Files
OpenMAIC/instrumentation.ts
T
AbelWangYaBoandwyuc 05d782090f feat(persistence): resolve single-tenant requests to a shared owner id (#1639)
Course documents, folders, materials and agent sessions are partitioned by an
owner id that comes from a 30-day anonymous cookie, because there is no host
auth layer. For one team behind one ACCESS_CODE that partition buys nothing —
those visitors already share the site password — and it costs: a second browser
shows an empty course list, and POST /api/stages/[id]/publish refuses every
owner, because publishing a cookie partition is not something the product
allows.

PERSISTENCE_SHARED_OWNER_ID resolves every request to that fixed id instead.
Unset — the default — changes nothing.

- **ACCESS_CODE is required.** Without one the middleware lets every request
  through, so a single shared owner would expose one readable, editable and
  publishable course library to anyone who can reach the deployment. That is
  worse than the per-browser partitioning this setting removes, so the
  combination refuses to start, from instrumentation.ts, with a message naming
  both variables. The alternative the issue allowed — ignore it with a warning —
  was rejected because a silently ignored setting is exactly the confusion this
  feature exists to remove.
- The value is validated rather than trusted: 1-128 characters of
  [A-Za-z0-9._-]. That excludes the reserved `anon:` prefix, which would still
  be refused by publish and would alias onto a cookie owner, and any character
  the material-key sanitiser rewrites, which could collide two ids onto one
  object key. A malformed value fails startup on the same path; an empty value
  is treated as unset, since KEY= in an env file cannot be told apart from an
  operator who meant to leave the feature off.
- An explicit authenticatedOwnerId still outranks it, so adding a real auth
  layer later does not require clearing the variable first.
- All four places that derive an owner honour it, including the Server Action
  in lib/workbench/workspace-actions.ts, which re-reads the cookie itself. That
  one is easy to miss and would make the workspace list and its row actions
  disagree about who owns a session.
- The env docs state the trust model next to the variable, cross-reference
  SECURITY.md, and note that this answers who owns a course while DATABASE_URL
  answers where one lives — a second browser sees an empty list on Postgres too.

This is the stopgap asked for in #1550 — a PR in the spirit of the closed
#1551 — with threading a real authenticated owner left as the long-term fix.

AI-assisted commit

Co-authored-by: wyuc <wang-yc24@mails.tsinghua.edu.cn>
2026-09-22 17:33:02 +08:00

130 lines
6.1 KiB
TypeScript

/**
* Process-scoped startup work.
*
* Next calls `register` once per server instance, before it serves a request.
* That makes it the only place in this app where a background schedule can
* live: a route module has no such guarantee — it can be instantiated more than
* once and gets no shutdown hook — so anything periodic started from one is
* really started per instantiation.
*
* `register` must return before the server is ready, so nothing here may block
* on I/O. Starting a timer does not.
*/
export async function register(): Promise<void> {
// Also invoked for the Edge runtime, which has neither `pg` nor timers we
// want; the persistence stack is Node-only.
if (process.env.NEXT_RUNTIME !== 'nodejs') return;
const { warnIfAccessCodeIsUnset } = await import('@/lib/server/access-code-warning');
warnIfAccessCodeIsUnset(process.env.ACCESS_CODE);
// The asset quota, read here rather than at the first persistence request.
// The provider that consumes it is lazy and memoised, so a malformed ceiling
// would otherwise let the process boot, pass its health check, and then fail
// every persistence request -- documents and runtime included -- until it was
// fixed and the process restarted. `register` runs before the server is
// ready, so throwing here is what makes a misconfigured deployment fail to
// start instead of failing to work. First, so the throw cannot skip the
// teardown registration for something this function has already started.
const { resolveAssetQuotaBytes } = await import('@/lib/persistence/asset-quota');
resolveAssetQuotaBytes();
// The pending-allocation window, for the same reason and at the same moment.
// Too short is worse than malformed: it silently expires allocations whose
// document write was still coming, so it must fail the process rather than
// the request that discovers it.
const { resolveAssetPendingTtlMs } = await import('@/lib/persistence/asset-pending-ttl');
resolveAssetPendingTtlMs();
// The shared owner id, for the same reason and at the same moment. A
// malformed value would otherwise boot, pass its health check, and then fail
// every owner-scoped request — and an operator who meant to share one course
// library has no way to tell from the outside that their setting was not
// accepted. `resolveSharedOwnerId` treats an empty value as unset, so this
// only rejects values that are present and unusable.
const { resolveSharedOwnerId } = await import('@/lib/server/agent-runtime/shared-owner');
resolveSharedOwnerId();
// Imported dynamically so the Edge bundle never pulls in `pg`.
const { startAssetCollectorSchedule } =
await import('@/lib/persistence/asset-collector-schedule');
const assetSchedule = startAssetCollectorSchedule();
// Warn-first boot-time validation of model routing config (MODEL_ROUTES,
// DEFAULT_MODEL, <PREFIX>_MODELS). Cheap and non-throwing: broken config
// surfaces here as [config] warnings instead of failing at request time.
// Imported dynamically so the Edge bundle never pulls in the fs/js-yaml
// backed provider config it reads.
const { validateServerConfig } = await import('@/lib/server/config-validation');
validateServerConfig();
let runner: import('@/lib/server/agent-runtime/runner').AgentRunnerHandle | undefined;
let extractionRunner:
| import('@/lib/server/material-extraction/runner').MaterialExtractionRunnerHandle
| undefined;
let stopAgentEventNotifyBus: (() => Promise<void>) | null = null;
try {
const { isAgentRuntimeConfigured } = await import('@/lib/config/feature-flags');
if (isAgentRuntimeConfigured()) {
// One dedicated LISTEN connection per application instance. The HTTP
// SSE routes and the runner share its in-process fanout registry; it is
// not a pool client and never scales with the number of streams.
const { startAgentEventNotifyBus } =
await import('@/lib/server/agent-runtime/event-notify-bus');
const eventNotifyBus = startAgentEventNotifyBus();
stopAgentEventNotifyBus = () => eventNotifyBus.stop();
// startAgentRunner only installs a timer. Store/schema initialization is
// retained behind the store's lazy promise and never blocks register().
const runtime = await import('@/lib/server/agent-runtime/runner');
runner = runtime.startAgentRunner();
const extraction = await import('@/lib/server/material-extraction/runner');
extractionRunner = extraction.startMaterialExtractionRunner();
}
} catch (error) {
console.error('[instrumentation] Agent runtime startup failed', error);
}
let shutdownPromise: Promise<void> | undefined;
const shutdown = (): Promise<void> => {
shutdownPromise ??= (async () => {
// Park sessions before any pool they use is closed. This preserves the
// last durable entry-tree checkpoint for immediate takeover.
try {
await extractionRunner?.stop();
} catch (error) {
console.error('[instrumentation] Material extraction runner drain failed', error);
}
try {
await runner?.stop();
} catch (error) {
console.error('[instrumentation] Agent runner drain failed', error);
}
try {
await stopAgentEventNotifyBus?.();
} catch (error) {
console.error('[instrumentation] Agent event notify bus drain failed', error);
}
try {
await assetSchedule?.stop();
} catch (error) {
console.error('[instrumentation] Asset collector drain failed', error);
}
const connectionString = process.env.DATABASE_URL?.trim();
if (connectionString) {
try {
const { getServerPersistenceProvider } =
await import('@/lib/persistence/server-provider');
const { pool } = await getServerPersistenceProvider(connectionString);
await pool.end();
} catch (error) {
console.error('[instrumentation] Persistence pool shutdown failed', error);
}
}
})();
return shutdownPromise;
};
process.once('SIGTERM', () => void shutdown());
process.once('SIGINT', () => void shutdown());
}