@vertesia/workflow 1.5.0-dev.20260714.072725Z → 1.5.0-dev.20260722.120446Z

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (58) hide show
  1. package/lib/activities/generateDocumentProperties.d.ts +54 -13
  2. package/lib/activities/generateDocumentProperties.d.ts.map +1 -1
  3. package/lib/activities/generateDocumentProperties.js +109 -34
  4. package/lib/activities/generateDocumentProperties.js.map +1 -1
  5. package/lib/activities/generateEmbeddings.d.ts.map +1 -1
  6. package/lib/activities/generateEmbeddings.js +14 -22
  7. package/lib/activities/generateEmbeddings.js.map +1 -1
  8. package/lib/activities/generateOrAssignContentType.d.ts +6 -16
  9. package/lib/activities/generateOrAssignContentType.d.ts.map +1 -1
  10. package/lib/activities/generateOrAssignContentType.js +70 -19
  11. package/lib/activities/generateOrAssignContentType.js.map +1 -1
  12. package/lib/activities/media/exec.d.ts +60 -0
  13. package/lib/activities/media/exec.d.ts.map +1 -0
  14. package/lib/activities/media/exec.js +212 -0
  15. package/lib/activities/media/exec.js.map +1 -0
  16. package/lib/activities/media/prepareAudio.d.ts.map +1 -1
  17. package/lib/activities/media/prepareAudio.js +7 -6
  18. package/lib/activities/media/prepareAudio.js.map +1 -1
  19. package/lib/activities/media/prepareVideo.d.ts +5 -0
  20. package/lib/activities/media/prepareVideo.d.ts.map +1 -1
  21. package/lib/activities/media/prepareVideo.js +37 -13
  22. package/lib/activities/media/prepareVideo.js.map +1 -1
  23. package/lib/activities/media/probeMediaStreams.d.ts.map +1 -1
  24. package/lib/activities/media/probeMediaStreams.js +7 -6
  25. package/lib/activities/media/probeMediaStreams.js.map +1 -1
  26. package/lib/activities/renditions/generateVideoRendition.d.ts.map +1 -1
  27. package/lib/activities/renditions/generateVideoRendition.js +7 -5
  28. package/lib/activities/renditions/generateVideoRendition.js.map +1 -1
  29. package/lib/dsl/dsl-workflow.js +3 -0
  30. package/lib/dsl/dsl-workflow.js.map +1 -1
  31. package/lib/dsl/dslProxyActivities.d.ts +1 -0
  32. package/lib/dsl/dslProxyActivities.d.ts.map +1 -1
  33. package/lib/dsl/dslProxyActivities.js +9 -0
  34. package/lib/dsl/dslProxyActivities.js.map +1 -1
  35. package/lib/errors.d.ts.map +1 -1
  36. package/lib/errors.js +5 -0
  37. package/lib/errors.js.map +1 -1
  38. package/lib/workflows-bundle.js +16111 -6668
  39. package/package.json +17 -16
  40. package/src/activities/generateDocumentProperties.test.ts +642 -0
  41. package/src/activities/generateDocumentProperties.ts +187 -43
  42. package/src/activities/generateEmbeddings.test.ts +114 -0
  43. package/src/activities/generateEmbeddings.ts +17 -24
  44. package/src/activities/generateOrAssignContentType.test.ts +196 -0
  45. package/src/activities/generateOrAssignContentType.ts +115 -30
  46. package/src/activities/media/exec.test.ts +194 -0
  47. package/src/activities/media/exec.ts +265 -0
  48. package/src/activities/media/prepareAudio.ts +12 -7
  49. package/src/activities/media/prepareVideo.test.ts +27 -0
  50. package/src/activities/media/prepareVideo.ts +54 -18
  51. package/src/activities/media/probeMediaStreams.test.ts +7 -8
  52. package/src/activities/media/probeMediaStreams.ts +11 -7
  53. package/src/activities/renditions/generateVideoRendition.ts +12 -6
  54. package/src/dsl/dsl-workflow.test.ts +4 -0
  55. package/src/dsl/dsl-workflow.ts +3 -0
  56. package/src/dsl/dslProxyActivities.test.ts +23 -0
  57. package/src/dsl/dslProxyActivities.ts +11 -0
  58. package/src/errors.ts +5 -0
@@ -0,0 +1,265 @@
1
+ import { type ExecFileOptions, execFile as execFileCallback, spawn } from 'node:child_process';
2
+ import { promisify } from 'node:util';
3
+ import { Context } from '@temporalio/activity';
4
+
5
+ const execFileAsync = promisify(execFileCallback);
6
+
7
+ /** 10MB buffer for ffmpeg output. */
8
+ export const FFMPEG_MAX_BUFFER = 1024 * 1024 * 10;
9
+
10
+ const MAX_COMMAND_TIMEOUT_MARGIN_MS = 60_000;
11
+
12
+ /** Terminate an ffmpeg child that makes no encoding progress for this long, so Temporal can retry it promptly. */
13
+ const FFMPEG_STALL_TIMEOUT_MS = 120_000;
14
+
15
+ // ffmpeg streams encoding progress as `frame=`/`time=` stats. We track the *advance* of these monotonic fields (rather
16
+ // than mere presence) so a wedged encoder that keeps re-emitting an identical status line is still detected as stalled.
17
+ const FFMPEG_TIME_PATTERN = /time=\s*(\d+):([0-5]?\d):([0-5]?\d(?:\.\d+)?)/;
18
+ const FFMPEG_FRAME_PATTERN = /frame=\s*(\d+)/;
19
+ // ffmpeg delimits stats snapshots with a carriage return, so the carry buffer normally stays tiny. These bounds are a
20
+ // defensive guard against an unterminated stream, retaining enough of a tail to reunite a marker split at the boundary.
21
+ const PROGRESS_CARRY_LIMIT = 64 * 1024;
22
+ const PROGRESS_CARRY_TAIL = 512;
23
+
24
+ const activityCommandDeadlines = new WeakMap<Context, number>();
25
+
26
+ /**
27
+ * Anchor the child-process deadline at the moment the Activity starts, so every later command derives its timeout from a
28
+ * stable point regardless of how long input download took.
29
+ */
30
+ export function initializeActivityCommandDeadline(): number {
31
+ const context = Context.current();
32
+ const existingDeadlineMs = activityCommandDeadlines.get(context);
33
+ if (existingDeadlineMs) {
34
+ return existingDeadlineMs;
35
+ }
36
+ const deadlineMs = Date.now() + context.info.startToCloseTimeoutMs;
37
+ activityCommandDeadlines.set(context, deadlineMs);
38
+ return deadlineMs;
39
+ }
40
+
41
+ function getActivityCommandTimeoutMs(): number {
42
+ const context = Context.current();
43
+ const { startToCloseTimeoutMs } = context.info;
44
+ const timeoutMarginMs = Math.min(MAX_COMMAND_TIMEOUT_MARGIN_MS, startToCloseTimeoutMs / 10);
45
+ const activityDeadlineMs = initializeActivityCommandDeadline();
46
+ return Math.max(1, activityDeadlineMs - Date.now() - timeoutMarginMs);
47
+ }
48
+
49
+ /**
50
+ * Execute media tooling with Temporal cancellation propagated to the child process.
51
+ *
52
+ * Temporal can begin a retry after an attempt times out while the original worker process is still running. Passing the
53
+ * Activity cancellation signal prevents an orphaned ffmpeg process from competing with its retry for CPU.
54
+ */
55
+ export async function execActivityFile(
56
+ command: string,
57
+ args: string[],
58
+ options: ExecFileOptions = {},
59
+ ): Promise<{ stdout: string; stderr: string }> {
60
+ const activityTimeoutMs = getActivityCommandTimeoutMs();
61
+ const configuredTimeoutMs = options.timeout && options.timeout > 0 ? options.timeout : activityTimeoutMs;
62
+
63
+ return (await execFileAsync(command, args, {
64
+ ...options,
65
+ encoding: 'utf8',
66
+ signal: Context.current().cancellationSignal,
67
+ timeout: Math.min(configuredTimeoutMs, activityTimeoutMs),
68
+ })) as { stdout: string; stderr: string };
69
+ }
70
+
71
+ export interface ProgressAwareOptions {
72
+ /**
73
+ * Coarse upper bound on retained stdout/stderr, counted in UTF-16 code units (not UTF-8 bytes). ffmpeg output is
74
+ * ASCII-dominant so this closely tracks bytes; it only exists to keep captured output from growing without bound
75
+ * (older output is dropped — tail retained — rather than failing the command). Not a strict byte cap.
76
+ */
77
+ maxBuffer?: number;
78
+ stallTimeoutMs?: number;
79
+ }
80
+
81
+ function parseFfmpegTimeCentis(text: string): number | undefined {
82
+ const match = FFMPEG_TIME_PATTERN.exec(text);
83
+ if (!match) {
84
+ return undefined;
85
+ }
86
+ const totalSeconds = Number(match[1]) * 3600 + Number(match[2]) * 60 + Number(match[3]);
87
+ return Number.isFinite(totalSeconds) ? Math.round(totalSeconds * 100) : undefined;
88
+ }
89
+
90
+ function parseFfmpegFrame(text: string): number | undefined {
91
+ const match = FFMPEG_FRAME_PATTERN.exec(text);
92
+ return match ? Number(match[1]) : undefined;
93
+ }
94
+
95
+ export interface ProgressTracker {
96
+ /** Feed a raw output chunk; returns true when a monotonic progress field advanced past its previous maximum. */
97
+ observe(chunk: string): boolean;
98
+ }
99
+
100
+ /**
101
+ * Track ffmpeg forward progress across arbitrarily-chunked output.
102
+ *
103
+ * Node delivers stream data in arbitrary slices, so a marker such as `frame=42` can arrive split across two chunks. The
104
+ * tracker buffers partial records until a `\r`/`\n` delimiter completes them, then reports advancement only when the
105
+ * parsed frame count or media timestamp exceeds the previous maximum — repeated identical stats from a wedged encoder do
106
+ * not count as progress.
107
+ */
108
+ export function createFfmpegProgressTracker(): ProgressTracker {
109
+ let carry = '';
110
+ let maxTimeCentis = -1;
111
+ let maxFrame = -1;
112
+
113
+ const noteSegment = (segment: string): boolean => {
114
+ let advanced = false;
115
+ const timeCentis = parseFfmpegTimeCentis(segment);
116
+ if (timeCentis !== undefined && timeCentis > maxTimeCentis) {
117
+ maxTimeCentis = timeCentis;
118
+ advanced = true;
119
+ }
120
+ const frame = parseFfmpegFrame(segment);
121
+ if (frame !== undefined && frame > maxFrame) {
122
+ maxFrame = frame;
123
+ advanced = true;
124
+ }
125
+ return advanced;
126
+ };
127
+
128
+ return {
129
+ observe(chunk: string): boolean {
130
+ carry += chunk;
131
+ const segments = carry.split(/[\r\n]+/);
132
+ carry = segments.pop() ?? '';
133
+ let advanced = false;
134
+ for (const segment of segments) {
135
+ advanced = noteSegment(segment) || advanced;
136
+ }
137
+ if (carry.length > PROGRESS_CARRY_LIMIT) {
138
+ advanced = noteSegment(carry) || advanced;
139
+ carry = carry.slice(carry.length - PROGRESS_CARRY_TAIL);
140
+ }
141
+ return advanced;
142
+ },
143
+ };
144
+ }
145
+
146
+ /**
147
+ * Execute a long-running media command while enforcing a progress-stall watchdog on top of the Activity deadline and
148
+ * cancellation.
149
+ *
150
+ * The worker heartbeats every Activity on a fixed interval, so a wedged ffmpeg child that stops emitting progress would
151
+ * otherwise keep the Activity alive until its multi-hour Start-to-Close timeout. Watching ffmpeg's progress markers lets
152
+ * us terminate a stalled encode within `stallTimeoutMs` so Temporal retries it promptly (escalating to a faster preset
153
+ * on the next attempt). A stall is surfaced as a `killed` error so it propagates as an Activity failure via
154
+ * {@link rethrowIfActivityStopped} instead of being swallowed into a missing rendition.
155
+ */
156
+ export async function execActivityFileWithProgress(
157
+ command: string,
158
+ args: string[],
159
+ { maxBuffer = FFMPEG_MAX_BUFFER, stallTimeoutMs = FFMPEG_STALL_TIMEOUT_MS }: ProgressAwareOptions = {},
160
+ ): Promise<{ stdout: string; stderr: string }> {
161
+ const activityTimeoutMs = getActivityCommandTimeoutMs();
162
+ const stallWindowMs = Math.max(1, Math.min(stallTimeoutMs, activityTimeoutMs));
163
+
164
+ return await new Promise<{ stdout: string; stderr: string }>((resolve, reject) => {
165
+ const child = spawn(command, args, {
166
+ signal: Context.current().cancellationSignal,
167
+ timeout: activityTimeoutMs,
168
+ killSignal: 'SIGKILL',
169
+ });
170
+
171
+ let stdout = '';
172
+ let stderr = '';
173
+ let settled = false;
174
+ let stalled = false;
175
+ let stallTimer: NodeJS.Timeout | undefined;
176
+
177
+ const armStallTimer = () => {
178
+ if (stallTimer) {
179
+ clearTimeout(stallTimer);
180
+ }
181
+ stallTimer = setTimeout(() => {
182
+ stalled = true;
183
+ child.kill('SIGKILL');
184
+ }, stallWindowMs);
185
+ // Never keep the event loop alive solely for the watchdog timer.
186
+ stallTimer.unref?.();
187
+ };
188
+
189
+ const append = (current: string, chunk: string): string => {
190
+ const combined = current + chunk;
191
+ return combined.length > maxBuffer ? combined.slice(combined.length - maxBuffer) : combined;
192
+ };
193
+
194
+ // ffmpeg writes progress to stderr by default, but track both streams so the watchdog works regardless of where
195
+ // a given command reports it. Each stream keeps its own line buffer.
196
+ const stdoutProgress = createFfmpegProgressTracker();
197
+ const stderrProgress = createFfmpegProgressTracker();
198
+
199
+ child.stdout?.setEncoding('utf8');
200
+ child.stderr?.setEncoding('utf8');
201
+ child.stdout?.on('data', (chunk: string) => {
202
+ stdout = append(stdout, chunk);
203
+ if (stdoutProgress.observe(chunk)) {
204
+ armStallTimer();
205
+ }
206
+ });
207
+ child.stderr?.on('data', (chunk: string) => {
208
+ stderr = append(stderr, chunk);
209
+ if (stderrProgress.observe(chunk)) {
210
+ armStallTimer();
211
+ }
212
+ });
213
+
214
+ armStallTimer();
215
+
216
+ const settle = (finish: () => void) => {
217
+ if (settled) {
218
+ return;
219
+ }
220
+ settled = true;
221
+ if (stallTimer) {
222
+ clearTimeout(stallTimer);
223
+ }
224
+ finish();
225
+ };
226
+
227
+ child.on('error', (error) => settle(() => reject(error)));
228
+
229
+ child.on('close', (code, signal) =>
230
+ settle(() => {
231
+ if (stalled) {
232
+ const seconds = Math.round(stallWindowMs / 1000);
233
+ reject(
234
+ Object.assign(new Error(`${command} made no progress for ${seconds}s and was terminated`), {
235
+ killed: true,
236
+ stalled: true,
237
+ }),
238
+ );
239
+ } else if (code === 0) {
240
+ resolve({ stdout, stderr });
241
+ } else {
242
+ const detail = signal ? `signal ${signal}` : `code ${code}`;
243
+ reject(
244
+ Object.assign(new Error(`${command} exited with ${detail}: ${stderr.slice(-2000)}`), {
245
+ killed: child.killed || signal != null,
246
+ code,
247
+ signal,
248
+ }),
249
+ );
250
+ }
251
+ }),
252
+ );
253
+ });
254
+ }
255
+
256
+ /**
257
+ * Re-throw errors that represent the Activity being stopped (cancellation, Start-to-Close timeout, or a progress-stall
258
+ * kill) so they are not swallowed by rendition fallback handling. Genuine tool failures still return `null` upstream.
259
+ */
260
+ export function rethrowIfActivityStopped(error: unknown): void {
261
+ const commandTimedOut = typeof error === 'object' && error !== null && 'killed' in error && error.killed === true;
262
+ if (Context.current().cancellationSignal.aborted || commandTimedOut) {
263
+ throw error;
264
+ }
265
+ }
@@ -1,8 +1,6 @@
1
- import { execFile as execFileCallback } from 'node:child_process';
2
1
  import fs from 'node:fs';
3
2
  import os from 'node:os';
4
3
  import path from 'node:path';
5
- import { promisify } from 'node:util';
6
4
  import { log } from '@temporalio/activity';
7
5
  import { RequestError } from '@vertesia/api-fetch-client';
8
6
  import type { VertesiaClient } from '@vertesia/client';
@@ -17,12 +15,15 @@ import {
17
15
  import { setupActivity } from '../../dsl/setup/ActivityContext.js';
18
16
  import { DocumentNotFoundError, InvalidContentTypeError } from '../../errors.js';
19
17
  import { saveBlobToTempFile } from '../../utils/blobs.js';
20
-
21
- const execFileAsync = promisify(execFileCallback);
18
+ import {
19
+ execActivityFile,
20
+ execActivityFileWithProgress,
21
+ initializeActivityCommandDeadline,
22
+ rethrowIfActivityStopped,
23
+ } from './exec.js';
22
24
 
23
25
  // Default configuration constants
24
26
  const DEFAULT_AUDIO_BITRATE = '128k'; // Default audio bitrate for AAC encoding
25
- const FFMPEG_MAX_BUFFER = 1024 * 1024 * 10; // 10MB buffer for ffmpeg output
26
27
 
27
28
  export interface PrepareAudioParams {
28
29
  audioBitrate?: string; // Audio bitrate for AAC encoding, default '128k'
@@ -79,7 +80,7 @@ export interface PrepareAudioResult {
79
80
  async function getAudioMetadata(audioPath: string): Promise<AudioMetadataExtended> {
80
81
  try {
81
82
  const args = ['-v', 'quiet', '-print_format', 'json', '-show_format', '-show_streams', audioPath];
82
- const { stdout } = await execFileAsync('ffprobe', args);
83
+ const { stdout } = await execActivityFile('ffprobe', args);
83
84
  const metadata = JSON.parse(stdout.toString()) as FFProbeOutput;
84
85
 
85
86
  const audioStream = metadata.streams.find((stream) => stream.codec_type === 'audio');
@@ -96,6 +97,7 @@ async function getAudioMetadata(audioPath: string): Promise<AudioMetadataExtende
96
97
 
97
98
  return { duration, codec, bitrate, sampleRate, channels };
98
99
  } catch (error) {
100
+ rethrowIfActivityStopped(error);
99
101
  log.error(`Failed to get audio metadata: ${error instanceof Error ? error.message : 'Unknown error'}`);
100
102
  throw new Error(`Failed to probe audio metadata: ${error instanceof Error ? error.message : 'Unknown error'}`);
101
103
  }
@@ -128,7 +130,7 @@ async function generateAudioRendition(
128
130
  log.info('Generating web audio rendition (AAC M4A)', { command: 'ffmpeg', args: command, audioBitrate });
129
131
 
130
132
  try {
131
- const { stderr } = await execFileAsync('ffmpeg', command, { maxBuffer: FFMPEG_MAX_BUFFER });
133
+ const { stderr } = await execActivityFileWithProgress('ffmpeg', command);
132
134
  const stderrText = stderr.toString();
133
135
 
134
136
  if (stderrText && !stderrText.includes('frame=')) {
@@ -145,6 +147,7 @@ async function generateAudioRendition(
145
147
  return null;
146
148
  }
147
149
  } catch (error) {
150
+ rethrowIfActivityStopped(error);
148
151
  log.error(`Failed to generate audio rendition: ${error instanceof Error ? error.message : 'Unknown error'}`);
149
152
  return null;
150
153
  }
@@ -200,6 +203,7 @@ async function uploadAudioAsRendition(
200
203
  export async function prepareAudio(
201
204
  payload: DSLActivityExecutionPayload<PrepareAudioParams>,
202
205
  ): Promise<PrepareAudioResult> {
206
+ initializeActivityCommandDeadline();
203
207
  const { client, objectId, params } = await setupActivity<PrepareAudioParams>(payload);
204
208
 
205
209
  const audioBitrate = params.audioBitrate ?? DEFAULT_AUDIO_BITRATE;
@@ -294,6 +298,7 @@ export async function prepareAudio(
294
298
  status: 'success',
295
299
  };
296
300
  } catch (error) {
301
+ rethrowIfActivityStopped(error);
297
302
  const errorMessage = error instanceof Error ? error.message : 'Unknown error';
298
303
  log.error(`Error preparing audio: ${errorMessage}`, { error });
299
304
 
@@ -0,0 +1,27 @@
1
+ import { describe, expect, it } from 'vitest';
2
+ import { resolveVideoPreset, shouldTranscodeVideo } from './prepareVideo.js';
3
+
4
+ describe('shouldTranscodeVideo', () => {
5
+ it('transcodes by default', () => {
6
+ expect(shouldTranscodeVideo()).toBe(true);
7
+ expect(shouldTranscodeVideo(false)).toBe(true);
8
+ });
9
+
10
+ it('skips transcoding when explicitly requested', () => {
11
+ expect(shouldTranscodeVideo(true)).toBe(false);
12
+ });
13
+ });
14
+
15
+ describe('resolveVideoPreset', () => {
16
+ it.each([
17
+ { preset: 'medium', attempt: 1, expected: 'medium' },
18
+ { preset: 'medium', attempt: 2, expected: 'fast' },
19
+ { preset: 'medium', attempt: 3, expected: 'veryfast' },
20
+ { preset: 'medium', attempt: 5, expected: 'veryfast' },
21
+ { preset: 'fast', attempt: 1, expected: 'fast' },
22
+ { preset: 'fast', attempt: 2, expected: 'veryfast' },
23
+ { preset: 'ultrafast', attempt: 2, expected: 'ultrafast' },
24
+ ] as const)('uses $expected for $preset on attempt $attempt', ({ preset, attempt, expected }) => {
25
+ expect(resolveVideoPreset(preset, attempt)).toBe(expected);
26
+ });
27
+ });
@@ -1,9 +1,7 @@
1
- import { execFile as execFileCallback } from 'node:child_process';
2
1
  import fs from 'node:fs';
3
2
  import os from 'node:os';
4
3
  import path from 'node:path';
5
- import { promisify } from 'node:util';
6
- import { ApplicationFailure, log } from '@temporalio/activity';
4
+ import { ApplicationFailure, Context, log } from '@temporalio/activity';
7
5
  import { RequestError } from '@vertesia/api-fetch-client';
8
6
  import type { VertesiaClient } from '@vertesia/client';
9
7
  import {
@@ -19,8 +17,12 @@ import {
19
17
  import { setupActivity } from '../../dsl/setup/ActivityContext.js';
20
18
  import { DocumentNotFoundError, InvalidContentTypeError } from '../../errors.js';
21
19
  import { saveBlobToTempFile } from '../../utils/blobs.js';
22
-
23
- const execFileAsync = promisify(execFileCallback);
20
+ import {
21
+ execActivityFile,
22
+ execActivityFileWithProgress,
23
+ initializeActivityCommandDeadline,
24
+ rethrowIfActivityStopped,
25
+ } from './exec.js';
24
26
 
25
27
  // Default configuration constants
26
28
  const DEFAULT_MAX_RESOLUTION = 1920; // Max resolution for video rendition (produces 1080p)
@@ -35,7 +37,6 @@ const POSTER_TIMESTAMP_MAX = 2; // Maximum poster timestamp in seconds
35
37
  const MIN_SCREENSHOT_TIMESTAMP = 1; // Minimum timestamp for screenshots in seconds
36
38
 
37
39
  // FFmpeg configuration constants
38
- const FFMPEG_MAX_BUFFER = 1024 * 1024 * 10; // 10MB buffer for ffmpeg output
39
40
  const VIDEO_CRF = '23'; // Constant Rate Factor for video quality (18-28, lower = better)
40
41
  const AUDIO_BITRATE = '128k'; // Audio bitrate for AAC encoding
41
42
  const JPEG_QUALITY = '2'; // JPEG quality for screenshots (1-31, lower = better)
@@ -45,6 +46,24 @@ export interface PrepareVideoParams {
45
46
  thumbnailSize?: number; // Max size (longest side) for thumbnail image, default 256
46
47
  posterSize?: number; // Max size (longest side) for poster image, default 1920
47
48
  generateAudio?: boolean; // Generate audio-only rendition (AAC), default true
49
+ videoPreset?: VideoPreset; // FFmpeg encoding preset, default medium
50
+ skipTranscode?: boolean; // Skip the main video re-encode, default false
51
+ }
52
+
53
+ export type VideoPreset = 'medium' | 'fast' | 'veryfast' | 'ultrafast';
54
+
55
+ export function shouldTranscodeVideo(skipTranscode?: boolean): boolean {
56
+ return skipTranscode !== true;
57
+ }
58
+
59
+ export function resolveVideoPreset(preset: VideoPreset, attempt: number): VideoPreset {
60
+ if (preset === 'ultrafast') {
61
+ return preset;
62
+ }
63
+
64
+ const retryPresets: VideoPreset[] = ['medium', 'fast', 'veryfast'];
65
+ const initialPresetIndex = retryPresets.indexOf(preset);
66
+ return retryPresets[Math.min(initialPresetIndex + attempt - 1, retryPresets.length - 1)];
48
67
  }
49
68
 
50
69
  export interface PrepareVideo extends DSLActivitySpec<PrepareVideoParams> {
@@ -85,7 +104,7 @@ interface FFProbeOutput {
85
104
  async function getVideoMetadata(videoPath: string): Promise<VideoMetadataExtended> {
86
105
  try {
87
106
  const args = ['-v', 'quiet', '-print_format', 'json', '-show_format', '-show_streams', videoPath];
88
- const { stdout } = await execFileAsync('ffprobe', args);
107
+ const { stdout } = await execActivityFile('ffprobe', args);
89
108
  const metadata = JSON.parse(stdout.toString()) as FFProbeOutput;
90
109
 
91
110
  const videoStream = metadata.streams.find((stream) => stream.codec_type === 'video');
@@ -113,6 +132,7 @@ async function getVideoMetadata(videoPath: string): Promise<VideoMetadataExtende
113
132
 
114
133
  return { duration, width, height, codec, bitrate, fps, hasAudio };
115
134
  } catch (error) {
135
+ rethrowIfActivityStopped(error);
116
136
  log.error(`Failed to get video metadata: ${error instanceof Error ? error.message : 'Unknown error'}`);
117
137
  if (error instanceof ApplicationFailure) {
118
138
  throw error;
@@ -184,6 +204,7 @@ async function generateRenditionWithDimensions(
184
204
  outputDir: string,
185
205
  metadata: VideoMetadataExtended,
186
206
  maxResolution: number,
207
+ videoPreset: VideoPreset,
187
208
  ): Promise<MediaResult | null> {
188
209
  const outputFile = path.join(outputDir, `rendition_${maxResolution}p.mp4`);
189
210
 
@@ -207,7 +228,7 @@ async function generateRenditionWithDimensions(
207
228
  '-c:v',
208
229
  'libx264', // H.264 codec
209
230
  '-preset',
210
- 'medium', // Balance between speed and compression
231
+ videoPreset,
211
232
  '-crf',
212
233
  VIDEO_CRF, // Constant Rate Factor (18-28, lower = better quality)
213
234
  '-pix_fmt',
@@ -226,7 +247,7 @@ async function generateRenditionWithDimensions(
226
247
  log.info(`Generating ${maxResolution}p video rendition`, { command: 'ffmpeg', args: command });
227
248
 
228
249
  try {
229
- const { stderr } = await execFileAsync('ffmpeg', command, { maxBuffer: FFMPEG_MAX_BUFFER });
250
+ const { stderr } = await execActivityFileWithProgress('ffmpeg', command);
230
251
  const stderrText = stderr.toString();
231
252
 
232
253
  if (stderrText && !stderrText.includes('frame=')) {
@@ -247,6 +268,7 @@ async function generateRenditionWithDimensions(
247
268
  return null;
248
269
  }
249
270
  } catch (error) {
271
+ rethrowIfActivityStopped(error);
250
272
  log.error(`Failed to generate video rendition: ${error instanceof Error ? error.message : 'Unknown error'}`);
251
273
  return null;
252
274
  }
@@ -273,7 +295,7 @@ async function generateAudioRendition(videoPath: string, outputDir: string): Pro
273
295
  log.info('Generating audio-only rendition', { command: 'ffmpeg', args: command });
274
296
 
275
297
  try {
276
- const { stderr } = await execFileAsync('ffmpeg', command, { maxBuffer: FFMPEG_MAX_BUFFER });
298
+ const { stderr } = await execActivityFileWithProgress('ffmpeg', command);
277
299
  const stderrText = stderr.toString();
278
300
 
279
301
  if (stderrText && !stderrText.includes('frame=')) {
@@ -290,6 +312,7 @@ async function generateAudioRendition(videoPath: string, outputDir: string): Pro
290
312
  return null;
291
313
  }
292
314
  } catch (error) {
315
+ rethrowIfActivityStopped(error);
293
316
  log.error(`Failed to generate audio rendition: ${error instanceof Error ? error.message : 'Unknown error'}`);
294
317
  return null;
295
318
  }
@@ -331,7 +354,7 @@ async function generateScreenshot(
331
354
  log.info(`Generating ${name} at ${timestamp}s`, { command: 'ffmpeg', args: command });
332
355
 
333
356
  try {
334
- const { stderr } = await execFileAsync('ffmpeg', command);
357
+ const { stderr } = await execActivityFile('ffmpeg', command);
335
358
  const stderrText = stderr.toString();
336
359
 
337
360
  if (stderrText && !stderrText.includes('frame=')) {
@@ -352,6 +375,7 @@ async function generateScreenshot(
352
375
  return null;
353
376
  }
354
377
  } catch (error) {
378
+ rethrowIfActivityStopped(error);
355
379
  log.error(`Failed to generate ${name}: ${error instanceof Error ? error.message : 'Unknown error'}`);
356
380
  return null;
357
381
  }
@@ -412,17 +436,22 @@ async function uploadMediaAsRendition(
412
436
  export async function prepareVideo(
413
437
  payload: DSLActivityExecutionPayload<PrepareVideoParams>,
414
438
  ): Promise<PrepareVideoResult> {
439
+ initializeActivityCommandDeadline();
415
440
  const { client, objectId, params } = await setupActivity<PrepareVideoParams>(payload);
416
441
 
417
442
  const maxResolution = params.maxResolution ?? DEFAULT_MAX_RESOLUTION;
418
443
  const thumbnailSize = params.thumbnailSize ?? DEFAULT_THUMBNAIL_SIZE;
419
444
  const posterSize = params.posterSize ?? DEFAULT_POSTER_SIZE;
420
445
  const generateAudio = params.generateAudio ?? DEFAULT_GENERATE_AUDIO;
446
+ const skipTranscode = params.skipTranscode ?? false;
447
+ const videoPreset = resolveVideoPreset(params.videoPreset ?? 'medium', Context.current().info.attempt);
421
448
 
422
449
  log.info(`Preparing video for ${objectId}`, {
423
450
  maxResolution,
424
451
  thumbnailSize,
425
452
  posterSize,
453
+ skipTranscode,
454
+ videoPreset,
426
455
  });
427
456
 
428
457
  // Retrieve the content object
@@ -454,13 +483,19 @@ export async function prepareVideo(
454
483
  const metadata = await getVideoMetadata(videoFile);
455
484
 
456
485
  // Step 2: Generate video rendition
457
- log.info('Generating video rendition');
458
- const renditionResult = await generateRenditionWithDimensions(
459
- videoFile,
460
- tempOutputDir,
461
- metadata,
462
- maxResolution,
463
- );
486
+ let renditionResult: MediaResult | null = null;
487
+ if (!shouldTranscodeVideo(skipTranscode)) {
488
+ log.info('Skipping video rendition transcode by configuration');
489
+ } else {
490
+ log.info('Generating video rendition');
491
+ renditionResult = await generateRenditionWithDimensions(
492
+ videoFile,
493
+ tempOutputDir,
494
+ metadata,
495
+ maxResolution,
496
+ videoPreset,
497
+ );
498
+ }
464
499
 
465
500
  // Step 3 & 4: Generate thumbnail and poster in parallel
466
501
  log.info('Generating thumbnail and poster');
@@ -578,6 +613,7 @@ export async function prepareVideo(
578
613
  status: 'success',
579
614
  };
580
615
  } catch (error) {
616
+ rethrowIfActivityStopped(error);
581
617
  const errorMessage = error instanceof Error ? error.message : 'Unknown error';
582
618
  log.error(`Error preparing video: ${errorMessage}`, { error });
583
619
 
@@ -3,6 +3,7 @@ import type { VertesiaClient } from '@vertesia/client';
3
3
  import { ContentEventName, type DSLActivityExecutionPayload } from '@vertesia/common';
4
4
  import { beforeAll, beforeEach, describe, expect, it, vi } from 'vitest';
5
5
  import type { ActivityContext } from '../../dsl/setup/ActivityContext.js';
6
+ import { execActivityFile } from './exec.js';
6
7
  import { type ProbeMediaStreamsParams, type ProbeMediaStreamsResult, probeMediaStreams } from './probeMediaStreams.js';
7
8
 
8
9
  vi.mock('../../dsl/setup/ActivityContext.js', async (importOriginal) => {
@@ -10,14 +11,12 @@ vi.mock('../../dsl/setup/ActivityContext.js', async (importOriginal) => {
10
11
  return { ...actual, setupActivity: vi.fn() };
11
12
  });
12
13
 
13
- // child_process.exec uses util.promisify.custom to return { stdout, stderr }.
14
- // vi.hoisted ensures these are defined before the vi.mock factory runs.
15
- const { execMock, execCustom } = vi.hoisted(() => {
16
- const custom = vi.fn();
17
- const mock = Object.assign(vi.fn(), { [Symbol.for('nodejs.util.promisify.custom')]: custom });
18
- return { execMock: mock, execCustom: custom };
14
+ // probeMediaStreams delegates ffprobe to the shared exec helper, so mock that seam and drive the parsed output
15
+ // directly. The helper's cancellation/timeout wrapping is covered by exec.test.ts.
16
+ vi.mock('./exec.js', async (importOriginal) => {
17
+ const actual = await importOriginal<typeof import('./exec.js')>();
18
+ return { ...actual, execActivityFile: vi.fn() };
19
19
  });
20
- vi.mock('child_process', () => ({ exec: execMock }));
21
20
 
22
21
  let testEnv: MockActivityEnvironment;
23
22
 
@@ -44,7 +43,7 @@ const createPayload = (objectId = 'test-object-id'): DSLActivityExecutionPayload
44
43
  });
45
44
 
46
45
  function mockExec(stdout: string) {
47
- execCustom.mockResolvedValue({ stdout, stderr: '' });
46
+ vi.mocked(execActivityFile).mockResolvedValue({ stdout, stderr: '' });
48
47
  }
49
48
 
50
49
  async function setupMockContext(objectId: string, signedUrl: string): Promise<void> {
@@ -1,12 +1,9 @@
1
- import { exec } from 'node:child_process';
2
- import { promisify } from 'node:util';
3
1
  import { ApplicationFailure, log } from '@temporalio/activity';
4
2
  import { RequestError } from '@vertesia/api-fetch-client';
5
3
  import type { DSLActivityExecutionPayload, DSLActivitySpec } from '@vertesia/common';
6
4
  import { setupActivity } from '../../dsl/setup/ActivityContext.js';
7
5
  import { DocumentNotFoundError } from '../../errors.js';
8
-
9
- const execAsync = promisify(exec);
6
+ import { execActivityFile, initializeActivityCommandDeadline, rethrowIfActivityStopped } from './exec.js';
10
7
 
11
8
  const FFPROBE_MAX_BUFFER = 1024 * 1024; // 1MB is more than enough for stream metadata JSON
12
9
 
@@ -32,6 +29,7 @@ interface FFProbeOutput {
32
29
  export async function probeMediaStreams(
33
30
  payload: DSLActivityExecutionPayload<ProbeMediaStreamsParams>,
34
31
  ): Promise<ProbeMediaStreamsResult> {
32
+ initializeActivityCommandDeadline();
35
33
  const { client, objectId } = await setupActivity<ProbeMediaStreamsParams>(payload);
36
34
 
37
35
  const inputObject = await client.objects.retrieve(objectId).catch((err: unknown) => {
@@ -54,12 +52,18 @@ export async function probeMediaStreams(
54
52
 
55
53
  // ffprobe reads only the container headers via HTTP range requests.
56
54
  // -probesize 32k caps the amount read from the network to ~32 KB.
55
+ // ffprobe emits no frame=/time= progress, so the stall watchdog does not apply here; the shared exec wrapper still
56
+ // propagates Activity cancellation and bounds the read by the Activity deadline, and passing the URL as an argv
57
+ // entry (not a shell string) removes the shell-interpolation surface.
57
58
  let stdout: string;
58
59
  try {
59
- ({ stdout } = await execAsync(`ffprobe -v quiet -probesize 32k -print_format json -show_streams "${url}"`, {
60
- maxBuffer: FFPROBE_MAX_BUFFER,
61
- }));
60
+ ({ stdout } = await execActivityFile(
61
+ 'ffprobe',
62
+ ['-v', 'quiet', '-probesize', '32k', '-print_format', 'json', '-show_streams', url],
63
+ { maxBuffer: FFPROBE_MAX_BUFFER },
64
+ ));
62
65
  } catch (err: unknown) {
66
+ rethrowIfActivityStopped(err);
63
67
  const message = err instanceof Error ? err.message : String(err);
64
68
  log.error(`ffprobe failed for object ${objectId}: ${message}`);
65
69
  throw new Error(`Failed to probe media streams for object ${objectId}: ${message}`);