sgl-project/sglang · warning · HTTPException

Video not found

Error message

Video not found

What it means

HTTP 404 from GET /v1/videos/{video_id} when the in-memory VIDEO_STORE contains no job with that id. The store only holds jobs created by this server process (or its configured backend), so unknown, expired, or restarted-away ids all miss.

Source

Thrown at python/sglang/multimodal_gen/runtime/entrypoints/openai/video_api.py:915

    return VideoResponse(**job)


@router.get("", response_model=VideoListResponse)
async def list_videos(
    after: Optional[str] = Query(None),
    limit: Optional[int] = Query(None, ge=1, le=100),
    order: Optional[str] = Query("desc"),
):
    jobs = await VIDEO_STORE.list_page(after=after, limit=limit, order=order)
    items = [VideoResponse(**job) for job in jobs]
    return VideoListResponse(data=items)


@router.get("/{video_id}", response_model=VideoResponse)
async def retrieve_video(video_id: str = Path(...)):
    job = await VIDEO_STORE.get(video_id)
    if not job:
        raise HTTPException(status_code=404, detail="Video not found")
    return VideoResponse(**job)


# TODO: support aborting a job.
@router.delete("/{video_id}", response_model=VideoResponse)
async def delete_video(video_id: str = Path(...)):
    job = await VIDEO_STORE.pop(video_id)
    if not job:
        raise HTTPException(status_code=404, detail="Video not found")
    # Mark as deleted in response semantics
    job["status"] = "deleted"
    return VideoResponse(**job)


def _select_video_variant_path(job: dict, variant: str | None) -> str | None:
    file_paths = job.get("file_paths")
    if file_paths:
        try:

View on GitHub (pinned to 0132848349)

Solutions

1. Verify the id matches exactly what the POST /v1/videos response returned
2. If the server restarted, resubmit the generation — in-memory state is not persisted
3. Check VIDEO_STORE retention/capacity settings and poll sooner or persist ids externally
4. Ensure sticky routing / same host when multiple server replicas serve the API

null

const job = await get(id).catch(e => { if (e.status === 404) return null; throw e; });
if (!job) { /* resubmit or surface 'expired' */ }

When it happens

Trigger: GET /v1/videos/{id} with a typo'd or truncated id, an id from a previous server process (store lost on restart), or a job evicted after the store's retention policy/capacity limit.

Common situations: Server restarted between creating the job and polling status; copy-pasting an id with whitespace or missing characters; polling long after completion when the entry was evicted; hitting a different server instance behind a load balancer than the one that created the job.

Understand the failure class

Background: 'Could not be found', 'does not exist', 'not found in database': the resource-not-found family when an ID, slug, key, or URI lookup comes back empty — this error's family across 20 libraries.

Related errors

• Video generation failed: {error_msg}
• Lost connection to server after {consecutive_errors} consecu
• Network error after {consecutive_errors} consecutive failure
• Video generation timed out after {max_wait_time} seconds
• Failed to generate video: {str(e)}

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/7f8b58dbc133ab6d. Report an issue: GitHub.