fb1cf0f6 deleted the todo router, the board and the screen but did NOT remove the
imports of them. so a checkout of that commit has hono.ts importing
`./api/todos/router`, which is not there — and the server cannot start.
fofos hit it within a minute of pulling. it was down until this commit.
WHY IT PASSED EVERY CHECK HERE: the files were modified in my working tree and
never staged, so this machine typechecked, tested and ran against the fix while the
commit shipped without it. every green light I reported was true of the tree in
front of me and false of the thing I pushed.
HOW IT HAPPENED, precisely, because the shape is worth knowing: the `git add` naming
these three also named a path that no longer existed. `git add` ABORTS ON AN
UNMATCHED PATHSPEC — it does not stage the rest and warn. It staged nothing. the
`git status` I printed straight afterwards showed them as ` M`, unstaged, and I read
the output as confirmation rather than as the warning it was.
the lesson is not "look harder at git status". it is that a commit which deletes a
module and a commit which stops importing it must be the same commit, and the only
thing that proves it is a check against the COMMIT rather than the working tree.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Officer-Session: 656f5c85-e49a-4f1f-b9f3-042a04b4d043
660 lines
35 KiB
TypeScript
660 lines
35 KiB
TypeScript
// The permission registry: the single enumeration of what this platform can do, and the unit the owner
|
|
// grants to a role.
|
|
//
|
|
// ── Why permissions and not routes ──
|
|
//
|
|
// The obvious model is "list the routes a role may call". It does not survive contact with this codebase.
|
|
// A sweep of all 100 mutating platform routes on 2026-08-06 found reads permanently stuck on POST for two
|
|
// reasons that are not going away: bodies GET cannot carry (`/tts`, `/ocr` multipart audio and images —
|
|
// `/stt` and `/transcribe` were two more of these until 2026-08-24, when speech-to-text left for the
|
|
// `transcribe` plugin), and credentials that must not sit in a query string where access logs, shell
|
|
// history and Referer headers capture them (`/tts/voices` apiKey, the four `/test` endpoints,
|
|
// `/local-providers/probe`).
|
|
// Five genuinely free conversions were done in e54d71d; the rest are staying. So the METHOD alone cannot
|
|
// carry the read/write distinction — hence `readOnlyWrites` below, declared per permission.
|
|
//
|
|
// The deeper reason is that a route list is not what the owner is deciding. The owner decides "this person
|
|
// gets Gitea". A permission is that decision; the prefixes, sockets and screens it expands to are an
|
|
// implementation detail that belongs next to the decision rather than in the granting UI.
|
|
//
|
|
// ── The four kinds, and why `execution` can never be granted ──
|
|
//
|
|
// core every authenticated account, always. Not grantable because not deniable — signing in
|
|
// without them means a broken app, not a restricted one.
|
|
// app the grantable surface. This is what the owner hands out per role.
|
|
// confined execution-shaped, but the KERNEL enforces the boundary per account. Grantable, and only
|
|
// to an account that has a Linux user — see below.
|
|
// execution NEVER grantable. Owner only, structurally.
|
|
// admin owner only: the platform administering itself, and the owner's own money and network.
|
|
//
|
|
// ── `confined`, and why it is not just `app` ──
|
|
//
|
|
// Added 2026-08-11 with per-user Linux accounts (docs/per-user-linux-accounts.md). A confined permission
|
|
// touches the filesystem or runs a process, so calling it an `app` would be a lie — but it is no longer
|
|
// the OWNER'S filesystem, because the account has its own Linux user, its own home, and the kernel refusing
|
|
// everything above it.
|
|
//
|
|
// The distinction earns its keep in one place: a grant on a confined permission means NOTHING unless the
|
|
// account actually has that Linux user. `authorize.ts` drops confined grants for an account with no
|
|
// `osUser`, so "granted but unconfined" resolves to no access rather than to the owner's home. That rule
|
|
// lives there, once, instead of in each router that would otherwise have to remember it.
|
|
//
|
|
// Moving a permission from `execution` to `confined` is therefore a claim with a test attached: every path
|
|
// it reaches must resolve its directory from the CALLER, not from HOME_DIR.
|
|
//
|
|
// `execution` is the important one. Everything under it runs as the OWNER'S OS user in the owner's home
|
|
// directory: the terminal is a real shell, chat spawns `claude` with --dangerously-skip-permissions, tasks
|
|
// run arbitrary scripts, the file browser and code editor read and write the owner's disk, the desktop is
|
|
// the owner's physical screen. Granting any of them is not a feature flag, it is co-ownership of the
|
|
// machine. There is no level of "read" that makes a shell safe, which is why these have no level at all.
|
|
// Revisit only if per-user home confinement is ever solved — and that is a project, not a checkbox.
|
|
//
|
|
// ── Read by default ──
|
|
//
|
|
// A grant carries a level, `read` or `write`. `read` permits safe methods (GET/HEAD/OPTIONS) anywhere in
|
|
// the permission, plus mutations under `personal` — sub-paths that hold the CALLER'S own data and nothing
|
|
// else. `dashboards` is the worked example: a dashboard belongs to the account that made it, so the whole
|
|
// surface is personal and a read grant is really "your own, fully". So "may a member write here" is a
|
|
// property of the endpoint, not a policy knob someone has to remember to set.
|
|
//
|
|
// Music was this note's example until 2026-08-15, when it left with `plugins/music/`. A plugin declares
|
|
// the same thing through its manifest's `readOnlyWrites` — `isRequestAllowedAtLevel` merges the two lists,
|
|
// so they are one mechanism under two names.
|
|
|
|
export type PermissionKind = 'core' | 'app' | 'confined' | 'execution' | 'admin';
|
|
|
|
export type Permission = {
|
|
/** Stable identifier. Stored in the database as the grant's subject — renaming one is a data change. */
|
|
key: string;
|
|
label: string;
|
|
description: string;
|
|
kind: PermissionKind;
|
|
/**
|
|
* Path prefixes under `/api`, written exactly as they are mounted on protectedRouter in hono.ts —
|
|
* leading slash, no `/api`. The totality check pairs these against the real mount table, so a prefix
|
|
* here that nothing mounts is as much an error as a mount nothing claims.
|
|
*/
|
|
api: string[];
|
|
/** WebSocket providers, named as in server.tsx's `handlers` map. */
|
|
ws?: string[];
|
|
/** Frontend route prefixes. Filters the dock and the app registry; never a security boundary. */
|
|
routes?: string[];
|
|
/**
|
|
* Sub-paths, relative to each `api` prefix, that a READ grant may still mutate because they hold only
|
|
* the caller's own data. Matched as a prefix after the permission's own: `/devices` on the `notify`
|
|
* permission permits `POST /api/notify/devices/123`.
|
|
*/
|
|
personal?: string[];
|
|
/**
|
|
* Sub-paths, relative to each `api` prefix, that ONLY the owner may reach — at any level, by any
|
|
* method, however the permission is granted.
|
|
*
|
|
* For the administrative corner of an otherwise shareable surface. Jellyfin is the case: the plugin is
|
|
* exactly what a Member should have, and `/_config` — which stores a server's admin token and drives
|
|
* account provisioning — is exactly what they should not. Without this the only options were to grant
|
|
* the whole plugin or none of it.
|
|
*
|
|
* ── Why this is not a kind, and not `ownerOnly` on a permission ──
|
|
*
|
|
* `plugins/manifest.ts` refuses an `ownerOnly` PERMISSION, and that refusal still stands: a permission
|
|
* is a thing the owner delegates, and one nobody can delegate is a contradiction. This is narrower and
|
|
* different in kind — the permission stays fully grantable, and a few paths inside it are marked as
|
|
* administering the plugin rather than using it.
|
|
*
|
|
* ── Why it is declared here rather than enforced by each plugin ──
|
|
*
|
|
* A plugin could check this itself, and the first one would. The tenth would forget a route and nothing
|
|
* would notice — no boot failure, no warning, no test, exactly the shape `assertPermissionTotality`
|
|
* exists to prevent. Declared here it is enforced once, for every plugin, on the door they all pass
|
|
* through, and a route added later by someone who never read this is covered by default.
|
|
*
|
|
* Beaten by nothing. Checked before `selfService`, so a path that is somehow both is owner-only —
|
|
* fail closed, since the two together are a contradiction and the safe reading has to win.
|
|
*/
|
|
ownerOnly?: string[];
|
|
/**
|
|
* Reads that must stay POST — see the note at the top. A read grant permits these paths at any method.
|
|
* Written relative to the permission's `api` prefix, like `personal`.
|
|
*/
|
|
readOnlyWrites?: string[];
|
|
/**
|
|
* Routes any authenticated account may call even holding NO grant on this permission, because they act
|
|
* on the caller themselves. `METHOD /exact/path`, relative to the permission's prefix — exact, not a
|
|
* prefix, so this cannot widen by accident.
|
|
*
|
|
* One entry exists and it should stay that way. `PUT /api/users` is self-profile update (useAuth.ts
|
|
* calls it to change your own name and avatar) and has always lived on the same router as the owner-only
|
|
* account administration beside it. Moving it to `/api/user` would be tidier and would break every
|
|
* shipped mobile client, so the honest fix is to say out loud that this one route is not what the
|
|
* permission around it is.
|
|
*/
|
|
selfService?: string[];
|
|
};
|
|
|
|
/**
|
|
* The PLATFORM's own permissions. A plugin's are added on top at runtime — see `setPluginPermissions`.
|
|
*
|
|
* Kept separate from the live `PERMISSIONS` below so two invariants hold by construction rather than by
|
|
* anyone remembering them: a plugin can never become `core` (granted to everyone, undeniable), and a
|
|
* plugin is never in the fresh-install baseline.
|
|
*/
|
|
const CORE_REGISTRY: Permission[] = [
|
|
// ── core ────────────────────────────────────────────────────────────────────────────────────────
|
|
{
|
|
key: 'account',
|
|
label: 'Account',
|
|
description: 'Sign in, your own profile, password, preferences and API keys',
|
|
kind: 'core',
|
|
// `/api-keys` is core rather than app or admin because a key is not new authority — it is a second way
|
|
// to present the authority the account already has, so denying it would only force the holder to keep
|
|
// using a password in places a password should not go. What a key can then DO is decided by the same
|
|
// permission checks as any other request from that user; nothing here widens them.
|
|
// `/theme-favorites` sits here rather than on `/themes`, which is unauthenticated so the sign-in screen
|
|
// can paint. A starred theme is the same shape as a dock preference: the caller's own taste, granting
|
|
// nothing, and not a sensible thing to deny an account that can already choose a theme.
|
|
api: ['/user', '/dock', '/api-keys', '/theme-favorites'],
|
|
routes: ['/settings/profile'],
|
|
},
|
|
{
|
|
key: 'themes',
|
|
label: 'Themes',
|
|
description: 'Read the themes installed on this machine',
|
|
// `core`, and not deniable. Themes are how the interface looks; an account that could not read the
|
|
// list would render unstyled, which is not a permission decision anyone means to make. It is read-only
|
|
// — installing a theme is a directory on disk, not a request.
|
|
kind: 'core',
|
|
api: ['/themes'],
|
|
routes: [],
|
|
},
|
|
{
|
|
key: 'bug-report',
|
|
label: 'Report a problem',
|
|
description: 'Send the server owner a bug report',
|
|
kind: 'core',
|
|
api: ['/bug-report'],
|
|
},
|
|
|
|
// ── app: the grantable surface ──────────────────────────────────────────────────────────────────
|
|
{
|
|
key: 'memos',
|
|
label: 'Memos',
|
|
description: 'Notes',
|
|
kind: 'app',
|
|
// api: ['/memos'], // plugin — switched off 2026-08-13
|
|
api: [],
|
|
routes: ['/memos'],
|
|
},
|
|
{
|
|
key: 'notify',
|
|
label: 'Notifications',
|
|
description: 'Push notifications to your devices',
|
|
/*
|
|
CORE, not `app`, since 2026-08-24.
|
|
|
|
It was grantable and nothing granted it, so every Member's app got a 403 the moment it tried to
|
|
register its push token — silently, at launch, with no notification ever arriving and nothing
|
|
anywhere saying why. Invisible for days because the owner bypasses the check.
|
|
|
|
Making it grantable was the mistake rather than the missing grant. These are the caller's OWN
|
|
devices, and the send route resolves its target from the authenticated header rather than the body
|
|
(`sidecar/notify/resolve-user.ts`), so an account can only ever reach its own phone. There is no
|
|
version of "you may not subscribe to your own notifications" that is a coherent thing to configure.
|
|
*/
|
|
kind: 'core',
|
|
api: ['/notify'],
|
|
// Device registration is the caller's own — a phone subscribing to its own push channel.
|
|
//
|
|
// The `_officer/` segment is NOT decoration: these are the sidecar's paths, and `personal` is matched
|
|
// against what remains after the permission's own prefix is stripped. Written as `/devices` it matched
|
|
// nothing, so a read-level member could list their phones but not register or delete one — the exact
|
|
// opposite of the line above. Invisible in testing because the owner bypasses every check.
|
|
personal: ['/_officer/devices', '/_officer/subscriptions'],
|
|
/*
|
|
Broadcast is the one part of a core permission that is not the caller's own.
|
|
|
|
Everything else here acts on the account asking — its devices, its notifications — which is what
|
|
made `core` right. Sending to EVERY account is administration, so it needs the owner even though
|
|
the permission around it is not deniable.
|
|
*/
|
|
ownerOnly: ['/broadcast', '/user'],
|
|
},
|
|
// `vpn` (POST /api/vpn/enroll) was here until 2026-08-14 — the one member-grantable piece of headscale,
|
|
// minting a pre-auth key bound to the caller. Deleted because it had no caller anywhere: the standalone
|
|
// OffScale app gates it on `embedded`, which it never sets, and it never will — the app is permanently
|
|
// independent of the platform, since the thing that gets you to the platform cannot itself need it.
|
|
//
|
|
// Device enrolment did not go away, it moved out. A phone claims an invite from the Companion at
|
|
// `${invite.base}/api/v1/enroll/claim`, which is a different component entirely and does not involve
|
|
// Officer. Confirmed against the mobile monorepo and the companion repo before removal.
|
|
//
|
|
// What this cost: the tailnet's own permission was `admin`, so with `vpn` gone no member-grantable
|
|
// tailnet surface remained. That permission has since left with the offscale plugin, which declares its
|
|
// own — so a member-facing enrolment surface would be the PLUGIN's to add, not this file's.
|
|
// Core, not app — and this was a real defect, not a preference. `/api/dashboards` is not a feature, it is
|
|
// the per-user key-value store where EVERY workspace screen keeps its layout (`screens/files`,
|
|
// `ws-layout-*`, panel config). `WorkspaceView` renders nothing until that store has loaded, so gating it
|
|
// meant a member with `files` granted got a completely blank Files screen and no request to
|
|
// /api/file-browser at all — the panel never mounted. Same for Terminal, Chat and every other screen.
|
|
//
|
|
// It is entirely `personal` and always was: every row is keyed to the caller. There is nothing here to
|
|
// withhold, and withholding it does not restrict an account, it breaks it — which is exactly the
|
|
// definition of `core` at the top of this file.
|
|
{
|
|
key: 'dashboards',
|
|
label: 'Screen layouts and dashboards',
|
|
description: 'Where your own screen layouts and dashboards are saved',
|
|
kind: 'core',
|
|
api: ['/dashboards'],
|
|
routes: ['/dashboards'],
|
|
personal: ['/'],
|
|
},
|
|
// Core because there is nothing to grant a level of: a commander is the caller's own session, in the
|
|
// caller's own directory, in a table scoped to the caller's own user_id at every query. Withholding
|
|
// this from a role would not restrict what they can reach, it would just hide their own conversations
|
|
// from a second list of the same conversations — the same reasoning `dashboards` above rests on.
|
|
{
|
|
key: 'command-hq',
|
|
label: 'Command HQ',
|
|
description: 'Long-running commanders you started, organised into categories',
|
|
kind: 'core',
|
|
api: ['/command-hq'],
|
|
routes: ['/command-hq'],
|
|
personal: ['/'],
|
|
},
|
|
|
|
// ── execution: never grantable ──────────────────────────────────────────────────────────────────
|
|
// Confined since 2026-08-11. A member's shell is spawned by the pty sidecar through `sudo setpriv` as their
|
|
// own Linux account, in their own home, with the platform's environment cleared — so it is their shell, and
|
|
// the kernel decides what it can reach. The sidecar also records whose each session is, so `list` and `kill`
|
|
// scope to the caller instead of every shell on the box.
|
|
//
|
|
// What this is NOT is a jail. A member with a shell can `cd /` and read whatever the system leaves
|
|
// world-readable, like any account on any machine. It isolates members from each other and from the owner's
|
|
// files, which is the promise `confined` makes.
|
|
{
|
|
key: 'terminal',
|
|
label: 'Terminal',
|
|
description: 'A shell on this machine, as your own user',
|
|
kind: 'confined',
|
|
api: ['/terminal'],
|
|
ws: ['terminal'],
|
|
routes: ['/terminal'],
|
|
},
|
|
// Confined, and true since 2026-08-12: a member's turn spawns their own `claude` as their own Linux account
|
|
// through `sudo setpriv`, with their own `~/.claude` credential, their own transcripts and sessions that
|
|
// record whose they are. The two owner-only refusals that held this open — `api/chat/chat.ts` and the socket
|
|
// in `server.tsx` — were removed together once each of those was in place.
|
|
{
|
|
key: 'chat',
|
|
label: 'Chat',
|
|
description: 'The agent',
|
|
kind: 'confined',
|
|
// `/agent-status` is on this grant but deliberately NOT on `chatRouter`, because that router refuses
|
|
// non-owners wholesale — so an endpoint there could not be read by the accounts that need it. It reports
|
|
// two booleans about the caller's own home and nothing about anyone else, which is what lets a member be
|
|
// told "run `claude` once to sign in" instead of being shown a tile that 403s.
|
|
// `/claude-auth` is the `/login` flow — it spawns `claude auth login` AS THE CALLER and writes the
|
|
// credential into the caller's own home. It rides this permission rather than earning its own for the
|
|
// reason that makes the pairing exact: an account that may use the agent must be able to authenticate
|
|
// the agent, and an account that may not has nothing to authenticate. `confined` then carries the rest
|
|
// for free — `authorize.ts` drops a confined grant for an account with no `osUser`, which is precisely
|
|
// the account that cannot hold credentials of its own.
|
|
//
|
|
// NOT under `/auth`, which `totality.ts` exempts from the gate entirely because it is where a caller
|
|
// goes to BECOME authenticated. Mounted there, an unauthenticated request could start a login as the
|
|
// owner.
|
|
api: ['/chat', '/agent-status', '/claude-auth'],
|
|
// 'chat-notify' is a session-independent socket (the bell feature): it carries no sessionId and no
|
|
// transcript, just a ping that a chat this account can already reach has finished a turn or needs
|
|
// input. Riding this permission rather than a new one, because it is meaningless without `chat`
|
|
// already granted.
|
|
ws: ['chat', 'chat-notify'],
|
|
// `/bots` is a chat with a context, not a second feature: a bot IS a conversation whose working
|
|
// directory is `bots/`, armed through the background job any chat can arm. It earns no permission of
|
|
// its own because it adds no surface — the same routes, the same socket, the same confinement.
|
|
routes: ['/chat', '/bots'],
|
|
},
|
|
// Confined rather than execution since 2026-08-11. Every path under `/file-browser` resolves its root
|
|
// through `resolveHomeDir(userId)` in a middleware that refuses the request outright when the account has
|
|
// no Linux user — so a member sees their own home and `resolveUserPath`'s containment check stops them
|
|
// walking out of it. `/upload` was already per-caller: it writes only under
|
|
// `DATA_PATH/<email>/attachments`, never into a home.
|
|
{
|
|
key: 'files',
|
|
label: 'Files',
|
|
description: 'Your own home directory on this machine, and the code editor over it',
|
|
kind: 'confined',
|
|
api: ['/file-browser', '/upload'],
|
|
routes: ['/files', '/code-editor'],
|
|
// `POST /file-browser/download` zips a multi-selection and streams it back. It writes nothing — it is
|
|
// a POST only because a list of paths does not fit a query string. Without this a read-level member
|
|
// could download a whole folder (GET /download) but not two files, which is not a rule anyone chose.
|
|
//
|
|
// `isPrefixOf` is segment-aware, so this does NOT also permit `/download-video` — which really is a
|
|
// write, fetching a video onto the disk.
|
|
readOnlyWrites: ['/download'],
|
|
},
|
|
{
|
|
key: 'tasks',
|
|
label: 'Tasks and jobs',
|
|
description: 'Running permissions, pipelines and background jobs',
|
|
kind: 'execution',
|
|
api: ['/tasks', '/jobs', '/pipeline-jobs', '/queue'],
|
|
ws: ['task-runner', 'pipeline'],
|
|
routes: ['/jobs'],
|
|
},
|
|
/*
|
|
SendOff is `app`, not `confined`, and the distinction is the point.
|
|
|
|
It writes into a home directory, which normally means `confined` — but it writes exactly one shape of
|
|
path, `Documents/SendOff/<sender>/`, chosen by the server from the caller's token. There is no path in
|
|
the request to contain, nothing to traverse out of, and no way to read anything back. So the Linux-user
|
|
requirement that `confined` exists to enforce buys nothing here, and requiring it would mean an account
|
|
with no OS user could not be sent a file — which is most of the point of the feature.
|
|
|
|
WHO may send to whom is not decided here. It is a row per pair held by the recipient, because the
|
|
person who should decide whether somebody may put a file in your home is you, not whoever administers
|
|
roles. This grant only says the feature is reachable at all.
|
|
*/
|
|
/*
|
|
`confined`, and for the same reason chat is.
|
|
|
|
Spawning an agent runs a process in a home directory — the caller's own, resolved by `chatIdentity`
|
|
from their token. So it needs exactly what chat needs: a Linux account for a member, and nothing for
|
|
the owner, with the kernel doing the containing. `authorize.ts` drops a confined grant for an account
|
|
whose `osUser` is null, which is what stops "granted but unconfined" resolving to the owner's home.
|
|
|
|
Not `execution`: that kind exists for surfaces that can ONLY ever run as the owner. This one runs as
|
|
whoever called it, which is the whole distinction.
|
|
*/
|
|
{
|
|
key: 'sub-agents',
|
|
label: 'Sub-agents',
|
|
description: 'Start another agent from a prompt, subject to the account budget',
|
|
kind: 'confined',
|
|
api: ['/sub-agents'],
|
|
},
|
|
{
|
|
key: 'sendoff',
|
|
label: 'SendOff',
|
|
description: 'Send files to your own machines, and to accounts that have allowed you',
|
|
kind: 'app',
|
|
api: ['/sendoff'],
|
|
routes: ['/sendoff'],
|
|
},
|
|
{
|
|
key: 'items',
|
|
label: 'Permission authoring',
|
|
description: 'Skills, tools, agents and processes on disk',
|
|
kind: 'execution',
|
|
api: ['/skills', '/tools', '/agents', '/processes', '/rescan'],
|
|
},
|
|
// `desktop` left with the plugin on 2026-08-21. It is the first permission to make the round trip from
|
|
// `kind: 'execution'` to a plugin's own, and the property survived the move: a plugin permission has no
|
|
// kind, so "runs as the owner and cannot be shared" is expressed as `ownerOnly: ['/']` in the plugin's
|
|
// manifest instead. Same answer, stated where the paths are.
|
|
// The relay left for `plugins/browser` on 2026-08-21, taking `/browser` and its frontend route with it.
|
|
// What stays is `/scrape`, and the two were never the same thing wearing one word: scrape launches a
|
|
// headless chromium through playwright as the OWNER and knows nobody, which is why this is still
|
|
// `execution`. The plugin attaches to whoever is holding the keyboard, scoped per pairing, and is
|
|
// grantable for exactly that reason.
|
|
{
|
|
key: 'browser',
|
|
label: 'Scraping',
|
|
description: 'Fetches and renders a page in a headless browser, as the server owner',
|
|
kind: 'execution',
|
|
api: ['/scrape'],
|
|
},
|
|
|
|
// ── admin: the platform administering itself ────────────────────────────────────────────────────
|
|
{
|
|
key: 'app-store',
|
|
label: 'App store',
|
|
description: 'Install, enable and remove the sidecars this server runs',
|
|
// Admin, not app. Installing a sidecar starts a process on the machine and provisioning one starts
|
|
// containers — that is process control, not a feature a member can be granted a read of. The router
|
|
// gates on the owner in its own right as well; this entry is what makes the boot check pass and what
|
|
// keeps the surface visible in one enumeration.
|
|
kind: 'admin',
|
|
api: ['/app-store'],
|
|
routes: ['/app-store'],
|
|
},
|
|
{
|
|
key: 'plugins',
|
|
label: 'Plugins',
|
|
description: 'Install, enable and remove the plugins this server runs',
|
|
// Admin for the same reason as the app store above: installing a plugin mounts routes and starts a
|
|
// process, which is process control rather than a feature to grant a read of.
|
|
//
|
|
// Note this permission guards the MANAGEMENT surface, not the plugins themselves. A plugin declares
|
|
// its own permissions in its manifest, and those are what gate its routes — so a member can hold
|
|
// `offscale` at read without being able to install or remove anything.
|
|
kind: 'admin',
|
|
api: ['/plugins'],
|
|
// The API kept its name; only the screen was renamed. `/plugins` stays listed because the redirect
|
|
// is a real route — a client-side redirect still has to resolve before it can send anyone anywhere.
|
|
routes: ['/marketplace', '/plugins'],
|
|
},
|
|
{
|
|
key: 'server-admin',
|
|
label: 'Server settings',
|
|
description: 'Server configuration, integrations and the activity log',
|
|
kind: 'admin',
|
|
api: ['/server-settings', '/integrations', '/activity', '/system-monitor'],
|
|
routes: ['/settings/server', '/settings/system', '/activity', '/system-monitor'],
|
|
},
|
|
{
|
|
key: 'user-admin',
|
|
label: 'User management',
|
|
description: 'Accounts, roles and what each role may reach',
|
|
kind: 'admin',
|
|
api: ['/users'],
|
|
routes: ['/settings/user-management'],
|
|
// Changing your own name, username and avatar. Owner-only account administration is every other route
|
|
// on this router and stays owner-only — see ownerGate in users-router.ts, which is the second lock.
|
|
selfService: ['PUT /'],
|
|
},
|
|
// `headscale` lived here until 2026-08-15, when it left with the rest of offscale. A plugin declares
|
|
// its own permissions in its manifest and they are registered at install — see plugins/mount.ts. The
|
|
// platform no longer knows this permission exists, which is the entire point.
|
|
{
|
|
key: 'roles',
|
|
label: 'Roles',
|
|
description: 'Create, rename and delete the roles this server grants',
|
|
// Admin, and for the same reason `user-admin` beside it is: a role is not a member's own data, it is
|
|
// policy that governs every account holding it, so there is no level of "read" that makes creating or
|
|
// deleting one a feature to grant. `registry.test.ts` asserts admin can never be handed out — this
|
|
// entry is what makes that assertion mean something for roles specifically.
|
|
//
|
|
// Separate from `user-admin` rather than folded into it: that permission's `/users` is the account
|
|
// list and the grants a role holds (`/users/permissions`); this one is the roles TABLE itself — the
|
|
// name, the description, whether it is system-named. Two routers, two questions, so a boot-time check
|
|
// that a mounted prefix has exactly one owner stays meaningful for both.
|
|
kind: 'admin',
|
|
api: ['/roles'],
|
|
},
|
|
];
|
|
|
|
// ── Derived lookups ───────────────────────────────────────────────────────────────────────────────
|
|
//
|
|
// `let`, not `const`, because installing a plugin adds permissions and uninstalling removes them. ESM
|
|
// exports are live bindings, so an importer holding `PERMISSION_BY_KEY` sees the reassignment — every
|
|
// consumer reads these inside a function, never at module scope, which is what makes that safe.
|
|
//
|
|
// Nothing mutates the arrays in place. `setPluginPermissions` replaces them, for the same reason
|
|
// `buildHonoApp` replaces the app rather than adding routes to it: a half-updated lookup is worse than a
|
|
// stale one, and replacement has no half.
|
|
|
|
/** Core plus every installed plugin's. The list `permissionForApiPath` and the totality check read. */
|
|
export let PERMISSIONS: Permission[] = CORE_REGISTRY;
|
|
|
|
export let PERMISSION_BY_KEY = new Map(PERMISSIONS.map((c) => [c.key, c]));
|
|
|
|
/**
|
|
* The keys an owner may actually hand to a role. `core` is automatic; `execution` and `admin` are owner-only.
|
|
*
|
|
* `confined` is offered here, but a grant on one is inert for an account without a Linux user — that is
|
|
* enforced in `authorize.ts`, not by withholding it from this list. Withholding it would mean the owner
|
|
* could not pre-grant a role before provisioning the people in it, which is the normal order of operations.
|
|
*/
|
|
export let GRANTABLE_PERMISSIONS = PERMISSIONS.filter((c) => c.kind === 'app' || c.kind === 'confined');
|
|
|
|
/**
|
|
* What every role starts with on a fresh install: the three confined permissions, at `write`.
|
|
*
|
|
* These are the baseline the platform is FOR — a terminal, a file browser and chat. An account that can sign
|
|
* in and reach none of them is not a restricted account, it is a useless one, and making the owner grant them
|
|
* by hand before anyone can do anything is a step with no decision in it.
|
|
*
|
|
* Seeded as real rows rather than implied by absence, which keeps the table's one rule intact: a missing row
|
|
* means no access, always, with no exceptions to remember. So revoking one of these works exactly like
|
|
* revoking anything else — the row goes, and nothing puts it back.
|
|
*
|
|
* `app` permissions are deliberately NOT here. Those reach data the owner may not intend to share, and each
|
|
* needs a sidecar installed before it means anything anyway.
|
|
*/
|
|
// From CORE_REGISTRY, never from PERMISSIONS: a plugin must not be able to put itself in the baseline
|
|
// every new role starts with.
|
|
export const DEFAULT_ROLE_PERMISSIONS: string[] = CORE_REGISTRY.filter((c) => c.kind === 'confined').map((c) => c.key);
|
|
|
|
/**
|
|
* Available to every signed-in account without a grant.
|
|
*
|
|
* From CORE_REGISTRY, so a plugin declaring `kind: 'core'` — which its manifest cannot express, but which
|
|
* a future bug could smuggle in — still could not grant itself to everyone undeniably.
|
|
*/
|
|
export const CORE_PERMISSIONS = CORE_REGISTRY.filter((c) => c.kind === 'core');
|
|
|
|
/**
|
|
* Replace the plugin half of the registry. Called after every install, uninstall, enable and disable.
|
|
*
|
|
* Plugin permissions are dropped and re-added wholesale rather than diffed: the source of truth is the
|
|
* set of installed plugins, and computing a delta against it would be a second answer to the same
|
|
* question. A key colliding with a core one is REFUSED here rather than silently overriding — a plugin
|
|
* that could redefine `chat` or `terminal` could widen it.
|
|
*/
|
|
/**
|
|
* Which of these keys the PLATFORM already claims.
|
|
*
|
|
* Exported so install can refuse before it starts, rather than leaving `setPluginPermissions` to reject
|
|
* the key afterwards — that rejection is real but it happens at mount time and only reaches stderr, so
|
|
* the install reports success, the dock tile appears, and the route 404s. That exact sequence caught
|
|
* photos, soulseek and jellyfin on the same machine within a day: each had been a CORE permission until
|
|
* the extraction commit, so any server whose platform predates it still claims the key.
|
|
*/
|
|
export function corePermissionCollisions(keys: string[]): string[] {
|
|
const coreKeys = new Set(CORE_REGISTRY.map((c) => c.key));
|
|
return keys.filter((k) => coreKeys.has(k));
|
|
}
|
|
|
|
export function setPluginPermissions(pluginPermissions: Permission[]): { rejected: string[] } {
|
|
const coreKeys = new Set(CORE_REGISTRY.map((c) => c.key));
|
|
const seen = new Set<string>();
|
|
const accepted: Permission[] = [];
|
|
const rejected: string[] = [];
|
|
|
|
for (const permission of pluginPermissions) {
|
|
if (coreKeys.has(permission.key) || seen.has(permission.key)) {
|
|
rejected.push(permission.key);
|
|
continue;
|
|
}
|
|
seen.add(permission.key);
|
|
accepted.push(permission);
|
|
}
|
|
|
|
PERMISSIONS = [...CORE_REGISTRY, ...accepted];
|
|
PERMISSION_BY_KEY = new Map(PERMISSIONS.map((c) => [c.key, c]));
|
|
GRANTABLE_PERMISSIONS = PERMISSIONS.filter((c) => c.kind === 'app' || c.kind === 'confined');
|
|
return { rejected };
|
|
}
|
|
|
|
export type PermissionLevel = 'read' | 'write';
|
|
|
|
const isPrefixOf = (prefix: string, path: string): boolean =>
|
|
prefix === '/' || path === prefix || path.startsWith(`${prefix}/`);
|
|
|
|
/**
|
|
* Which permission owns this path? `path` is the full request path (`/api/gitea/...`).
|
|
*
|
|
* Longest prefix wins, so a permission may claim `/dav` while another claims `/dav/something` without the
|
|
* order of the array mattering. Returns null for a path no permission claims — which the totality check
|
|
* below is there to make impossible for anything mounted on protectedRouter.
|
|
*/
|
|
export function permissionForApiPath(path: string): Permission | null {
|
|
const rest = path.startsWith('/api') ? path.slice('/api'.length) : path;
|
|
let best: Permission | null = null;
|
|
let bestLength = -1;
|
|
for (const permission of PERMISSIONS) {
|
|
for (const prefix of permission.api) {
|
|
if (isPrefixOf(prefix, rest) && prefix.length > bestLength) {
|
|
best = permission;
|
|
bestLength = prefix.length;
|
|
}
|
|
}
|
|
}
|
|
return best;
|
|
}
|
|
|
|
export function permissionForWsProvider(provider: string): Permission | null {
|
|
return PERMISSIONS.find((c) => c.ws?.includes(provider)) ?? null;
|
|
}
|
|
|
|
/**
|
|
* Is this path inside one of the permission's owner-only sub-paths?
|
|
*
|
|
* Prefix-matched, like `personal` — `/_config` covers `/_config/3/users`. A settings surface is a tree,
|
|
* and enumerating its leaves is how one gets missed.
|
|
*/
|
|
export function isOwnerOnlyRoute(permission: Permission, path: string): boolean {
|
|
if (!permission.ownerOnly?.length) return false;
|
|
const rest = path.startsWith('/api') ? path.slice('/api'.length) : path;
|
|
return permission.api.some((prefix) => {
|
|
if (!isPrefixOf(prefix, rest)) return false;
|
|
const sub = rest.slice(prefix.length) || '/';
|
|
return permission.ownerOnly!.some((entry) => isPrefixOf(entry, sub));
|
|
});
|
|
}
|
|
|
|
/**
|
|
* Is this an exact self-service route — one any authenticated account may call without holding the
|
|
* permission at all? Matched exactly on method AND path, never as a prefix.
|
|
*/
|
|
export function isSelfServiceRoute(permission: Permission, method: string, path: string): boolean {
|
|
if (!permission.selfService?.length) return false;
|
|
const rest = path.startsWith('/api') ? path.slice('/api'.length) : path;
|
|
const upper = method.toUpperCase();
|
|
return permission.api.some((prefix) => {
|
|
if (!isPrefixOf(prefix, rest)) return false;
|
|
const sub = rest.slice(prefix.length) || '/';
|
|
return permission.selfService!.includes(`${upper} ${sub}`);
|
|
});
|
|
}
|
|
|
|
const SAFE_METHODS = new Set(['GET', 'HEAD', 'OPTIONS']);
|
|
|
|
/**
|
|
* May a caller holding `level` on `permission` make this request?
|
|
*
|
|
* `write` is unconditional within the permission. `read` permits safe methods, anything the permission
|
|
* declares as a `readOnlyWrites` read-in-POST-clothing, and mutations confined to `personal` sub-paths.
|
|
*/
|
|
export function isRequestAllowedAtLevel(
|
|
permission: Permission,
|
|
level: PermissionLevel,
|
|
method: string,
|
|
path: string,
|
|
): boolean {
|
|
if (level === 'write') return true;
|
|
if (SAFE_METHODS.has(method.toUpperCase())) return true;
|
|
|
|
const rest = path.startsWith('/api') ? path.slice('/api'.length) : path;
|
|
// Strip whichever of the permission's own prefixes matched, so `personal` entries are written relative
|
|
// to the permission rather than repeated per prefix.
|
|
const withinPermission = permission.api
|
|
.filter((prefix) => isPrefixOf(prefix, rest))
|
|
.map((prefix) => rest.slice(prefix.length) || '/');
|
|
|
|
const allowed = [...(permission.personal ?? []), ...(permission.readOnlyWrites ?? [])];
|
|
return withinPermission.some((sub) => allowed.some((entry) => isPrefixOf(entry, sub)));
|
|
}
|