Skip to content

Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS

fix: honor cancellation while cloud renders prepare (#12052)

Adam Eivy819b4f0f85f5
diff --git a/server/lib/README.md b/server/lib/README.md
index d7d7a5d355..e0fb3f7777 100644
--- a/server/lib/README.md
+++ b/server/lib/README.md
@@ -44,6 +44,7 @@ The barrel `server/lib/index.js` is a machine-checkable enumeration of every pub
| `taskDataInputCatalog.js` | Pure catalog and persisted ids for deterministic context sources that scheduled agent tasks can preload before dispatch. |
| `sharedSchemas.js` | Cross-domain Zod fragments that `validation.js` and the per-domain `*Validation.js` files both need, kept in a leaf module so the domain files never import back through `validation.js` (its hoisted `export * from` lines would TDZ). `grokVideoDurationSchema` (grok clip-length union), `cloudModelIdString(message)` (cloud-CLI model-id charset/bounds), `recordRenderPinFields` (the `imageMode`/`imageModelId` per-record render pin pair), and `isSafeSubdirFilter(v)` (relative path with no wildcard, `..` segment, or leading `/`), and `csvIdsParam({ max, maxIdLength, truncate })` (the `?ids=a,b,c` batch-by-id query param: trims, drops empties, reads all-blank as absent, and either 400s or silently slices an over-cap batch). Also the backup-exclude anchoring rule (#7241): `anchorUserExclude` / `anchorUserExcludes` prepend the rsync root anchor (`/`) to any user pattern that is not already anchored and is not deliberately `*`/`**`-led, and `isSafeExcludePattern` (defined as `anchorUserExclude(x) !== null`, so the two can never disagree) rejects a `..` segment, a NUL, or a pattern longer than `EXCLUDE_PATTERN_MAX_LENGTH` once anchored — one rule shared by `backupConfigSchema` and `services/backup.js#computeEffectiveExcludes`, since an unanchored `cache/` is an rsync filter that matches at EVERY depth. Mirrored to `client/src/lib/backupExcludes.js`. Also `resolveAutoUpdateConfig(settings.autoUpdate)` + the channel/interval constants — same sparse-slice rule for the automatic self-update: omitted `enabled` means OFF, omitted `channel` means `release`, `minIntervalHours` clamps into `[1, 720]`, `resolveBlockersWithAgent` defaults on — shared by `autoUpdateSettingsSchema`, `services/autoUpdateScheduler.js`, and the settings GET projection. |
| `generationModes.js` | Shared immutable image/video render-backend alphabets (`IMAGE_GEN_MODE[S]`, `QUEUEABLE_IMAGE_MODES`, `VIDEO_GEN_MODE[S]`, and cloud subsets) plus `mediaJobExecutionLane({ kind, mode, remote })` / `MEDIA_JOB_EXECUTION_LANES` — THE media-queue lane rule the scheduler and the public job projection share, kept below both validation and generation services so schemas never import orchestration modules. |
+| `mediaJobPreparation.js` | `createMediaJobPreparation()` tracks per-job cancellation before cloud providers install a render handle; `run`, `cancel`, `take`, and `ids` cover preparation, handoff, and bulk cancellation. |
| `imageGenCapabilities.js` | Per-backend image-gen capability literals, a dependency-free leaf beside `generationModes.js`: `EDIT_INCAPABLE_IMAGE_MODES`/`isEditCapableMode` and the derived `I2I_CAPABLE_MODES`, the shipped defaults (`CODEX_IMAGEGEN_DEFAULT_MODEL`/`_EFFORT`, `AGY_IMAGEGEN_DEFAULT_MODEL`, `AGY_IMAGEGEN_IMAGE_MODEL`), the `AGY_`/`GROK_ASPECT_RATIOS` alphabets, and `IMAGE_GEN_PROVIDER_CAPABILITIES` (`supportsModelOverride`, `maxInputImages`, `promptRequiredWithInputImage`) with its `supportsCloudModelOverride`/`maxInputImages`/`cloudPromptRequired` readers. The `services/imageGen/*` modules re-export these, and `client/src/lib/imageGenModes.js` imports them instead of copying them (#6590). |
| `falImageModels.js` | Curated fal.ai image model catalog, a dependency-free leaf the client imports: `FAL_IMAGE_FAMILIES` (text + `/edit` endpoint per family, per-model reference cap, sizing alphabet, pricing), `FAL_IMAGE_DEFAULT_MODEL`, `FAL_IMAGE_MODEL_IDS` (the only ids a fal `cloudModel` may name), `falImageFamily`/`isFalImageModel`, `falImagePriceLabel`, `buildFalImageRequest` (routes to `/edit` when input images are present; maps width/height to aspect-ratio + resolution tier or pixel `image_size`) and `estimateFalImageCostUsd` (with `FAL_MEGAPIXEL`, fal.ai's binary megapixel). |
| `imageRuntimeRemedies.js` | Local image-runtime vocabulary, a dependency-free leaf beside `imageGenCapabilities.js`: `IMAGE_RUNTIME_READINESS` (ready/unavailable/unknown), `IMAGE_RUNTIME` + `IMAGE_RUNTIME_LABELS` (the shared torch venv vs. the configurable mflux interpreter), `IMAGE_RUNTIME_REMEDY` (the one action that fixes each unavailable state), and `IMAGE_RUNTIME_UNAVAILABLE_CODE`. `services/imageGen/localRuntime.js` builds the diagnosis from these, `services/imageGen/local.js` tags its pre-flight render refusal with the same remedy, and `client/src/lib/imageGenModes.js` re-exports them so no component hand-copies a remedy kind. |
diff --git a/server/lib/index.js b/server/lib/index.js
index 7c1a3bc8b6..7f48508365 100644
--- a/server/lib/index.js
+++ b/server/lib/index.js
@@ -897,3 +897,5 @@ export * from './musicVideoImageEffects.js';
export * from './securityScanEvidence.js';
export * from './musicDesignerValidation.js';
+
+export * from './mediaJobPreparation.js';
diff --git a/server/lib/mediaJobPreparation.js b/server/lib/mediaJobPreparation.js
new file mode 100644
index 0000000000..f3cd12933f
--- /dev/null
+++ b/server/lib/mediaJobPreparation.js
@@ -0,0 +1,32 @@
+import { randomUUID } from 'node:crypto';
+
+/** Track per-job cancellation until a provider installs its render handle. */
+export function createMediaJobPreparation() {
+ const preparing = new Map();
+ return {
+ async run(params, generate) {
+ const jobId = params.jobId || randomUUID();
+ const entry = { aborted: false };
+ preparing.set(jobId, entry);
+ try {
+ return await generate({ ...params, jobId });
+ } finally {
+ if (preparing.get(jobId) === entry) preparing.delete(jobId);
+ }
+ },
+ cancel(jobId) {
+ const entry = preparing.get(jobId);
+ if (!entry) return false;
+ entry.aborted = true;
+ return 'pending';
+ },
+ // Called synchronously at the spawn/submit handoff; no await may separate
+ // this gate from installing the provider's active handle.
+ take(jobId) {
+ const entry = preparing.get(jobId);
+ preparing.delete(jobId);
+ return entry?.aborted === true;
+ },
+ ids: () => [...preparing.keys()],
+ };
+}
diff --git a/server/services/imageGen/agy.js b/server/services/imageGen/agy.js
index 44a2c78c91..508b5509af 100644
--- a/server/services/imageGen/agy.js
+++ b/server/services/imageGen/agy.js
@@ -1,3 +1,4 @@
+import { createMediaJobPreparation } from '../../lib/mediaJobPreparation.js';
import { withBackupAssetPublication } from '../../lib/backupSnapshotBoundary.js';
import { maintenance } from '../../lib/maintenanceAdmission.js';
/**
@@ -82,8 +83,12 @@ const terminate = (jobId, proc) => {
});
};
+const preparations = createMediaJobPreparation();
+
export const cancel = (jobId) => {
if (!jobId) throw new Error('agy.cancel requires a jobId — use agy.cancelAll() to terminate every in-flight render');
+ const pending = preparations.cancel(jobId);
+ if (pending) return pending;
const proc = activeProcs.get(jobId);
if (!proc) return false;
terminate(jobId, proc);
@@ -91,9 +96,9 @@ export const cancel = (jobId) => {
};
export const cancelAll = () => {
- const entries = [...activeProcs.entries()];
- if (!entries.length) return false;
- for (const [jobId, proc] of entries) terminate(jobId, proc);
+ const ids = [...preparations.ids(), ...activeProcs.keys()];
+ if (ids.length === 0) return false;
+ for (const id of ids) cancel(id);
return true;
};
@@ -276,7 +281,9 @@ export function agySentPrompt({ prompt = '', negativePrompt, width, height, stag
};
}
-export async function generateImage({
+export const generateImage = (params) => preparations.run(params, generateImagePrepared);
+
+async function generateImagePrepared({
promptAdditions = [],
agyPath,
model,
@@ -408,6 +415,10 @@ function runAgy(job, jobId, bin, args, {
maintenance.markCurrentUnsettled();
console.error(`❌ agy scratch cleanup failed: ${err.message}`);
});
+ if (preparations.take(jobId)) {
+ void removeScratch();
+ return finalizeJobFailure.canceled(job, jobId);
+ }
let proc;
try {
const resolvedBin = (!isAbsolute(bin) && (bin.includes('/') || bin.includes(sep))) ? pathResolve(bin) : bin;
diff --git a/server/services/imageGen/agy.test.js b/server/services/imageGen/agy.test.js
index c46dbedc9d..76f61a2b04 100644
--- a/server/services/imageGen/agy.test.js
+++ b/server/services/imageGen/agy.test.js
@@ -519,3 +519,19 @@ describe('agy image provider — launch refusal settlement', () => {
expect(agy.cancel(job.jobId)).toBe(false);
});
});
+
+it('cancels during preparation before starting a render', async () => {
+ const { imageGenEvents: events } = await import('../imageGenEvents.js');
+ const failed = vi.fn();
+ const completed = vi.fn();
+ events.on('failed', failed);
+ events.on('completed', completed);
+ const jobId = 'preamble-cancel-imageGen-agy';
+ const pending = agy.generateImage({ prompt: 'a quiet gate', jobId });
+ expect(agy.cancel(jobId)).toBe('pending');
+ await pending;
+ await vi.waitFor(() => expect(failed).toHaveBeenCalledWith(expect.objectContaining({ generationId: jobId, error: 'Canceled' })));
+ expect(completed).not.toHaveBeenCalled();
+ expect(spawnCalls).toHaveLength(0);
+ expect(agy.cancel(jobId)).toBe(false);
+});
diff --git a/server/services/imageGen/codex.js b/server/services/imageGen/codex.js
index 49637d06e4..98d709e071 100644
--- a/server/services/imageGen/codex.js
+++ b/server/services/imageGen/codex.js
@@ -1,3 +1,4 @@
+import { createMediaJobPreparation } from '../../lib/mediaJobPreparation.js';
import { withBackupAssetPublication } from '../../lib/backupSnapshotBoundary.js';
/**
* Image Gen — Codex CLI provider.
@@ -130,10 +131,14 @@ const sigtermWithEscalation = (id, proc) => {
// in-flight render), so callers have to be explicit. Use `cancelAll()` for
// the legacy "stop everything" path that the imageGen.cancel() dispatcher
// wires up.
+const preparations = createMediaJobPreparation();
+
export const cancel = (jobId) => {
if (!jobId) {
throw new Error("codex.cancel requires a jobId — use codex.cancelAll() to terminate every in-flight render");
}
+ const pending = preparations.cancel(jobId);
+ if (pending) return pending;
const proc = activeProcs.get(jobId);
if (!proc) return false;
sigtermWithEscalation(jobId, proc);
@@ -144,9 +149,9 @@ export const cancel = (jobId) => {
// dispatcher's "cancel everything" route — the per-job mediaJobQueue path
// always passes a specific jobId to `cancel()`.
export const cancelAll = () => {
- const entries = [...activeProcs.entries()];
- if (entries.length === 0) return false;
- for (const [id, proc] of entries) sigtermWithEscalation(id, proc);
+ const ids = [...preparations.ids(), ...activeProcs.keys()];
+ if (ids.length === 0) return false;
+ for (const id of ids) cancel(id);
return true;
};
@@ -237,7 +242,9 @@ export function codexSentPrompt({ prompt = '', negativePrompt, width, height, in
// Input images are attached via codex CLI's `-i <FILE>` flag (variadic, so all
// of them ride one flag) and the prompt is reshaped so the `$imagegen` skill
// feeds the attachments to `image_gen` as references.
-export async function generateImage({
+export const generateImage = (params) => preparations.run(params, generateImagePrepared);
+
+async function generateImagePrepared({
promptAdditions = [],
codexPath, model, modelIsShippedDefault = false, effort, prompt = '', width, height, negativePrompt,
initImagePath, initImageStrength, referenceImagePaths = [],
@@ -401,6 +408,10 @@ function removeScratchDirectory(scratchDir) {
// close handler to settle once it has physically exited.
function runCodex(job, jobId, bin, args, outputPath, filename, meta, { cleanC2PA = false, denoise = false, scratchDir } = {}) {
const removeScratch = () => removeScratchDirectory(scratchDir);
+ if (preparations.take(jobId)) {
+ void removeScratch();
+ return finalizeJobFailure.canceled(job, jobId);
+ }
let proc;
try {
const resolvedBin = !isAbsolute(bin) && (bin.includes('/') || bin.includes(sep)) ? resolve(bin) : bin;
diff --git a/server/services/imageGen/codex.test.js b/server/services/imageGen/codex.test.js
index 892c642567..4b7a7da58d 100644
--- a/server/services/imageGen/codex.test.js
+++ b/server/services/imageGen/codex.test.js
@@ -1099,3 +1099,19 @@ describe('Codex cached defaults and isolated workspace', () => {
expect(JSON.stringify(failed.mock.calls)).toContain('does not establish whether the tool is available');
});
});
+
+it('cancels during preparation before starting a render', async () => {
+ const { imageGenEvents: events } = await import('../imageGenEvents.js');
+ const failed = vi.fn();
+ const completed = vi.fn();
+ events.on('failed', failed);
+ events.on('completed', completed);
+ const jobId = 'preamble-cancel-imageGen-codex';
+ const pending = codex.generateImage({ prompt: 'a quiet gate', jobId });
+ expect(codex.cancel(jobId)).toBe('pending');
+ await pending;
+ await vi.waitFor(() => expect(failed).toHaveBeenCalledWith(expect.objectContaining({ generationId: jobId, error: 'Canceled' })));
+ expect(completed).not.toHaveBeenCalled();
+ expect(spawnCalls).toHaveLength(0);
+ expect(codex.cancel(jobId)).toBe(false);
+});
diff --git a/server/services/imageGen/fal.js b/server/services/imageGen/fal.js
index 2c59a07e23..e54bea9226 100644
--- a/server/services/imageGen/fal.js
+++ b/server/services/imageGen/fal.js
@@ -1,3 +1,4 @@
+import { createMediaJobPreparation } from '../../lib/mediaJobPreparation.js';
import { withBackupAssetPublication } from '../../lib/backupSnapshotBoundary.js';
/**
* Image Gen — fal.ai queue REST API provider.
@@ -96,10 +97,14 @@ export const attachSseClient = (jobId, res) => attachSse(jobs, jobId, res);
// Cancel one specific render. jobId is required — with parallel renders an
// anonymous cancel would stop every in-flight one; use cancelAll() for that.
+const preparations = createMediaJobPreparation();
+
export const cancel = (jobId) => {
if (!jobId) {
throw new Error('imageGen/fal.cancel requires a jobId — use cancelAll() to terminate every in-flight render');
}
+ const pending = preparations.cancel(jobId);
+ if (pending) return pending;
const entry = activeRequests.get(jobId);
if (!entry) return false;
entry.aborted = true;
@@ -109,7 +114,7 @@ export const cancel = (jobId) => {
};
export const cancelAll = () => {
- const ids = [...activeRequests.keys()];
+ const ids = [...preparations.ids(), ...activeRequests.keys()];
if (ids.length === 0) return false;
for (const id of ids) cancel(id);
return true;
@@ -218,7 +223,9 @@ export function falImageSentPrompt({ family, prompt = '', negativePrompt, inputI
};
}
-export async function generateImage({
+export const generateImage = (params) => preparations.run(params, generateImagePrepared);
+
+async function generateImagePrepared({
promptAdditions = [],
model, prompt = '', negativePrompt, width, height, seed,
initImagePath, initImageStrength, referenceImagePaths = [],
@@ -312,6 +319,7 @@ async function runFal(job, jobId, {
inputPaths, outputPath, filename, meta, cleanC2PA = false, denoise = false,
}) {
const entry = createFalRequestEntry(apiKey);
+ entry.aborted = preparations.take(jobId);
activeRequests.set(jobId, entry);
const deadline = Date.now() + FAL_IMAGE_TIMEOUT_MS;
let wroteOutput = false;
diff --git a/server/services/imageGen/fal.test.js b/server/services/imageGen/fal.test.js
index 4583cbf23e..2a1f4d0b13 100644
--- a/server/services/imageGen/fal.test.js
+++ b/server/services/imageGen/fal.test.js
@@ -249,3 +249,20 @@ describe('imageGen/fal — checkConnection', () => {
expect(await fal.checkConnection({})).toMatchObject({ connected: false, reason: expect.stringContaining('HTTP 401') });
});
});
+
+it('cancels during preparation before starting a render', async () => {
+ const { imageGenEvents: events } = await import('../imageGenEvents.js');
+ const failed = vi.fn();
+ const completed = vi.fn();
+ events.on('failed', failed);
+ events.on('completed', completed);
+ vi.stubGlobal('fetch', vi.fn());
+ const jobId = 'preamble-cancel-imageGen-fal';
+ const pending = fal.generateImage({ prompt: 'a quiet gate', jobId });
+ expect(fal.cancel(jobId)).toBe('pending');
+ await pending;
+ await vi.waitFor(() => expect(failed).toHaveBeenCalledWith(expect.objectContaining({ generationId: jobId, error: 'Canceled' })));
+ expect(completed).not.toHaveBeenCalled();
+ expect(fetch).not.toHaveBeenCalled();
+ expect(fal.cancel(jobId)).toBe(false);
+});
diff --git a/server/services/imageGen/grok.js b/server/services/imageGen/grok.js
index b75f2973c8..db0faa5726 100644
--- a/server/services/imageGen/grok.js
+++ b/server/services/imageGen/grok.js
@@ -1,3 +1,4 @@
+import { createMediaJobPreparation } from '../../lib/mediaJobPreparation.js';
import { withBackupAssetPublication } from '../../lib/backupSnapshotBoundary.js';
import { maintenance } from '../../lib/maintenanceAdmission.js';
/**
@@ -115,10 +116,14 @@ const sigtermWithEscalation = (id, proc) => {
// Cancel one specific grok render. jobId is required — with parallel renders
// an "anonymous cancel" would nuke every in-flight render. Use `cancelAll()`
// for the dispatcher's "stop everything" path.
+const preparations = createMediaJobPreparation();
+
export const cancel = (jobId) => {
if (!jobId) {
throw new Error("grok.cancel requires a jobId — use grok.cancelAll() to terminate every in-flight render");
}
+ const pending = preparations.cancel(jobId);
+ if (pending) return pending;
const proc = activeProcs.get(jobId);
if (!proc) return false;
sigtermWithEscalation(jobId, proc);
@@ -128,9 +133,9 @@ export const cancel = (jobId) => {
// Bulk terminate every in-flight grok render. Only used by the imageGen
// dispatcher's "cancel everything" route.
export const cancelAll = () => {
- const entries = [...activeProcs.entries()];
- if (entries.length === 0) return false;
- for (const [id, proc] of entries) sigtermWithEscalation(id, proc);
+ const ids = [...preparations.ids(), ...activeProcs.keys()];
+ if (ids.length === 0) return false;
+ for (const id of ids) cancel(id);
return true;
};
@@ -238,7 +243,9 @@ export function grokSentPrompt({ prompt = '', negativePrompt, width, height, asp
// `initImageStrength` maps to a fidelity phrase (grok's image_edit has no
// numeric denoise knob, same constraint as codex).
-export async function generateImage({
+export const generateImage = (params) => preparations.run(params, generateImagePrepared);
+
+async function generateImagePrepared({
promptAdditions = [],
grokPath, aspectRatio, prompt = '', width, height, negativePrompt,
initImagePath, initImageStrength, referenceImagePaths = [],
@@ -351,6 +358,10 @@ function runGrok(job, jobId, bin, baseArgs, {
console.error(`❌ grok scratch cleanup failed: ${err.message}`);
});
+ if (preparations.take(jobId)) {
+ void removeScratch();
+ return finalizeJobFailure.canceled(job, jobId);
+ }
let proc;
let useStdin;
try {
diff --git a/server/services/imageGen/grok.test.js b/server/services/imageGen/grok.test.js
index 0d3c42a67f..4497b58ec5 100644
--- a/server/services/imageGen/grok.test.js
+++ b/server/services/imageGen/grok.test.js
@@ -652,3 +652,19 @@ describe('grok provider — launch refusal settlement', () => {
expect(grok.cancel(job.jobId)).toBe(false);
});
});
+
+it('cancels during preparation before starting a render', async () => {
+ const { imageGenEvents: events } = await import('../imageGenEvents.js');
+ const failed = vi.fn();
+ const completed = vi.fn();
+ events.on('failed', failed);
+ events.on('completed', completed);
+ const jobId = 'preamble-cancel-imageGen-grok';
+ const pending = grok.generateImage({ prompt: 'a quiet gate', jobId });
+ expect(grok.cancel(jobId)).toBe('pending');
+ await pending;
+ await vi.waitFor(() => expect(failed).toHaveBeenCalledWith(expect.objectContaining({ generationId: jobId, error: 'Canceled' })));
+ expect(completed).not.toHaveBeenCalled();
+ expect(spawnCalls).toHaveLength(0);
+ expect(grok.cancel(jobId)).toBe(false);
+});
diff --git a/server/services/mediaJobQueue/index.test.js b/server/services/mediaJobQueue/index.test.js
index 20b5d52319..bbabde58cb 100644
--- a/server/services/mediaJobQueue/index.test.js
+++ b/server/services/mediaJobQueue/index.test.js
@@ -1060,6 +1060,25 @@ describe('mediaJobQueue', () => {
for (const path of [uploadedTempPath, audioFilePath, extraPath]) expect(existsSync(path)).toBe(false);
});
+ it('accepts cancellation while a cloud provider is preparing before started', async () => {
+ let finishPreparation;
+ stubs.generateVideoGrok.mockImplementationOnce(({ jobId }) => new Promise(resolve => {
+ finishPreparation = () => {
+ videoGenEvents.emit('failed', { generationId: jobId, error: 'Canceled' });
+ resolve({ jobId });
+ };
+ }));
+ const job = await mediaJobQueue.enqueueJob({ kind: 'video', params: { mode: 'grok', prompt: 'preparing render' } });
+ await waitFor(() => typeof finishPreparation === 'function');
+ stubs.cancelVideo.mockReturnValueOnce('pending');
+ try {
+ await expect(mediaJobQueue.cancelJob(job.jobId)).resolves.toEqual({ ok: true, status: 'canceling' });
+ } finally {
+ finishPreparation();
+ }
+ await waitFor(() => mediaJobQueue.getJob(job.jobId).status === 'canceled');
+ });
+
it('restores cancellation bookkeeping when a provider refuses during finalization', async () => {
const job = await mediaJobQueue.enqueueJob({ kind: 'video', params: {
mode: 'grok', prompt: 'finishing render', videoProduction: { submissionUncertain: false },
diff --git a/server/services/videoGen/fal.js b/server/services/videoGen/fal.js
index ec8499a927..ed0d171122 100644
--- a/server/services/videoGen/fal.js
+++ b/server/services/videoGen/fal.js
@@ -23,6 +23,7 @@
* cost is recorded as unknown.
*/
+import { createMediaJobPreparation } from '../../lib/mediaJobPreparation.js';
import { randomUUID } from 'crypto';
import { readFile, writeFile, unlink } from 'fs/promises';
import { join, relative, resolve, isAbsolute } from 'path';
@@ -98,10 +99,14 @@ export const getActiveJob = () => {
export const attachSseClient = (jobId, res) => attachSse(jobs, jobId, res);
+const preparations = createMediaJobPreparation();
+
export const cancel = (jobId) => {
if (!jobId) {
throw new Error("videoGen/fal.cancel requires a jobId — use cancelAll() to terminate every in-flight render");
}
+ const pending = preparations.cancel(jobId);
+ if (pending) return pending;
const entry = activeRequests.get(jobId);
if (!entry) return false;
entry.aborted = true;
@@ -111,7 +116,7 @@ export const cancel = (jobId) => {
};
export const cancelAll = () => {
- const ids = [...activeRequests.keys()];
+ const ids = [...preparations.ids(), ...activeRequests.keys()];
if (ids.length === 0) return false;
for (const id of ids) cancel(id);
return true;
@@ -181,7 +186,9 @@ export function falVideoSentPrompt({ hasImage = false, hasEndImage = false, hasA
return { planned, record };
}
-export async function generateVideo({
+export const generateVideo = (params) => preparations.run(params, generateVideoPrepared);
+
+async function generateVideoPrepared({
apiKey: providedApiKey, settings, modelId: requestedModelId,
prompt = '', negativePrompt, duration, aspectRatio, width, height,
sourceImagePath = null, lastImagePath = null, jobId: providedJobId = null,
@@ -314,6 +321,7 @@ async function runFalVideo(job, jobId, {
frameAspect = null,
}) {
const entry = createFalRequestEntry(apiKey);
+ entry.aborted = preparations.take(jobId);
const publication = {};
// Every terminal path shares the same cleanup attempt. A publication error
// can enter the catch after the success path has already settled its inputs.
diff --git a/server/services/videoGen/fal.test.js b/server/services/videoGen/fal.test.js
index 02e826a0fe..e934637792 100644
--- a/server/services/videoGen/fal.test.js
+++ b/server/services/videoGen/fal.test.js
@@ -934,3 +934,20 @@ describe('hosted inputs', () => {
expect(JSON.stringify(posted)).not.toContain('data:');
});
});
+
+it('cancels during preparation before starting a render', async () => {
+ const { videoGenEvents: events } = await import('./events.js');
+ const failed = vi.fn();
+ const completed = vi.fn();
+ events.on('failed', failed);
+ events.on('completed', completed);
+ vi.stubGlobal('fetch', vi.fn());
+ const jobId = 'preamble-cancel-videoGen-fal';
+ const pending = fal.generateVideo({ prompt: 'a quiet gate', jobId, apiKey: 'test-key', modelId: 'fal-ai/veo3.1/fast' });
+ expect(fal.cancel(jobId)).toBe('pending');
+ await pending;
+ await vi.waitFor(() => expect(failed).toHaveBeenCalledWith(expect.objectContaining({ generationId: jobId, error: 'Canceled' })));
+ expect(completed).not.toHaveBeenCalled();
+ expect(fetch).not.toHaveBeenCalled();
+ expect(fal.cancel(jobId)).toBe(false);
+});
diff --git a/server/services/videoGen/grok.js b/server/services/videoGen/grok.js
index a5914eca46..640556a5bb 100644
--- a/server/services/videoGen/grok.js
+++ b/server/services/videoGen/grok.js
@@ -21,6 +21,7 @@
* disabled/missing-binary up front; this module assumes it's enabled.
*/
+import { createMediaJobPreparation } from '../../lib/mediaJobPreparation.js';
import { spawn } from '../../lib/childProcess.js';
import { mkdir, open, stat } from 'fs/promises';
import { isAbsolute, join, resolve as pathResolve, sep } from 'path';
@@ -83,10 +84,14 @@ const sigtermWithEscalation = (id, proc) => {
killWithEscalation(proc, { label: 'grok video child', delayMs: 5000, stillRunning: () => activeProcs.get(id) === proc });
};
+const preparations = createMediaJobPreparation();
+
export const cancel = (jobId) => {
if (!jobId) {
throw new Error("videoGen/grok.cancel requires a jobId — use cancelAll() to terminate every in-flight render");
}
+ const pending = preparations.cancel(jobId);
+ if (pending) return pending;
const proc = activeProcs.get(jobId);
if (!proc) return false;
sigtermWithEscalation(jobId, proc);
@@ -94,9 +99,9 @@ export const cancel = (jobId) => {
};
export const cancelAll = () => {
- const entries = [...activeProcs.entries()];
- if (entries.length === 0) return false;
- for (const [id, proc] of entries) sigtermWithEscalation(id, proc);
+ const ids = [...preparations.ids(), ...activeProcs.keys()];
+ if (ids.length === 0) return false;
+ for (const id of ids) cancel(id);
return true;
};
@@ -141,7 +146,9 @@ export function grokVideoSentPrompt({
};
}
-export async function generateVideo({
+export const generateVideo = (params) => preparations.run(params, generateVideoPrepared);
+
+async function generateVideoPrepared({
grokPath, aspectRatio, prompt = '', negativePrompt, width, height,
sourceImagePath = null, duration,
uploadedTempPath = null,
@@ -220,15 +227,6 @@ export async function generateVideo({
async function runGrokVideo(job, jobId, bin, args, {
useStdin, fullPrompt, cleanupPromptFile, scratchDir, stagingPath, outputPath, filename, meta, uploadedTempPath,
}) {
- // Resolve a path-shaped grokPath against the PortOS cwd before the child's
- // cwd moves to the scratch dir; bare names stay bare for PATH lookup.
- const resolvedBin = (!isAbsolute(bin) && (bin.includes('/') || bin.includes(sep))) ? pathResolve(bin) : bin;
- const { command: spawnBin, args: spawnArgs } = prepareCliSpawn(resolvedBin, args);
- // Pin PWD to the spawn cwd — see withSpawnCwdEnv (#3193). grok reads
- // process.cwd(), so this is defensive rather than a live fix; it keeps every
- // scratch-dir spawn telling the child one consistent story about where it is.
- const proc = spawn(spawnBin, spawnArgs, { cwd: scratchDir, env: withSpawnCwdEnv(process.env, scratchDir), shell: false, stdio: [useStdin ? 'pipe' : 'ignore', 'pipe', 'pipe'] });
- activeProcs.set(jobId, proc);
const retainOwnership = (error) => {
maintenance.markCurrentUnsettled();
maintenance.markResourceUnsettled('media', jobId);
@@ -249,6 +247,20 @@ async function runGrokVideo(job, jobId, bin, args, {
}
});
+ if (preparations.take(jobId)) {
+ await cleanup();
+ return finalizeJobFailure.canceled(job, jobId);
+ }
+ // Resolve a path-shaped grokPath against the PortOS cwd before the child's
+ // cwd moves to the scratch dir; bare names stay bare for PATH lookup.
+ const resolvedBin = (!isAbsolute(bin) && (bin.includes('/') || bin.includes(sep))) ? pathResolve(bin) : bin;
+ const { command: spawnBin, args: spawnArgs } = prepareCliSpawn(resolvedBin, args);
+ // Pin PWD to the spawn cwd — see withSpawnCwdEnv (#3193). grok reads
+ // process.cwd(), so this is defensive rather than a live fix; it keeps every
+ // scratch-dir spawn telling the child one consistent story about where it is.
+ const proc = spawn(spawnBin, spawnArgs, { cwd: scratchDir, env: withSpawnCwdEnv(process.env, scratchDir), shell: false, stdio: [useStdin ? 'pipe' : 'ignore', 'pipe', 'pipe'] });
+ activeProcs.set(jobId, proc);
+
if (useStdin) {
proc.stdin.on('error', () => {});
proc.stdin.write(fullPrompt);
diff --git a/server/services/videoGen/grok.test.js b/server/services/videoGen/grok.test.js
index eb457e6068..6f12e1adfd 100644
--- a/server/services/videoGen/grok.test.js
+++ b/server/services/videoGen/grok.test.js
@@ -415,3 +415,19 @@ describe('videoGen/grok — cancel', () => {
expect(grok.cancelAll()).toBe(false);
});
});
+
+it('cancels during preparation before starting a render', async () => {
+ const { videoGenEvents: events } = await import('./events.js');
+ const failed = vi.fn();
+ const completed = vi.fn();
+ events.on('failed', failed);
+ events.on('completed', completed);
+ const jobId = 'preamble-cancel-videoGen-grok';
+ const pending = grok.generateVideo({ prompt: 'a quiet gate', jobId });
+ expect(grok.cancel(jobId)).toBe('pending');
+ await pending;
+ await vi.waitFor(() => expect(failed).toHaveBeenCalledWith(expect.objectContaining({ generationId: jobId, error: 'Canceled' })));
+ expect(completed).not.toHaveBeenCalled();
+ expect(spawnCalls).toHaveLength(0);
+ expect(grok.cancel(jobId)).toBe(false);
+});
diff --git a/server/services/videoGen/reactor.js b/server/services/videoGen/reactor.js
index d7e039aa26..d3f6b3e4dd 100644
--- a/server/services/videoGen/reactor.js
+++ b/server/services/videoGen/reactor.js
@@ -1,4 +1,5 @@
/** Reactor FastH3 renders one bounded SDK/WebRTC session per job. */
+import { createMediaJobPreparation } from '../../lib/mediaJobPreparation.js';
import { randomUUID } from 'crypto';
import { stat, rm } from 'fs/promises';
import { spawn } from '../../lib/childProcess.js';
@@ -79,10 +80,14 @@ export async function mintReactorToken(apiKey) {
return { jwt: payload.jwt, expiresAt: payload.expires_at || null };
}
+const preparations = createMediaJobPreparation();
+
export const cancel = (jobId) => {
if (!jobId) {
throw new Error("videoGen/reactor.cancel requires a jobId — use cancelAll() to terminate every in-flight render");
}
+ const pending = preparations.cancel(jobId);
+ if (pending) return pending;
const entry = activeRequests.get(jobId);
if (!entry || entry.finalizing) return false;
entry.aborted = true;
@@ -91,7 +96,7 @@ export const cancel = (jobId) => {
};
export const cancelAll = () => {
- const ids = [...activeRequests.keys()];
+ const ids = [...preparations.ids(), ...activeRequests.keys()];
if (ids.length === 0) return false;
for (const id of ids) cancel(id);
return true;
@@ -211,7 +216,9 @@ function captureClip(entry, input, pythonPath, job, jobId) {
});
}
-export async function generateVideo({
+export const generateVideo = (params) => preparations.run(params, generateVideoPrepared);
+
+async function generateVideoPrepared({
settings, prompt = '', negativePrompt,
continueFromClipId, seconds, seed, aspect,
sourceImagePath = null, jobId: providedJobId = null,
@@ -272,6 +279,7 @@ async function runReactorVideo(job, jobId, {
apiKey, prompt, seconds, seed, aspect, continueFromClipId, sourceImagePath, outputPath, filename, meta,
}) {
const entry = { aborted: false, finalizing: false, stop: null };
+ entry.aborted = preparations.take(jobId);
activeRequests.set(jobId, entry);
// Declared outside the try so the finally can still delete a fitted frame
// whose render failed after it was written.
diff --git a/server/services/videoGen/reactor.test.js b/server/services/videoGen/reactor.test.js
index 8a82702b0c..f6b6577dae 100644
--- a/server/services/videoGen/reactor.test.js
+++ b/server/services/videoGen/reactor.test.js
@@ -405,3 +405,20 @@ describe('Reactor SDK adapter', () => {
});
});
+
+it('cancels during preparation before starting a render', async () => {
+ const { videoGenEvents: events } = await import('./events.js');
+ const failed = vi.fn();
+ const completed = vi.fn();
+ events.on('failed', failed);
+ events.on('completed', completed);
+ const jobId = 'preamble-cancel-videoGen-reactor';
+ const pending = reactor.generateVideo({ prompt: 'a quiet gate', jobId });
+ expect(reactor.cancel(jobId)).toBe('pending');
+ await pending;
+ await vi.waitFor(() => expect(failed).toHaveBeenCalledWith(expect.objectContaining({ generationId: jobId, error: 'Canceled' })));
+ expect(completed).not.toHaveBeenCalled();
+ expect(mocks.spawn).not.toHaveBeenCalled();
+ expect(fetch).not.toHaveBeenCalled();
+ expect(reactor.cancel(jobId)).toBe(false);
+});