifecycle.boot()')],
+  ]
+
+  for (let i = 1; i < steps.length; i += 1) {
+    assert.ok(
+      steps[i][1] > steps[i - 1][1],
+      `${steps[i][0]} must come after ${steps[i - 1][0]} in server.js`,
+    )
+  }
+})
+
+test('app.js is required inside start(), never at the top of the file', () => {
+  // The whole mechanism depends on this. A top-level require runs when server.js
+  // is loaded — before a single line of start() — and the scan would then happen
+  // before the declared set had a chance to put anything on the volume, silently
+  // and with no error anywhere.
+  assert.ok(
+    at("require('./app')") > at('async function start()'),
+    'requiring ./app at the top of server.js scans the volume before the declared set is resolved',
+  )
+})
+
+test('ensureSchema is asked NOT to replay fragments, and something else does', () => {
+  // Both halves matter. Passing the flag without replaying elsewhere is the
+  // original defect with an explicit spelling; replaying without the flag runs
+  // the fragments twice, the second time being the one that matters.
+  assert.match(SERVER, /ensureSchema\(\{ replayModules: false \}\)/)
+  assert.match(SERVER, /modules\/schema'\)\.replayFragments\(\)/)
+})

server/test/loginProtection.test.js

@@ -60,7 +60,13 @@ test('backoffGuard returns a generic 429 while locked out', async () => {
     a.post('/login', lp.backoffGuard, (req, res) => res.json({ ok: true }))
   })
   try {
-    lp.recordFailure('203.0.113.40') // lock the test client IP
+    // Lock the test client IP. FIVE failures, not one: the lock is
+    // `BASE_MS * 2 ** (count - 1)`, so a single failure locks for exactly one
+    // second and this test then races the round trip. It lost that race on CI
+    // (200 instead of 429, request arriving 1,456 ms after the lock). Five
+    // failures lock for sixteen seconds, which is not a race. What is under
+    // test is the guard's ANSWER while locked out, and that is unchanged.
+    for (let i = 0; i < 5; i += 1) lp.recordFailure('203.0.113.40')
     const res = await fetch(`${app.url}/login`, {
       method: 'POST',
       headers: { 'X-Forwarded-For': '203.0.113.40' },

server/test/moduleDeclared.test.js

@@ -0,0 +1,391 @@
+// modules/declared.js — resolving the module set an environment declares.
+//
+// Phase 4, slice 3 of MODULE_SYSTEM.md §2.7.2, decision 4. Two claims are worth
+// more than the rest and both are about what does NOT happen:
+//
+//   - a module already unpacked at the declared version does not touch the
+//     network, because that is what makes a restart with the internet down come
+//     up unchanged;
+//   - a module that cannot be resolved does not fail the boot, and does not stop
+//     the next declaration from resolving.
+//
+// The last test in the file runs the whole path through the real install.js
+// against a fake transport — a real gzipped tar, really hashed, really unpacked
+// — because everything above it stubs `installImpl` and would keep passing if
+// the two files stopped agreeing about what an install returns.
+process.env.DB_HOST = '127.0.0.1'
+process.env.DB_PORT = '59999'
+
+const crypto = require('crypto')
+const fs = require('fs')
+const os = require('os')
+const path = require('path')
+const zlib = require('zlib')
+
+const { test, beforeEach, after } = require('node:test')
+const assert = require('node:assert/strict')
+
+const db = require('../src/utils/db')
+
+after(() => db.close())
+
+const HOSTS = ['releases.example.com']
+const URL_030 = 'https://releases.example.com/mod/uo-0.3.0.json'
+
+let tmpRoot
+let declared
+
+/**
+ * A fresh declared.js bound to a fresh modules directory.
+ *
+ * loader.js resolves MODULES_DIR once at require time and install.js reads
+ * `loader.dir()`, so all three have to come back together — the same dance
+ * moduleInstall.test.js and moduleLifecycle.test.js do.
+ */
+function fresh(dir) {
+  process.env.MODULES_DIR = dir
+  for (const m of ['loader', 'install', 'declared']) {
+    delete require.cache[require.resolve(`../src/modules/${m}`)]
+  }
+  // eslint-disable-next-line global-require
+  return require('../src/modules/declared')
+}
+
+/** Put a module directory on the volume, as an unpack would leave it. */
+function place(id, version) {
+  const dir = path.join(tmpRoot, id)
+  fs.mkdirSync(dir, { recursive: true })
+  fs.writeFileSync(
+    path.join(dir, 'module.json'),
+    JSON.stringify({ id, name: 'Ultima Online', version, coreApi: '^1.0.0' }),
+  )
+  return dir
+}
+
+/**
+ * A stub install service.
+ *
+ * Records every call, so "did this reach the network at all" is a question the
+ * tests can ask directly rather than inferring from an outcome.
+ */
+function fakeInstall({ version = '0.3.0', throws = null } = {}) {
+  const calls = []
+  return {
+    calls,
+    InstallError: Error,
+    async install(args) {
+      calls.push(args)
+      if (throws) throw new Error(throws)
+      return {
+        id: 'uo',
+        name: 'Ultima Online',
+        version,
+        sha256: 'a'.repeat(64),
+        source: args.url,
+        bytes: 1024,
+        replaced: false,
+      }
+    },
+  }
+}
+
+/** A modules.model stand-in that remembers what provenance it was given. */
+function fakeModel() {
+  const recorded = []
+  return { recorded, async recordInstalled(row) { recorded.push(row); return row } }
+}
+
+beforeEach(() => {
+  tmpRoot = fs.mkdtempSync(path.join(os.tmpdir(), 'rg-declared-'))
+  declared = fresh(tmpRoot)
+})
+
+// ── Parsing ────────────────────────────────────────────────────────────────
+
+test('parses id@version=url entries separated by whitespace or commas', () => {
+  const { entries, errors } = declared.parse(
+    ` uo@0.3.0=https://a.example/uo.json,\n  market@1.2.3=https://a.example/market.json `,
+  )
+  assert.deepEqual(errors, [])
+  assert.deepEqual(entries, [
+    { id: 'uo', version: '0.3.0', url: 'https://a.example/uo.json' },
+    { id: 'market', version: '1.2.3', url: 'https://a.example/market.json' },
+  ])
+})
+
+test('an empty or unset declaration parses to nothing, without complaint', () => {
+  for (const value of [undefined, '', '   ']) {
+    const { entries, errors } = declared.parse(value)
+    assert.deepEqual(entries, [])
+    assert.deepEqual(errors, [])
+  }
+})
+
+test('a malformed entry is refused on its own, leaving the others', () => {
+  // A bare URL, a missing version, and an id that is not one — each of which an
+  // operator can plausibly type, and none of which should cost the module that
+  // was written correctly.
+  const { entries, errors } = declared.parse(
+    'https://a.example/uo.json uo@=https://a.example/uo.json Uo@1=https://a.example/uo.json'
+      + ' good@1.0.0=https://a.example/good.json',
+  )
+  assert.equal(entries.length, 1)
+  assert.equal(entries[0].id, 'good')
+  assert.equal(errors.length, 3)
+})
+
+test('a duplicate id keeps the first and says so', () => {
+  const { entries, errors } = declared.parse(
+    'uo@1.0.0=https://a.example/one.json uo@2.0.0=https://a.example/two.json',
+  )
+  assert.equal(entries.length, 1)
+  assert.equal(entries[0].version, '1.0.0')
+  assert.match(errors[0], /declared more than once/)
+})
+
+// ── Resolution ─────────────────────────────────────────────────────────────
+
+test('a module already at the declared version is a no-op that never fetches', async () => {
+  place('uo', '0.3.0')
+  const installer = fakeInstall()
+  const model = fakeModel()
+
+  const results = await declared.resolve({
+    value: `uo@0.3.0=${URL_030}`,
+    hosts: HOSTS,
+    model,
+    installImpl: installer,
+  })
+
+  assert.deepEqual(results.map((r) => r.action), ['noop'])
+  // The offline guarantee, stated as an assertion: nothing was fetched and
+  // nothing was written. A restart with no route to the internet is this case.
+  assert.equal(installer.calls.length, 0)
+  assert.equal(model.recorded.length, 0)
+})
+
+test('a missing module is installed, with its provenance recorded', async () => {
+  const installer = fakeInstall()
+  const model = fakeModel()
+
+  const results = await declared.resolve({
+    value: `uo@0.3.0=${URL_030}`,
+    hosts: HOSTS,
+    model,
+    installImpl: installer,
+  })
+
+  assert.deepEqual(results.map((r) => r.action), ['installed'])
+  assert.equal(installer.calls[0].url, URL_030)
+  assert.deepEqual(installer.calls[0].hosts, HOSTS)
+  // The declaration is handed down so install.js can refuse a URL that turns out
+  // to be another module or another version BEFORE it downloads it.
+  assert.deepEqual(installer.calls[0].expect, { id: 'uo', version: '0.3.0' })
+  // Written exactly as the admin route writes it — the reason this runs in the
+  // server process rather than in a script that cannot reach the database.
+  assert.deepEqual(model.recorded, [
+    { id: 'uo', name: 'Ultima Online', version: '0.3.0', source: URL_030, sha256: 'a'.repeat(64) },
+  ])
+})
+
+test('a module at a different version is re-resolved', async () => {
+  place('uo', '0.2.0')
+  const installer = fakeInstall()
+
+  const results = await declared.resolve({
+    value: `uo@0.3.0=${URL_030}`,
+    hosts: HOSTS,
+    model: fakeModel(),
+    installImpl: installer,
+  })
+
+  assert.deepEqual(results.map((r) => r.action), ['installed'])
+  assert.equal(installer.calls.length, 1)
+})
+
+test('a directory with no readable module.json counts as absent', async () => {
+  fs.mkdirSync(path.join(tmpRoot, 'uo'), { recursive: true })
+  fs.writeFileSync(path.join(tmpRoot, 'uo', 'module.json'), '{ this is not json')
+  const installer = fakeInstall()
+
+  await declared.resolve({
+    value: `uo@0.3.0=${URL_030}`,
+    hosts: HOSTS,
+    model: fakeModel(),
+    installImpl: installer,
+  })
+
+  // Whatever is in that directory, it is not the declared module — so the
+  // declared module is fetched rather than assumed to be there.
+  assert.equal(installer.calls.length, 1)
+})
+
+test('a failure is carried, not thrown, and does not stop the next module', async () => {
+  const model = fakeModel()
+  const installer = {
+    calls: [],
+    async install(args) {
+      installer.calls.push(args)
+      if (args.expect.id === 'uo') throw new Error('could not reach releases.example.com')
+      return {
+        id: args.expect.id,
+        name: 'Market',
+        version: args.expect.version,
+        sha256: 'b'.repeat(64),
+        source: args.url,
+        bytes: 10,
+        replaced: false,
+      }
+    },
+  }
+
+  const results = await declared.resolve({
+    value: `uo@0.3.0=${URL_030} market@1.0.0=https://releases.example.com/market.json`,
+    hosts: HOSTS,
+    model,
+    installImpl: installer,
+  })
+
+  assert.deepEqual(results.map((r) => r.action), ['failed', 'installed'])
+  assert.match(results[0].message, /could not reach/)
+  // The second module still installed: one unreachable release host costs that
+  // module, not the deployment.
+  assert.equal(model.recorded.length, 1)
+  assert.equal(model.recorded[0].id, 'market')
+})
+
+test('a failed resolution leaves whatever was already on the volume', async () => {
+  place('uo', '0.2.0')
+  const installer = fakeInstall({ throws: 'the host is down' })
+
+  const results = await declared.resolve({
+    value: `uo@0.3.0=${URL_030}`,
+    hosts: HOSTS,
+    model: fakeModel(),
+    installImpl: installer,
+  })
+
+  assert.equal(results[0].action, 'failed')
+  // The site comes up serving the version it already had rather than not at all.
+  assert.equal(declared.installedVersion('uo'), '0.2.0')
+})
+
+test('resolution without a model records nothing and still installs', async () => {
+  // `npm run seed` and the test harness both reach code paths with no model to
+  // hand; the volume half must not depend on the database half.
+  const installer = fakeInstall()
+  const results = await declared.resolve({ value: `uo@0.3.0=${URL_030}`, hosts: HOSTS, installImpl: installer })
+  assert.deepEqual(results.map((r) => r.action), ['installed'])
+})
+
+test('state() reports the last resolution and is replaced by the next', async () => {
+  const installer = fakeInstall()
+  await declared.resolve({ value: `uo@0.3.0=${URL_030}`, hosts: HOSTS, installImpl: installer })
+  assert.equal(declared.state().length, 1)
+  assert.equal(declared.state()[0].id, 'uo')
+
+  await declared.resolve({ value: '', hosts: HOSTS, installImpl: installer })
+  assert.deepEqual(declared.state(), [])
+})
+
+// ── The whole path, once, for real ─────────────────────────────────────────
+
+const BLOCK = 512
+
+function octal(value, width) {
+  return Number(value).toString(8).padStart(width - 1, '0') + '\0'
+}
+
+function member({ name, type = '0', body = '' }) {
+  const header = Buffer.alloc(BLOCK, 0)
+  const data = Buffer.from(body, 'utf8')
+  header.write(name, 0, 100, 'utf8')
+  header.write(octal(0o644, 8), 100, 8, 'ascii')
+  header.write(octal(0, 8), 108, 8, 'ascii')
+  header.write(octal(0, 8), 116, 8, 'ascii')
+  header.write(octal(data.length, 12), 124, 12, 'ascii')
+  header.write(octal(0, 12), 136, 12, 'ascii')
+  header.write('        ', 148, 8, 'ascii')
+  header.write(type, 156, 1, 'ascii')
+  header.write('ustar\0', 257, 6, 'ascii')
+  header.write('00', 263, 2, 'ascii')
+  let sum = 0
+  for (const byte of header) sum += byte
+  header.write(`${sum.toString(8).padStart(6, '0')}\0 `, 148, 8, 'ascii')
+  const padding = Buffer.alloc((BLOCK - (data.length % BLOCK)) % BLOCK, 0)
+  return Buffer.concat([header, data, padding])
+}
+
+/** A bundle tarball, named the way module-uo's release workflow names one. */
+function bundle(version) {
+  const root = `module-uo-${version}`
+  return zlib.gzipSync(Buffer.concat([
+    member({ name: `${root}/`, type: '5' }),
+    member({
+      name: `${root}/module.json`,
+      body: JSON.stringify({ id: 'uo', name: 'Ultima Online', version, coreApi: '^1.0.0', server: 'server/index.js' }),
+    }),
+    member({ name: `${root}/server/`, type: '5' }),
+    member({ name: `${root}/server/index.js`, body: 'module.exports = () => {}\n' }),
+    Buffer.alloc(BLOCK * 2, 0),
+  ]))
+}
+
+test('end to end: a declaration installs a real bundle through the real install path', async (t) => {
+  const tarball = bundle('0.3.0')
+  const artifactUrl = 'https://releases.example.com/mod/uo-0.3.0.tar.gz'
+  const manifest = JSON.stringify({
+    schema: 1,
+    id: 'uo',
+    name: 'Ultima Online',
+    version: '0.3.0',
+    coreApi: '^1.0.0',
+    artifact: 'uo-0.3.0.tar.gz',
+    url: artifactUrl,
+    sha256: crypto.createHash('sha256').update(tarball).digest('hex'),
+    size: tarball.length,
+  })
+  const routes = { [URL_030]: manifest, [artifactUrl]: tarball }
+  const fetched = []
+  const fetchImpl = async (url) => {
+    fetched.push(String(url))
+    const body = routes[String(url)]
+    return body === undefined ? new Response('nope', { status: 404 }) : new Response(body, { status: 200 })
+  }
+
+  // The real install.js, with only the transport replaced — same seam the
+  // install tests use, and the same reason: what is under test is the decisions,
+  // not Node's TLS.
+  // eslint-disable-next-line global-require
+  const install = require('../src/modules/install')
+  const installImpl = { install: (args) => install.install({ ...args, fetchImpl }) }
+  const model = fakeModel()
+
+  const first = await declared.resolve({
+    value: `uo@0.3.0=${URL_030}`,
+    hosts: HOSTS,
+    model,
+    installImpl,
+  })
+  assert.deepEqual(first.map((r) => r.action), ['installed'])
+  // Unpacked, with the release's top-level directory stripped, under the id.
+  assert.equal(
+    JSON.parse(fs.readFileSync(path.join(tmpRoot, 'uo', 'module.json'), 'utf8')).version,
+    '0.3.0',
+  )
+  assert.equal(model.recorded[0].sha256, JSON.parse(manifest).sha256)
+
+  // And again, which is what every restart after the first one is.
+  const before = fetched.length
+  const second = await declared.resolve({ value: `uo@0.3.0=${URL_030}`, hosts: HOSTS, model, installImpl })
+  assert.deepEqual(second.map((r) => r.action), ['noop'])
+  assert.equal(fetched.length, before, 'the second resolution must not fetch anything')
+
+  // A declaration whose URL resolves to a different version is refused BEFORE
+  // the artifact is downloaded — the module on the volume is left alone.
+  const wrong = await declared.resolve({ value: `uo@0.4.0=${URL_030}`, hosts: HOSTS, model, installImpl })
+  assert.equal(wrong[0].action, 'failed')
+  assert.match(wrong[0].message, /v0\.4\.0 was asked for/)
+  assert.equal(declared.installedVersion('uo'), '0.3.0')
+  t.diagnostic(`fetched ${fetched.length} URL(s) across three resolutions`)
+})