videobook 0.3.0 → 0.3.1

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 (57) hide show
  1. package/CHANGELOG.md +73 -0
  2. package/dist/asset-media-projection.js +161 -0
  3. package/dist/command-reference.js +3 -1
  4. package/dist/diagnostics.js +1 -1
  5. package/dist/generation-reference-mentions.js +19 -0
  6. package/dist/goal-plan-v3.js +11 -1
  7. package/dist/goal-plan.js +1 -1
  8. package/dist/goal-store.js +2 -2
  9. package/dist/har-capture.js +36 -3
  10. package/dist/jobs/_image-edit-adapter.js +2 -0
  11. package/dist/jobs/assess-media-probes.js +94 -0
  12. package/dist/jobs/complete-provider-task.js +2 -5
  13. package/dist/jobs/generate-media.js +3 -2
  14. package/dist/jobs/image-subjects.js +2 -0
  15. package/dist/jobs/kie-paygate.js +3 -3
  16. package/dist/jobs/register-all.js +2 -0
  17. package/dist/jobs/run-notebook.js +16 -1
  18. package/dist/library-asset-projection.js +141 -0
  19. package/dist/lightning-defaults.js +4 -4
  20. package/dist/media/catalog.js +8 -2
  21. package/dist/media/descriptor.js +7 -4
  22. package/dist/media/frontier-gaps.js +0 -1
  23. package/dist/media/har-endpoint.js +25 -0
  24. package/dist/media/json-schema.js +3 -0
  25. package/dist/media/preferred-defaults.js +161 -0
  26. package/dist/media/prompt-dialects.js +9 -1
  27. package/dist/media-library.js +158 -153
  28. package/dist/notebook-sticky-settings.js +60 -0
  29. package/dist/tools/reconcile-project-work.js +34 -181
  30. package/dist/tools/videobook-domain.js +16 -1
  31. package/dist/usage-providers.js +3 -3
  32. package/package.json +15 -15
  33. package/ui/assets/{alive-captions-Cj8C1tCN.js → alive-captions-CnLK22ED.js} +1 -1
  34. package/ui/assets/confirm-bodies-BaalBxp1.js +1 -0
  35. package/ui/assets/{fullscreen-DyEncSdm.js → fullscreen-DTJQJ6dX.js} +1 -1
  36. package/ui/assets/{goal-ui-BdFeoSWh.js → goal-ui-CN-kZTEO.js} +1 -1
  37. package/ui/assets/index-CTPstLYq.css +1 -0
  38. package/ui/assets/index-Dp81yChE.js +37 -0
  39. package/ui/assets/inspector-jeyma7HL.js +1 -0
  40. package/ui/assets/{jobs-ui-BudI4w0r.js → jobs-ui-D6S3LJ4j.js} +1 -1
  41. package/ui/assets/{music-video-CT4C55ok.js → music-video-CDqr0X6G.js} +1 -1
  42. package/ui/assets/{ui-mq0uIS5P.css → ui-BBM3YCY6.css} +1 -1
  43. package/ui/assets/ui-Cqs1-YNg.js +2 -0
  44. package/ui/assets/vendor-BWf-WCsY.js +19 -0
  45. package/ui/assets/{vendor-flow-nqTFQmTh.js → vendor-flow-DPvi8ynA.js} +1 -1
  46. package/ui/assets/vendor-icons-D3KrGMJY.js +1 -0
  47. package/ui/assets/{vendor-mcp-D_C3EZPv.js → vendor-mcp-BSR9acpG.js} +2 -2
  48. package/ui/assets/videobook-b9VODNOA.js +21 -0
  49. package/ui/index.html +14 -14
  50. package/ui/assets/confirm-bodies-BxWXwDf0.js +0 -1
  51. package/ui/assets/index-DWfR79lA.css +0 -1
  52. package/ui/assets/index-QZ8lbemN.js +0 -37
  53. package/ui/assets/inspector-BC6xWkJU.js +0 -1
  54. package/ui/assets/ui-z5r-p0K9.js +0 -2
  55. package/ui/assets/vendor-D8XcMav4.js +0 -19
  56. package/ui/assets/vendor-icons-CGqhCIXV.js +0 -1
  57. package/ui/assets/videobook-CsW453sb.js +0 -21
package/CHANGELOG.md CHANGED
@@ -6,6 +6,79 @@ and the package uses [Semantic Versioning](https://semver.org/).
6
6
 
7
7
  ## [Unreleased]
8
8
 
9
+ ## [0.3.1] - 2026-10-25
10
+
11
+ ### Added
12
+
13
+ - KIE Kling 4.0 reference-to-video and keyframe-to-video, 3 to 30 seconds at
14
+ 720p or 1080p. Reference-to-video takes up to ten images and five videos,
15
+ and generates from the prompt alone when none are connected. Keyframe-to-video
16
+ takes a first frame, up to eight middle keyframes and a last frame, as many
17
+ as the duration allows. Prompts name connected media the way Kling 4.0 does:
18
+ `@image_1`, `@video_1`, `@first_frame`, `@keyframe_1` and `@last_frame`.
19
+ - KIE PixVerse V6 Max text-to-video and image-to-video, 1 to 15 seconds at
20
+ 360p to 1080p, with optional generated audio and a seed. Image-to-video
21
+ follows the image's aspect ratio.
22
+ - KIE OmniHuman 2.0, which animates one image to a driving audio track at 720p
23
+ for as long as the audio runs, with an optional prompt of up to 300
24
+ characters. KIE's watermark stays on by default (`enable_watermark`).
25
+ - The KIE Grok Imagine Video 1.5 preview, the KIE Grok Imagine endpoint that
26
+ takes first and last frames, a loop image (`loop_url`) and up to four timed
27
+ `keyframes`. It shares the Grok Imagine Video 1.5 image-to-video frontier
28
+ ranking. Kling 4.0, PixVerse V6 Max and OmniHuman 2.0 are on no top-10 board
29
+ and appear in the unranked group.
30
+
31
+ ### Changed
32
+
33
+ - Upgrade all `@apicity/*` dependencies to 0.18.1, which prices the new KIE
34
+ models. Its endpoint catalog now lists fal `v1.queue.status` and
35
+ `v1.queue.result`, with the other fal SDK methods written as method
36
+ shorthand, so `apicity har report` from `@apicity/cli` 0.18.1 can resolve the
37
+ fal polling captures Videobook writes. The requests sent to fal are
38
+ unchanged.
39
+ - Image and video models default to a 12-second duration (or the nearest
40
+ allowed length), a 9:16 aspect ratio and their highest resolution, quality
41
+ and bitrate. Each notebook remembers the last duration, aspect ratio,
42
+ resolution and quality a cell used, and applies them to new cells and model
43
+ switches. Goals saved before this change keep their 2K scene; new goals use
44
+ 4K.
45
+ - On desktop, Settings shows AI providers in one column and accounts, storage
46
+ and diagnostics in another; on phones the columns stack behind section jump
47
+ buttons. The chat input stays a plain bar on Settings.
48
+ - The Library opens from an Assets button in the notebook toolbar instead of
49
+ the header, and Arrange is renamed Clean.
50
+ - Library browsing pages assets before building their rows and no longer runs
51
+ ffprobe while listing; proxy checks and duration backfills run as background
52
+ jobs.
53
+ - A Post can take its published link without being saved to iPhone or Dropbox
54
+ first.
55
+
56
+ ### Fixed
57
+
58
+ - HAR captures of KIE's nested Veo and Suno SDK calls are named by their
59
+ apicity catalog dot paths, such as `api.v1.veo.generate` and
60
+ `api.v1.generate.sounds`, instead of their SDK paths, so `apicity har report`
61
+ resolves them. Sub-provider prefixes that are URL segments stay:
62
+ `kie.claude.post.v1.messages` is named `claude.v1.messages`.
63
+ - A generation that fails before reaching its provider, such as a request the
64
+ schema refuses, is captured under the provider's catalog endpoint, and fal,
65
+ Alibaba and KIE file uploads are captured as `storage.upload.initiate`,
66
+ `api.v1.uploads` and `api.fileStreamUpload`, so `apicity har report` resolves
67
+ them.
68
+ - A notebook generation whose provider fails after accepting the request now
69
+ marks its plan failed with the provider error. `list_notebook_generation_plans`
70
+ and the Jobs screen kept reporting such plans as awaiting output.
71
+ - v3 goals saved with a 2K scene load again and keep that resolution when they
72
+ are edited, approved or retried.
73
+ - Image subject edits check the provider key before uploading the source image,
74
+ so a missing or placeholder key no longer sends the image to a public file
75
+ host for a request that cannot run.
76
+ - Settings loads when the page opens on it, or when it opens before the MCP
77
+ connection is ready.
78
+ - The multyculty workflow and video poster recovery tests no longer time out on
79
+ loaded hosts, and the chat-composer browser test no longer reads the runner's
80
+ `.env` or provider keys and matches the ranked model picker.
81
+
9
82
  ## [0.3.0] - 2026-10-24
10
83
 
11
84
  ### Added
@@ -0,0 +1,161 @@
1
+ import { createHash } from "node:crypto";
2
+ import { findAudioFile, findPrimaryMediaFile, findVideoFile } from "videobook-engine";
3
+ import { logger } from "./logger.js";
4
+ import { queue } from "./queue.js";
5
+ import { scheduleMediaDurationBackfill, scheduleVideoProxyAssessment, } from "./jobs/assess-media-probes.js";
6
+ import { videoProxyAssessmentMatches, videoProxyMetadataMatches, } from "./video-proxy.js";
7
+ export const PARTIAL_DOWNLOAD_RE = /\.(mp4|mov)\.part$/i;
8
+ const DURATION_PROBE_EXT_RE = /\.(mp4|mov|webm|mkv|mp3|wav|m4a|aac|ogg|flac)$/i;
9
+ const IMAGE_FILE_RE = /\.(jpg|jpeg|png|webp|gif)$/i;
10
+ export function findAssetMediaFile(files, asset) {
11
+ return asset.kind === "audio"
12
+ ? findAudioFile(files, asset.kind)
13
+ : asset.kind === "video"
14
+ ? findVideoFile(files.filter((file) => file.name !== "proxy.mp4"))
15
+ : findPrimaryMediaFile(files, asset.kind);
16
+ }
17
+ // Content fingerprint for cache-busting media URLs. Pure function of the
18
+ // manifest tuples; partial downloads are excluded because their size grows on
19
+ // every poll while no served URL points at them.
20
+ function computeAssetFingerprint(files) {
21
+ const settled = files.filter((f) => !PARTIAL_DOWNLOAD_RE.test(f.name));
22
+ if (settled.length === 0)
23
+ return "";
24
+ const sorted = [...settled].sort((a, b) => a.name.localeCompare(b.name));
25
+ const hash = createHash("sha1");
26
+ for (const f of sorted) {
27
+ hash.update(f.name);
28
+ hash.update("\0");
29
+ hash.update(String(f.sizeBytes ?? 0));
30
+ hash.update("\0");
31
+ hash.update(f.objectHash);
32
+ hash.update("\0");
33
+ }
34
+ return hash.digest("hex").slice(0, 12);
35
+ }
36
+ export function indexByArtifactId(rows) {
37
+ const map = new Map();
38
+ for (const row of rows)
39
+ map.set(row.artifactId, row);
40
+ return map;
41
+ }
42
+ export function groupByArtifactId(rows) {
43
+ const map = new Map();
44
+ for (const row of rows) {
45
+ const list = map.get(row.artifactId);
46
+ if (list)
47
+ list.push(row);
48
+ else
49
+ map.set(row.artifactId, [row]);
50
+ }
51
+ return map;
52
+ }
53
+ function streamSeconds(stream) {
54
+ return stream.durationTicks * stream.timeBase.numerator / stream.timeBase.denominator;
55
+ }
56
+ export function assetStreamProfile(streams, meta) {
57
+ const videoStream = streams.find((stream) => stream.kind === "video" && stream.video);
58
+ const audioStream = streams.find((stream) => stream.kind === "audio" && stream.audio);
59
+ const frameRate = videoStream?.video?.averageFrameRate
60
+ ?? videoStream?.video?.nominalFrameRate;
61
+ const streamDuration = videoStream ? streamSeconds(videoStream) : 0;
62
+ const metadataDuration = Number(meta?.duration);
63
+ const sourceDuration = streamDuration > 0
64
+ ? streamDuration
65
+ : Number.isFinite(metadataDuration) ? metadataDuration : 0;
66
+ const sourceProfile = sourceDuration > 0
67
+ ? {
68
+ videoDuration: sourceDuration,
69
+ avgFps: frameRate ? frameRate.numerator / frameRate.denominator : null,
70
+ hasAudio: audioStream !== undefined || meta?.hasAudio === true,
71
+ }
72
+ : null;
73
+ return { videoStream, audioStream, frameRate, sourceProfile };
74
+ }
75
+ function isActiveWork(work) {
76
+ return work !== null && (work.status === "pending" || work.status === "working");
77
+ }
78
+ export function servedVideoProxyFile(input) {
79
+ if (input.artifact.kind !== "video" || !input.mediaFile)
80
+ return null;
81
+ const proxyFile = input.files.find((file) => file.name === "proxy.mp4");
82
+ if (!proxyFile)
83
+ return null;
84
+ const metadata = input.proxyMetadata;
85
+ if (metadata.ok && videoProxyMetadataMatches(metadata.value, input.mediaFile, proxyFile)) {
86
+ return proxyFile;
87
+ }
88
+ const assessed = metadata.ok
89
+ && videoProxyAssessmentMatches(metadata.value, input.mediaFile, proxyFile);
90
+ if (!assessed && !isActiveWork(input.work) && input.sourceProfile) {
91
+ scheduleVideoProxyAssessment({
92
+ bookId: input.bookId,
93
+ artifactId: input.artifact.artifactId,
94
+ sourceFile: input.mediaFile,
95
+ proxyFile,
96
+ sourceProfile: input.sourceProfile,
97
+ });
98
+ }
99
+ return null;
100
+ }
101
+ export function assetDuration(input) {
102
+ const stored = input.meta?.duration;
103
+ if (typeof stored === "number")
104
+ return stored;
105
+ const stream = input.profile.videoStream ?? input.profile.audioStream;
106
+ const measured = stream ? streamSeconds(stream) : 0;
107
+ if (measured > 0)
108
+ return measured;
109
+ if (input.meta && input.mediaFile && DURATION_PROBE_EXT_RE.test(input.mediaFile.name)) {
110
+ scheduleMediaDurationBackfill({
111
+ bookId: input.bookId,
112
+ artifactId: input.artifactId,
113
+ mediaFile: input.mediaFile,
114
+ });
115
+ }
116
+ return null;
117
+ }
118
+ function mediaCacheBust(kind, files, mediaFile) {
119
+ if (kind === "video" || kind === "image") {
120
+ const fingerprint = computeAssetFingerprint(files);
121
+ return fingerprint ? `?v=${fingerprint}` : "";
122
+ }
123
+ const mediaHash = mediaFile?.objectHash;
124
+ return mediaHash ? `?v=${mediaHash.slice(0, 12)}` : "";
125
+ }
126
+ export function assetMediaUrls(input) {
127
+ const { artifact, files } = input;
128
+ const base = `/files/${input.bookId}/${artifact.artifactId}`;
129
+ const cacheBust = mediaCacheBust(artifact.kind, files, input.mediaFile);
130
+ const characterImage = artifact.kind === "character"
131
+ ? files.find((file) => IMAGE_FILE_RE.test(file.name)) ?? null
132
+ : null;
133
+ const served = characterImage ?? input.proxyFile ?? input.mediaFile;
134
+ const media_url = served ? `${base}/${served.name}${cacheBust}` : null;
135
+ if (characterImage)
136
+ return { media_url, thumbnail_url: media_url };
137
+ const hasThumbnail = files.some((file) => file.name === "original.jpg");
138
+ const thumbnail_url = hasThumbnail
139
+ ? `${base}/original.jpg${cacheBust}`
140
+ : artifact.kind === "image" ? media_url : null;
141
+ return { media_url, thumbnail_url };
142
+ }
143
+ export async function scheduleMissingVideoPoster(engine, input) {
144
+ const { artifact, mediaFile } = input;
145
+ if (artifact.kind !== "video" || input.status !== "ready" || !mediaFile)
146
+ return;
147
+ if (input.files.some((file) => file.name === "original.jpg"))
148
+ return;
149
+ const recovery = await engine.metadata.artifacts.read(artifact.artifactId, "poster-recovery");
150
+ if (recovery.ok && recovery.value.failed && recovery.value.sourceHash === mediaFile.objectHash)
151
+ return;
152
+ void queue.enqueue(`extract-video-poster-${input.bookId}-${artifact.artifactId}-${mediaFile.objectHash}`, {
153
+ type: "extract_video_poster",
154
+ bookId: input.bookId,
155
+ posterArtifactId: artifact.artifactId,
156
+ sourceFilename: mediaFile.name,
157
+ sourceHash: mediaFile.objectHash,
158
+ }).catch((error) => {
159
+ logger.error("poster-enqueue-failed", { assetId: artifact.artifactId, error });
160
+ });
161
+ }
@@ -90,7 +90,7 @@ export function renderGenerationModels() {
90
90
 
91
91
  Generated by \`npm run docs:commands\` from the media catalog (\`server/src/media/catalog.ts\`). Call \`generate_image\`, \`generate_video\` or \`generate_audio\` with one of these model ids; \`list_generation_models\` returns the same list with every option, enum, range and default read from the apicity schemas.
92
92
 
93
- Image and video pickers rank exact catalog endpoints using the committed top-10 snapshots in @apicity/cost 0.17.1, originally sourced from Artificial Analysis. No leaderboard is fetched at runtime. Rankings are per operation: t2i, i2i, t2v, i2v, v2v, and r2v (derived from the text/image-to-video boards). Each ranked option shows rank, Elo, and a bundled price label; prices are indicative ranges, not a quote for the selected settings.
93
+ Image and video pickers rank exact catalog endpoints using the committed top-10 snapshots in @apicity/cost 0.18.1, originally sourced from Artificial Analysis. No leaderboard is fetched at runtime. Rankings are per operation: t2i, i2i, t2v, i2v, v2v, and r2v (derived from the text/image-to-video boards). Each ranked option shows rank, Elo, and a bundled price label; prices are indicative ranges, not a quote for the selected settings.
94
94
 
95
95
  Configured frontier models appear first; other models retain their existing order in a separate group. Providers without credentials are marked unavailable. Saved selections and defaults stay unchanged. Catalog variants on a frontier board are included even if previously hidden; non-frontier hidden variants remain hidden. A model version or lite variant never inherits another endpoint's score.
96
96
 
@@ -98,6 +98,8 @@ Kling 4 supports text, frames, references and video editing; it appears in the u
98
98
 
99
99
  HeyGen Avatar 3 and 4 digital twins are text-to-video with a stock avatar (Abigail Office Front) and voice (Ivy); the prompt is the spoken script. Avatar 5 uses the SDK defaults Abigail Sofa Front and Warm Pro Narrator. Avatar 4 image-to-video animates an image with the script and defaults to Ivy. Connected audio overrides speech synthesis. Video Agent v2 and v3 accept a production prompt. HeyGen lipsync, translation, filler-word removal, HeyGen Voice TTS and Ideogram vector endpoints are outside this catalog addition.
100
100
 
101
+ Kie Kling 4.0 reference-to-video and keyframe-to-video, PixVerse V6 Max text-to-video and image-to-video, and OmniHuman 2.0 are on no top-10 board, so they appear in the unranked group. Kling 4.0 prompts name connected media as Kling 4.0 documents them: \`@image_1\` and \`@video_1\` for references, \`@first_frame\`, \`@keyframe_1\` and \`@last_frame\` for frames; with no media connected, both endpoints generate from the prompt alone. PixVerse V6 Max image-to-video follows the image's aspect ratio. OmniHuman 2.0 renders as long as its driving audio. The Kie Grok Imagine Video 1.5 preview takes first and last frames, a loop image (\`loop_url\`) and up to four timed \`keyframes\`, and shares the Grok Imagine Video 1.5 image-to-video ranking.
102
+
101
103
  Top-10 coverage policy: every registry endpoint must map exactly to a catalog entry or appear in the known gaps below. Models with no provider endpoint in the registry: ${unmapped.join(", ")}. Catalog gaps do not imply provider API support is missing; adding those endpoints is separate work.
102
104
 
103
105
  | Known endpoint gap | Reason |
@@ -14,7 +14,7 @@ const EVENTS = [
14
14
  "process-video-enqueue-failed", "waveform-write-failed", "timeline-reencode",
15
15
  "pad-video-started", "video-analysis-failed", "description-truncated",
16
16
  "description-cache-failed", "watermark-placement-fallback", "waveform-enqueue-failed",
17
- "poster-enqueue-failed",
17
+ "poster-enqueue-failed", "proxy-assessment-enqueue-failed", "duration-backfill-enqueue-failed",
18
18
  "watermark-probe-fallback", "generation-error-clear-failed", "timeline-orientation-failed",
19
19
  "generation-cost-estimate", "job-error", "semantic-index-deferred",
20
20
  "tag-asset-failed", "voice-session-closed", "alive-captions-regions-failed",
@@ -65,6 +65,25 @@ export function rewriteSeedanceAssetMentions(prompt, refs) {
65
65
  }
66
66
  return replaceNotebookMentions(prompt, replacements);
67
67
  }
68
+ function numberedMentions(refs, marker) {
69
+ return refMentionEntries(refs).map(({ ref, index }) => ({
70
+ reference: ref,
71
+ replacement: `@${marker}_${index + 1}`,
72
+ }));
73
+ }
74
+ export function rewriteKling4ReferenceMentions(prompt, imageRefs, videoRefs) {
75
+ return replaceNotebookMentions(prompt, [
76
+ ...numberedMentions(imageRefs, "image"),
77
+ ...numberedMentions(videoRefs, "video"),
78
+ ]);
79
+ }
80
+ export function rewriteKling4FrameMentions(prompt, frames) {
81
+ return replaceNotebookMentions(prompt, [
82
+ ...frames.first.slice(0, 1).map((reference) => ({ reference, replacement: "@first_frame" })),
83
+ ...numberedMentions(frames.keyframes, "keyframe"),
84
+ ...frames.last.slice(0, 1).map((reference) => ({ reference, replacement: "@last_frame" })),
85
+ ]);
86
+ }
68
87
  export function rewriteViduReferenceMentions(prompt, imageRefs, audioRefs) {
69
88
  return replaceNotebookMentions(prompt, [
70
89
  ...refMentionEntries(imageRefs).map(({ ref, index }) => ({
@@ -4,6 +4,8 @@ import { goalPromptWords } from "./goal-prompt-budget.js";
4
4
  import { LIGHTNING_COLD_OPEN_SECONDS, LIGHTNING_SCENE_ASPECT_RATIO, LIGHTNING_SCENE_MODEL, LIGHTNING_SCENE_OUTPUT_FORMAT, LIGHTNING_SCENE_PROMPT_WORDS, LIGHTNING_SCENE_RESOLUTION, LIGHTNING_VIDEO_DURATION_SECONDS, LIGHTNING_VIDEO_GENERATE_AUDIO, LIGHTNING_VIDEO_OUTPUT_FORMAT, LIGHTNING_VIDEO_PROMPT_WORDS, LIGHTNING_VIDEO_RESOLUTION, LIGHTNING_VIDEO_MODEL, LIGHTNING_WATERMARK_PROFILE, preservedPrompt, } from "./lightning-defaults.js";
5
5
  const IMAGE_LIMIT = 30;
6
6
  const VIDEO_LIMIT = 10;
7
+ const LEGACY_SCENE_RESOLUTION = "2K";
8
+ const SCENE_RESOLUTIONS = [LEGACY_SCENE_RESOLUTION, "4K"];
7
9
  export const goalMediaRefSchema = z.object({
8
10
  kind: z.enum(["image", "video"]),
9
11
  artifactId: z.string().min(1),
@@ -22,7 +24,7 @@ export const goalPlanV3Schema = z.object({
22
24
  tool: z.literal("generate_image"),
23
25
  model: z.literal(LIGHTNING_SCENE_MODEL),
24
26
  aspectRatio: z.enum(["16:9", "9:16", "1:1"]),
25
- resolution: z.literal(LIGHTNING_SCENE_RESOLUTION),
27
+ resolution: z.enum(SCENE_RESOLUTIONS),
26
28
  outputFormat: z.literal(LIGHTNING_SCENE_OUTPUT_FORMAT),
27
29
  imageRefs: z.array(goalMediaRefSchema).max(IMAGE_LIMIT),
28
30
  }).strict(),
@@ -59,6 +61,14 @@ export const goalPlanV3Schema = z.object({
59
61
  }).strict(),
60
62
  }).strict(),
61
63
  }).strict();
64
+ function withLegacySceneResolution(plan) {
65
+ if (typeof plan !== "object" || plan === null || !("version" in plan) || plan.version !== 3)
66
+ return plan;
67
+ if (!("scene" in plan) || typeof plan.scene !== "object" || plan.scene === null || "resolution" in plan.scene)
68
+ return plan;
69
+ return { ...plan, scene: { ...plan.scene, resolution: LEGACY_SCENE_RESOLUTION } };
70
+ }
71
+ export const savedGoalPlanV3Schema = z.preprocess(withLegacySceneResolution, goalPlanV3Schema);
62
72
  export function isGoalPlanV3(plan) {
63
73
  return plan?.version === 3;
64
74
  }
package/dist/goal-plan.js CHANGED
@@ -2,7 +2,7 @@ import { z } from "zod";
2
2
  import { LIGHTNING_VIDEO_MODEL } from "./lightning-defaults.js";
3
3
  import { GOAL_PROMPT_BUDGET, goalPromptWords } from "./goal-prompt-budget.js";
4
4
  import { GOAL_STYLE_IDS, goalStyleFragment, resolveGoalStyleOptions } from "./goal-styles.js";
5
- export { compileGoalV3Approval, compileGoalV3Steps, goalMediaRefSchema, goalPlanV3Schema, isGoalPlanV3, refreshGoalPlanV3 } from "./goal-plan-v3.js";
5
+ export { compileGoalV3Approval, compileGoalV3Steps, goalMediaRefSchema, goalPlanV3Schema, isGoalPlanV3, refreshGoalPlanV3, savedGoalPlanV3Schema } from "./goal-plan-v3.js";
6
6
  export { GOAL_PROMPT_BUDGET, goalPromptWords } from "./goal-prompt-budget.js";
7
7
  export const GOAL_VIDEO_TOOL = "generate_video";
8
8
  export const GOAL_VIDEO_MODEL = LIGHTNING_VIDEO_MODEL;
@@ -1,6 +1,6 @@
1
1
  import { z } from "zod";
2
2
  import { firstEmptyNotebookGridSlot, notebookOccupiedSlots } from "videobook-engine/notebook-grid";
3
- import { goalMediaRefSchema, goalPlanSchema, goalPlanV3Schema, goalStyleSchema, legacyGoalPlanSchema } from "./goal-plan.js";
3
+ import { goalMediaRefSchema, goalPlanSchema, goalStyleSchema, legacyGoalPlanSchema, savedGoalPlanV3Schema } from "./goal-plan.js";
4
4
  import { mutateNotebook } from "./notebook-mutations.js";
5
5
  import { arrangeGoalNotebook } from "./notebook-goal-layout.js";
6
6
  const KEY = "application.goals.v2";
@@ -68,7 +68,7 @@ function mediaFrom(value) {
68
68
  }
69
69
  export function goalContent(cell) {
70
70
  const inputs = cell.inputs ?? {};
71
- const plan = z.union([goalPlanV3Schema, goalPlanSchema, legacyGoalPlanSchema]).safeParse(inputs.plan);
71
+ const plan = z.union([savedGoalPlanV3Schema, goalPlanSchema, legacyGoalPlanSchema]).safeParse(inputs.plan);
72
72
  const requestedStyle = goalStyleSchema.safeParse(inputs.requestedStyle);
73
73
  const media = mediaFrom(inputs.media);
74
74
  return {
@@ -4,6 +4,7 @@ import { createHarCapture } from "@apicity/har";
4
4
  import { currentBookContext } from "./book-context.js";
5
5
  import { harDirectory, readHarSettings } from "./har-settings.js";
6
6
  import { logger } from "./logger.js";
7
+ const localFailures = new AsyncLocalStorage();
7
8
  const pending = new Set();
8
9
  let maintenance = Promise.resolve();
9
10
  export async function maintainHarCaptures(operation) {
@@ -51,11 +52,36 @@ export async function withHarCapture(meta, operation, base = globalThis.fetch, c
51
52
  try {
52
53
  return await runCapture(meta, operation, base, context);
53
54
  }
55
+ catch (error) {
56
+ localFailures.getStore()?.add(error);
57
+ throw error;
58
+ }
54
59
  finally {
55
60
  finish();
56
61
  pending.delete(finished);
57
62
  }
58
63
  }
64
+ export async function withHarLocalFailure(meta, operation) {
65
+ const failures = localFailures.getStore() ?? new Set();
66
+ return localFailures.run(failures, async () => {
67
+ try {
68
+ return await operation();
69
+ }
70
+ catch (error) {
71
+ const seen = new Set();
72
+ let cause = error;
73
+ while (!seen.has(cause)) {
74
+ if (failures.has(cause))
75
+ throw error;
76
+ seen.add(cause);
77
+ if (!(cause instanceof Error) || cause.cause === undefined)
78
+ break;
79
+ cause = cause.cause;
80
+ }
81
+ return withHarCapture(meta, () => { throw error; });
82
+ }
83
+ });
84
+ }
59
85
  async function runCapture(meta, operation, base, context) {
60
86
  const settings = await readHarSettings();
61
87
  if (settings.mode === "off")
@@ -99,8 +125,15 @@ async function runCapture(meta, operation, base, context) {
99
125
  }
100
126
  }
101
127
  const HTTP_METHODS = new Set(["get", "post", "put", "patch", "delete", "head", "options"]);
102
- function endpointDotPath(path) {
103
- return (path.length > 1 && HTTP_METHODS.has(path[0] ?? "") ? path.slice(1) : path).join(".");
128
+ const SDK_ONLY_NAMESPACES = new Map([["kie", new Set(["veo", "suno"])]]);
129
+ function endpointDotPath(provider, path) {
130
+ const [head = "", next = ""] = path;
131
+ if (path.length > 1 && HTTP_METHODS.has(head))
132
+ return path.slice(1).join(".");
133
+ if (path.length < 3 || !HTTP_METHODS.has(next))
134
+ return path.join(".");
135
+ const rest = path.slice(2);
136
+ return (SDK_ONLY_NAMESPACES.get(provider)?.has(head) ? rest : [head, ...rest]).join(".");
104
137
  }
105
138
  const providerCalls = new WeakMap();
106
139
  export function withProviderCall(provider, meta, operation) {
@@ -119,7 +152,7 @@ export function capturedProvider(provider, factory, base = globalThis.fetch) {
119
152
  function wrap(target, path) {
120
153
  return new Proxy(target, {
121
154
  apply(fn, receiver, args) {
122
- return withHarCapture({ provider, endpoint: endpointDotPath(path), input: args.length === 1 ? args[0] : args }, (fetch) => transports.run(fetch, () => Reflect.apply(fn, receiver, args)), base);
155
+ return withHarCapture({ provider, endpoint: endpointDotPath(provider, path), input: args.length === 1 ? args[0] : args }, (fetch) => transports.run(fetch, () => Reflect.apply(fn, receiver, args)), base);
123
156
  },
124
157
  get(object, property, receiver) {
125
158
  const value = Reflect.get(object, property, receiver);
@@ -99,6 +99,7 @@ function kieImageEditAdapter(model, build) {
99
99
  });
100
100
  return {
101
101
  id: model.apiId,
102
+ envKey: "KIE_API_KEY",
102
103
  uploadAsset: (engine, artifactId, filename) => uploadAssetToKie(engine, artifactId, filename),
103
104
  buildSemanticRequest: (input) => buildRequest({ ...input, imageUrls: [input.imageUrls[0]] }),
104
105
  buildRegionRequest: buildRequest,
@@ -166,6 +167,7 @@ function falImageEditAdapter(model) {
166
167
  const buildRequest = (input) => toProviderRequest(model.apiId, catalogEditRequest(model, input));
167
168
  return {
168
169
  id: model.apiId,
170
+ envKey: "FAL_API_KEY",
169
171
  uploadAsset: (engine, artifactId, filename) => uploadAssetToFal(engine, artifactId, filename),
170
172
  buildSemanticRequest: (input) => buildRequest({ ...input, imageUrls: [input.imageUrls[0]] }),
171
173
  buildRegionRequest: buildRequest,
@@ -0,0 +1,94 @@
1
+ import { join } from "node:path";
2
+ import { probeVideo } from "../ffmpeg.js";
3
+ import { logger } from "../logger.js";
4
+ import { queue, registerJobHandler } from "../queue.js";
5
+ import { invalidVideoProxyMetadata, probeStoredVideoFile, videoProxyAssessmentMatches, videoProxyMetadata, videoProxyValidationError, } from "../video-proxy.js";
6
+ const PROXY_FILENAME = "proxy.mp4";
7
+ export function scheduleVideoProxyAssessment(input) {
8
+ const jobId = `assess-video-proxy-${input.bookId}-${input.artifactId}-${input.proxyFile.objectHash}`;
9
+ void queue.enqueue(jobId, {
10
+ type: "assess_video_proxy",
11
+ bookId: input.bookId,
12
+ proxyAssetId: input.artifactId,
13
+ sourceFilename: input.sourceFile.name,
14
+ sourceHash: input.sourceFile.objectHash,
15
+ proxyHash: input.proxyFile.objectHash,
16
+ sourceProfile: input.sourceProfile,
17
+ }).catch((error) => {
18
+ logger.error("proxy-assessment-enqueue-failed", { assetId: input.artifactId, error });
19
+ });
20
+ }
21
+ export function scheduleMediaDurationBackfill(input) {
22
+ const jobId = `backfill-media-duration-${input.bookId}-${input.artifactId}-${input.mediaFile.objectHash}`;
23
+ void queue.enqueue(jobId, {
24
+ type: "backfill_media_duration",
25
+ bookId: input.bookId,
26
+ durationAssetId: input.artifactId,
27
+ mediaFilename: input.mediaFile.name,
28
+ mediaHash: input.mediaFile.objectHash,
29
+ }).catch((error) => {
30
+ logger.error("duration-backfill-enqueue-failed", { assetId: input.artifactId, error });
31
+ });
32
+ }
33
+ async function currentFile(engine, artifactId, name, objectHash) {
34
+ const manifest = await engine.files.manifest(artifactId);
35
+ if (!manifest.ok)
36
+ return null;
37
+ return manifest.value.files.find((file) => file.name === name && file.objectHash === objectHash) ?? null;
38
+ }
39
+ export async function assessStoredVideoProxy(engine, assetId, sourceFile, proxyFile, sourceProfile) {
40
+ let assessment;
41
+ try {
42
+ const proxy = await probeStoredVideoFile(engine, assetId, proxyFile.name);
43
+ const error = videoProxyValidationError(sourceProfile, proxy);
44
+ assessment = error
45
+ ? invalidVideoProxyMetadata(sourceFile, proxyFile, error)
46
+ : videoProxyMetadata(sourceFile, proxyFile, proxy);
47
+ }
48
+ catch (cause) {
49
+ const error = cause instanceof Error ? cause.message : String(cause);
50
+ assessment = invalidVideoProxyMetadata(sourceFile, proxyFile, error);
51
+ }
52
+ await engine.metadata.artifacts.write(assetId, "proxy", assessment);
53
+ return assessment.valid;
54
+ }
55
+ async function runVideoProxyAssessment(engine, payload) {
56
+ const { proxyAssetId } = payload;
57
+ const sourceFile = await currentFile(engine, proxyAssetId, payload.sourceFilename, payload.sourceHash);
58
+ const proxyFile = await currentFile(engine, proxyAssetId, PROXY_FILENAME, payload.proxyHash);
59
+ if (!sourceFile || !proxyFile)
60
+ return { skipped: true };
61
+ const existing = await engine.metadata.artifacts.read(proxyAssetId, "proxy");
62
+ if (existing.ok && videoProxyAssessmentMatches(existing.value, sourceFile, proxyFile)) {
63
+ return { skipped: true };
64
+ }
65
+ return {
66
+ valid: await assessStoredVideoProxy(engine, proxyAssetId, sourceFile, proxyFile, payload.sourceProfile),
67
+ };
68
+ }
69
+ async function runMediaDurationBackfill(engine, payload, signal) {
70
+ const { durationAssetId } = payload;
71
+ const metadata = await engine.metadata.artifacts.read(durationAssetId, "original");
72
+ if (!metadata.ok || typeof metadata.value.duration === "number")
73
+ return { skipped: true };
74
+ if (!await currentFile(engine, durationAssetId, payload.mediaFilename, payload.mediaHash)) {
75
+ return { skipped: true };
76
+ }
77
+ const dir = await engine.workspaces.resolveArtifact(durationAssetId);
78
+ if (!dir.ok)
79
+ throw new Error(`backfill_media_duration: ${dir.error.message}`);
80
+ const probe = await probeVideo(join(dir.value, payload.mediaFilename), { signal });
81
+ if (!Number.isFinite(probe.duration) || probe.duration <= 0)
82
+ return { skipped: true };
83
+ const written = await engine.metadata.artifacts.write(durationAssetId, "original", {
84
+ ...metadata.value,
85
+ duration: probe.duration,
86
+ });
87
+ if (!written.ok)
88
+ throw new Error(written.error.message);
89
+ return { duration: probe.duration };
90
+ }
91
+ export function registerAssessMediaProbeJobs(engine) {
92
+ registerJobHandler("assess_video_proxy", async (payload) => runVideoProxyAssessment(engine, payload));
93
+ registerJobHandler("backfill_media_duration", async (payload, { signal }) => runMediaDurationBackfill(engine, payload, signal));
94
+ }
@@ -1,4 +1,3 @@
1
- import { withHarCapture } from "../har-capture.js";
2
1
  import { join } from "node:path";
3
2
  import { tmpdir } from "node:os";
4
3
  import { writeFile, unlink, readFile as readFsFile } from "node:fs/promises";
@@ -48,9 +47,7 @@ async function reclaimFailedCompletion(engine, task, ctx) {
48
47
  return lease.ownerId;
49
48
  }
50
49
  export function registerCompleteProviderTaskJob(engine) {
51
- registerJobHandler("complete_kie_task", async (payload, ctx) => withHarCapture({
52
- provider: String(payload.taskType).split("_")[0] ?? "unknown", endpoint: "generation.complete", input: payload,
53
- }, async () => {
50
+ registerJobHandler("complete_kie_task", async (payload, ctx) => {
54
51
  const { taskId, taskType, assetId, bookId, meta, resultUrls, resultObject, } = payload;
55
52
  // Pattern C: inherit the provider lease via getOwner, abort if the row
56
53
  // already terminated, and register the lease so the wrapper's catch path
@@ -128,7 +125,7 @@ export function registerCompleteProviderTaskJob(engine) {
128
125
  clearInterval(heartbeat);
129
126
  await engine.jobs.locks.release(assetDir);
130
127
  }
131
- }, globalThis.fetch, { providerTaskId: payload.taskId, model: payload.meta.model, runId: payload.meta.runId, attemptId: payload.meta.attemptId }));
128
+ });
132
129
  }
133
130
  async function mediaOutputs(taskType, taskId, urls) {
134
131
  if (taskType === MEDIA_TASK_TYPES.machgen)
@@ -1,4 +1,5 @@
1
- import { withHarCapture } from "../har-capture.js";
1
+ import { withHarLocalFailure } from "../har-capture.js";
2
+ import { generationHarEndpoint } from "../media/har-endpoint.js";
2
3
  import { withBookContext } from "../book-context.js";
3
4
  import { isSeedanceEdit } from "../media/seedance-edit.js";
4
5
  import { registerJobHandler } from "../queue.js";
@@ -87,7 +88,7 @@ async function generateMedia(engine, job, ctx) {
87
88
  }
88
89
  export async function runGenerateMedia(engine, job, ctx) {
89
90
  return withBookContext({ bookId: job.bookId, engine, operation: "generate_media", artifactId: job.assetId,
90
- attemptId: job.generationId ?? job.assetId, runId: job.runId, model: job.model }, () => withHarCapture({ provider: resolveModel(job.model).provider, endpoint: "generation.submit", input: job }, () => generateMedia(engine, job, ctx)));
91
+ attemptId: job.generationId ?? job.assetId, runId: job.runId, model: job.model }, () => withHarLocalFailure({ provider: resolveModel(job.model).provider, endpoint: generationHarEndpoint(resolveModel(job.model)), input: job }, () => generateMedia(engine, job, ctx)));
91
92
  }
92
93
  export function registerGenerateMediaJob(engine) {
93
94
  registerJobHandler("generate_media", (payload, ctx) => runGenerateMedia(engine, payload, ctx));
@@ -4,6 +4,7 @@ import sharp from "sharp";
4
4
  import { KIE_RESULT_TIMEOUT_MS, kieProvider, parseResultUrls, waitForKieResult, } from "./_kie-poll.js";
5
5
  import { GROK_EXACT_MASK_ADAPTER_ID, imageEditAdapter, IMAGE_EDIT_ADAPTERS, } from "./_image-edit-adapter.js";
6
6
  import { DEFAULT_IMAGE_EDIT_MODEL_ID } from "../image-edit-models.js";
7
+ import { requireApiKey } from "../api-key.js";
7
8
  import { compositeImageRegionEdit, imageRegionEditPrompt } from "../image-region-edit.js";
8
9
  import { IMAGE_SUBJECT_MAP_VERSION, applyGrokSegmentsToSubjects, composeImageSubjectRemoval, exactEditMaskIndexes, grokMaskIndex, parseGrokSegmentMapResult, persistSeedreamSubjectLayers, readImageSubjectMapRecord, resolveImageSubjectSource, writeCurrentImageSubjectAnalysisRecord, } from "../image-subject-map.js";
9
10
  import { enqueueSearchIndex } from "./enqueue-search-index.js";
@@ -534,6 +535,7 @@ export async function previewImageSubjectEditRequest(engine, payload, imageUrls)
534
535
  : adapter.buildSemanticRequest(input);
535
536
  }
536
537
  async function runAdapterImageSubjectEdit(engine, payload, adapter, route, prepared, sourceFilename, signal) {
538
+ requireApiKey(adapter.envKey);
537
539
  const stored = await readStoredImageSubjectEditTask(engine, payload.imageAssetId);
538
540
  const storedMatch = stored?.adapterId === adapter.id ? stored : null;
539
541
  let request = storedMatch?.request ?? null;
@@ -1,4 +1,4 @@
1
- import { withHarCapture } from "../har-capture.js";
1
+ import { withHarLocalFailure } from "../har-capture.js";
2
2
  import { createKie } from "../usage-providers.js";
3
3
  import { submitMediaJob } from "../usage-providers.js";
4
4
  import { mintOtp } from "@apicity/cost";
@@ -177,8 +177,8 @@ async function submitQueuedKieTaskInner(engine, task) {
177
177
  return writeResult.value.task;
178
178
  }
179
179
  export async function submitKieTask(engine, ctx, args) {
180
- return withHarCapture({ provider: "kie", endpoint: "generation.paygate", input: args.request }, () => submitKieTaskInner(engine, ctx, args));
180
+ return withHarLocalFailure({ provider: "kie", endpoint: "api.v1.jobs.createTask", input: args.request }, () => submitKieTaskInner(engine, ctx, args));
181
181
  }
182
182
  export async function submitQueuedKieTask(engine, task) {
183
- return withHarCapture({ provider: "kie", endpoint: "generation.queued", input: task }, () => submitQueuedKieTaskInner(engine, task));
183
+ return withHarLocalFailure({ provider: "kie", endpoint: "api.v1.jobs.createTask", input: task }, () => submitQueuedKieTaskInner(engine, task));
184
184
  }
@@ -38,6 +38,7 @@ import { registerLabelMusicSectionsJob } from "./label-music-sections.js";
38
38
  import { registerEditingMomentsJob } from "./analyze-editing-moments.js";
39
39
  import { registerExtractWaveformJob } from "./extract-waveform.js";
40
40
  import { registerExtractVideoPosterJob } from "./extract-video-poster.js";
41
+ import { registerAssessMediaProbeJobs } from "./assess-media-probes.js";
41
42
  import { registerReplaceVideoAudioJob } from "./replace-video-audio.js";
42
43
  import { registerSpliceVideoJob } from "./splice-video.js";
43
44
  import { registerSmoothSceneConcatJob } from "./smooth-scene-concat.js";
@@ -109,6 +110,7 @@ export function registerAllJobHandlers(engine, books = bookManager) {
109
110
  registerExtractAudioJob(engine);
110
111
  registerExtractWaveformJob(engine);
111
112
  registerExtractVideoPosterJob(engine);
113
+ registerAssessMediaProbeJobs(engine);
112
114
  registerAnalyzeMusicJob(engine);
113
115
  registerPlanGoalJob(engine);
114
116
  registerAssembleGoalJob(engine);