mirror of
https://github.com/shumaiOne/shumai.git
synced 2026-11-22 20:27:37 +00:00
* feat(webui): remove original-file display fallbacks, use proxies only WebUI now never uses the raw original file for display. All viewers (image, video, compare, pdf) render exclusively from transcoded proxies; when no proxy exists the viewer shows an unavailable placeholder instead of falling back to media.original. - getBestTranscode: drop the isRaw fallback candidates - video-viewer / compare-video-pane: remove the synthetic 'Original' resolution entry and the original.downloadUrl hasMedia gate - image-viewer / compare-image-pane: remove original.downloadUrl default - pdf-viewer: require pdfTranscode.url (no original fallback) - download menus (video-control-bar, breadcrumb-nav): use media.original.key for the 'Original' download option instead of the isRaw transcode marker entry Backend: - transcode-video workflow no longer pushes the isRaw:true marker entry into videoTranscodes (it pointed at the original file key) - remove media.original.downloadUrl from the DTO, asset service mapping, MediaInfo DB type, and transcode workflows/activities (it was only consumed by the removed display fallbacks; downloads already use key via the download-url endpoints) Tests/fixtures updated; docs updated to state the proxy-only display policy. * refactor: remove isRaw field via data migration, clean legacy media JSON Add a Prisma data migration that cleans legacy media JSONB on both assets and watermark_files: - drop the raw-original marker entries (isRaw: true) from videoTranscodes / imageTranscodes — they pointed at the original file key and were only used by the removed display fallbacks / download menus, which now use media.original.key directly - strip the isRaw field from remaining transcode entries - drop original.downloadUrl (removed from the AssetInfo DTO) The migration is generated via prisma migrate dev --create-only and applied automatically by migrate deploy in production. It is guarded by a test (packages/db/src/clean-media-migration.test.ts) that seeds legacy media JSON, runs the migration SQL, and asserts the transform on both tables. With the data cleaned, remove the isRaw field entirely from the codebase: - DTO schemas + ImageTranscode/VideoTranscode interfaces - PrismaJson MediaInfo transcode types - watermark activity isRaw skip (and its test), watermark proxy mapping, agent-embedding transcode selection, webui transcode filters, and all test fixtures AGENTS.md: clarify that the no-manual-migration rule applies to DDL; data migrations (pure SQL transforms, no schema change) are allowed inside a --create-only generated migration and must be tested.
195 lines
5.9 KiB
TypeScript
195 lines
5.9 KiB
TypeScript
import type { AssetInfo } from '@shumai/dtos'
|
|
|
|
/**
|
|
* Deterministic metadata for the committed fixture video
|
|
* (packages/webui/e2e/fixtures/sample.mp4).
|
|
*
|
|
* Generated with:
|
|
* ffmpeg -f lavfi -i testsrc=size=16x16:rate=30 -t 5 -pix_fmt yuv420p sample.mp4
|
|
*
|
|
* Verified via ffprobe: 16x16, 30/1 fps, duration 5.0s, 150 frames.
|
|
* These constants are shared by the harness and the Playwright specs so the
|
|
* frame-accurate assertions have known-good expected values.
|
|
*/
|
|
export const SAMPLE_FRAME_RATE = 30
|
|
export const SAMPLE_TOTAL_FRAMES = 150
|
|
export const SAMPLE_DURATION = 5
|
|
export const SAMPLE_WIDTH = 16
|
|
export const SAMPLE_HEIGHT = 16
|
|
|
|
/**
|
|
* The fixture video is served statically by the harness Vite server
|
|
* (publicDir points at packages/webui/e2e/fixtures), so it is reachable at
|
|
* the site root. No backend is involved.
|
|
*/
|
|
export const SAMPLE_VIDEO_URL = '/sample.mp4'
|
|
|
|
export const sampleVideoAsset: AssetInfo = {
|
|
id: 'e2e-sample-asset',
|
|
name: 'sample.mp4',
|
|
sizeByte: 6170,
|
|
fileCount: 1,
|
|
type: 'file',
|
|
status: 'ready',
|
|
proxyType: 'video',
|
|
createdAt: '2024-01-01T00:00:00.000Z',
|
|
updatedAt: '2024-01-01T00:00:00.000Z',
|
|
media: {
|
|
proxyType: 'video',
|
|
original: {
|
|
key: 'sample.mp4',
|
|
},
|
|
videoTranscodes: [
|
|
{
|
|
id: 'sample-proxy',
|
|
url: SAMPLE_VIDEO_URL,
|
|
key: 'sample.mp4',
|
|
width: SAMPLE_WIDTH,
|
|
height: SAMPLE_HEIGHT,
|
|
size: 6170,
|
|
},
|
|
],
|
|
metadata: {
|
|
duration: SAMPLE_DURATION,
|
|
originalWidth: SAMPLE_WIDTH,
|
|
originalHeight: SAMPLE_HEIGHT,
|
|
frameRate: SAMPLE_FRAME_RATE,
|
|
totalFrames: SAMPLE_TOTAL_FRAMES,
|
|
},
|
|
},
|
|
}
|
|
|
|
/**
|
|
* Metadata variant reproducing the real-world case where the container/format
|
|
* duration is LONGER than the actual video stream (audio padding / container
|
|
* overhead). This mirrors an observed asset whose ffprobe reported:
|
|
* - video stream: nb_frames=268, duration=8.9333s
|
|
* - format: duration=9.008s (~2 frames longer than the stream)
|
|
*
|
|
* Here the backing file is still the real 150-frame sample.mp4 (video stream =
|
|
* 5.0s, 150 real decodable frames), but the stored metadata reports a container
|
|
* duration of 5.075s — i.e. ~2.25 frames longer than the 150-frame stream.
|
|
*
|
|
* `metadata.totalFrames` stays at the accurate stream frame count (150, from
|
|
* ffprobe `nb_frames`), exactly as the backend stores it.
|
|
*/
|
|
export const CONTAINER_LONGER_DURATION = 5.075
|
|
|
|
/**
|
|
* The frame count the player's current heuristic derives from the inflated
|
|
* container duration: round(5.075 * 30) = 152, i.e. 2 frames MORE than the 150
|
|
* frames that actually exist in the stream. Frames 150 and 151 are phantom.
|
|
*/
|
|
export const CONTAINER_LONGER_COMPUTED_TOTAL_FRAMES = Math.round(
|
|
CONTAINER_LONGER_DURATION * SAMPLE_FRAME_RATE,
|
|
)
|
|
|
|
/** The last frame index that actually exists in the stream (0-based): 149. */
|
|
export const REAL_LAST_FRAME = SAMPLE_TOTAL_FRAMES - 1
|
|
|
|
export const containerLongerVideoAsset: AssetInfo = {
|
|
...sampleVideoAsset,
|
|
id: 'e2e-container-longer-asset',
|
|
media: {
|
|
...sampleVideoAsset.media,
|
|
original: {
|
|
key: 'sample.mp4',
|
|
},
|
|
metadata: {
|
|
duration: CONTAINER_LONGER_DURATION,
|
|
originalWidth: SAMPLE_WIDTH,
|
|
originalHeight: SAMPLE_HEIGHT,
|
|
frameRate: SAMPLE_FRAME_RATE,
|
|
totalFrames: SAMPLE_TOTAL_FRAMES,
|
|
},
|
|
},
|
|
}
|
|
|
|
/**
|
|
* Metadata variant for the LARGE-gap ("videoB") case: the audio track is much
|
|
* longer than the video track, so the container tail is long enough for the
|
|
* playhead to actually traverse it during playback.
|
|
*
|
|
* Backed by the real committed fixture `sample-long-audio.mp4` (verified via
|
|
* ffprobe):
|
|
* - video stream: nb_frames=150, duration=5.000s, 30fps
|
|
* - audio stream: duration=6.500s
|
|
* - format: duration=6.500s (1.5s / 45-frame tail — well above the
|
|
* ~100ms playback stall threshold)
|
|
*
|
|
* Because the tail exceeds the stall threshold, both the current code and the
|
|
* fix derive the frame count from the container duration
|
|
* (round(6.5 * 30) = 195). The playhead advances through the audio tail to the
|
|
* container end, so at natural end the readout reaches its maximum (194).
|
|
*/
|
|
export const LONG_AUDIO_VIDEO_URL = '/sample-long-audio.mp4'
|
|
export const LONG_AUDIO_DURATION = 6.5
|
|
|
|
/** Container-derived frame count for the long-audio fixture: round(6.5 * 30). */
|
|
export const LONG_AUDIO_COMPUTED_TOTAL_FRAMES = Math.round(LONG_AUDIO_DURATION * SAMPLE_FRAME_RATE)
|
|
|
|
export const longAudioVideoAsset: AssetInfo = {
|
|
...sampleVideoAsset,
|
|
id: 'e2e-long-audio-asset',
|
|
name: 'sample-long-audio.mp4',
|
|
media: {
|
|
...sampleVideoAsset.media,
|
|
original: {
|
|
key: 'sample-long-audio.mp4',
|
|
},
|
|
videoTranscodes: [
|
|
{
|
|
id: 'long-audio-proxy',
|
|
url: LONG_AUDIO_VIDEO_URL,
|
|
key: 'sample-long-audio.mp4',
|
|
width: SAMPLE_WIDTH,
|
|
height: SAMPLE_HEIGHT,
|
|
size: 6170,
|
|
},
|
|
],
|
|
metadata: {
|
|
duration: LONG_AUDIO_DURATION,
|
|
originalWidth: SAMPLE_WIDTH,
|
|
originalHeight: SAMPLE_HEIGHT,
|
|
frameRate: SAMPLE_FRAME_RATE,
|
|
totalFrames: SAMPLE_TOTAL_FRAMES,
|
|
},
|
|
},
|
|
}
|
|
|
|
export const SAMPLE_AUDIO_URL = '/sample-audio.mp4'
|
|
export const SAMPLE_AUDIO_DURATION = 3.0
|
|
export const SAMPLE_AUDIO_COMPUTED_TOTAL_FRAMES = Math.round(
|
|
SAMPLE_AUDIO_DURATION * SAMPLE_FRAME_RATE,
|
|
)
|
|
|
|
export const sampleAudioAsset: AssetInfo = {
|
|
...sampleVideoAsset,
|
|
id: 'e2e-sample-audio-asset',
|
|
name: 'sample-audio.mp4',
|
|
proxyType: 'audio',
|
|
media: {
|
|
proxyType: 'audio',
|
|
original: {
|
|
key: 'sample-audio.mp4',
|
|
},
|
|
videoTranscodes: [
|
|
{
|
|
id: 'sample-audio-proxy',
|
|
url: SAMPLE_AUDIO_URL,
|
|
key: 'sample-audio.mp4',
|
|
width: 0,
|
|
height: 0,
|
|
size: 0,
|
|
},
|
|
],
|
|
metadata: {
|
|
duration: SAMPLE_AUDIO_DURATION,
|
|
originalWidth: 0,
|
|
originalHeight: 0,
|
|
frameRate: SAMPLE_FRAME_RATE,
|
|
totalFrames: SAMPLE_AUDIO_COMPUTED_TOTAL_FRAMES,
|
|
},
|
|
},
|
|
}
|