videofetch (Python SDK)

Official VideoFetch SDK — video ingestion API: give us a video URL, we deliver MP4/MP3 to your storage.

pip install videofetch-sdk

Usage

import videofetch

client = videofetch.VideoFetch(api_key="vf_live_sk_...")

# L1: create a job (returns immediately, status "queued")
job = client.downloads.create(
    url="https://www.youtube.com/watch?v=...",
    format="1080p",                 # 144p..2160p | mp3
    trim=videofetch.TrimSpec(start=120, end=420),   # optional clip window
    # destination={"type": "r2", "id": "..."},      # optional direct-to-bucket
)

# L2: wait for the terminal state (polls with backoff)
result = job.wait(timeout=120)
print(result.status)        # "completed"
print(result.download_url)  # presigned link, valid 7 days (url destination)
print(result.storage_key)   # "user://<bucket>/<key>" (bucket destination)
print(result.size_bytes, result.cost_usd)

# L3: one-shot
result = client.downloads.create_and_wait(url=..., format="1080p")

Async (FastAPI / Next.js backends):

from videofetch.asyncio import AsyncVideoFetch

async with AsyncVideoFetch(api_key="...") as client:
    job = await client.downloads.create(url=..., format="1080p")
    result = await job.wait(timeout=120)

Free metadata lookup:

info = client.info.lookup("https://www.youtube.com/watch?v=...")
print(info.title, info.duration, info.formats)   # available quality tiers

Webhooks (FastAPI):

from fastapi import Request, HTTPException
from videofetch.webhooks import construct_event, SignatureVerificationError

@app.post("/webhooks/videofetch")
async def on_event(request: Request):
    body = await request.body()
    try:
        event = construct_event(body, request.headers.get("X-VideoFetch-Signature"), WEBHOOK_SECRET)
    except SignatureVerificationError:
        raise HTTPException(status_code=400, detail="bad signature")
    if event["event"] == "download.completed":
        ...  # fetch the result via client.downloads.retrieve(event["id"])
    return {"ok": True}

Configuration

• api_key — required. Create keys in the Dashboard (vf_live_sk_...). Falls back to the VIDEOFETCH_API_KEY env var.
• base_url — defaults to the hosted API; override for local/single-server dev (http://localhost:8301). Env: VIDEOFETCH_BASE_URL.
• timeout (s), max_retries (default 2, automatic on 429/5xx).

Errors

All SDK errors derive from videofetch.VideoFetchError:

Serverless warning

Do not call job.wait() in a Vercel/Cloudflare/Lambda function — it burns billed execution time. Create the job, persist its id, then respond to the download.completed webhook and retrieve(id) the final result.

Development

pip install -e ".[dev]"
pytest

License

MIT

Usage, alerts and credits

u = client.usage.get()
u.plan            # "free" | "developer" | "growth" | "scale"
u.quota_gb        # account plan quota for this month (GB)
u.used_pct        # % of the monthly quota consumed
u.remaining_gb    # remaining GB before overage
u.active_jobs     # jobs currently queued + processing on your account
u.concurrency_limit  # max in-flight jobs per account
u.alert_level     # ok | warning | critical | exceeded
u.payg_balance_cents  # pay-as-you-go credit balance

a = client.usage.alerts()
a.state.level          # ok | warning | critical | exceeded
a.state.crossed        # e.g. [80, 95] — thresholds already crossed
a.state.next_threshold_pct
a.state.action         # "topup" | "upgrade" | None  → drive your own UI
a.fired                # alerts already delivered this month (kind/level/pct/created_at)
a.thresholds           # [80, 95, 100]
a.topup_amounts        # [10, 25, 50, 100]

Alerts are evaluated after every successful download and delivered to your webhook endpoints as quota.warning (80%/95%) and quota.exceeded (100%); a low credit balance emits balance.low. Each threshold fires at most once per calendar month, so your handler will not be spammed.

Managing webhook endpoints with your API key

wh = client.webhooks.create(
    "https://api.yourapp.com/videofetch/hook",
    events=["download.completed", "quota.exceeded", "balance.low"],
)
wh.secret            # whsec_... — FULL value, shown only on create. Store it now.

client.webhooks.list()          # secrets are masked here
client.webhooks.test(wh.id)     # send a webhook.test ping, check delivery
client.webhooks.delete(wh.id)

Available events: download.queued, download.processing, download.completed, download.failed, quota.warning, quota.exceeded, balance.low.

Verification mirrors the server exactly — HMAC-SHA256 over the raw request body:

from videofetch import construct_event   # or verify_webhook_signature(raw, header, secret)
event = construct_event(request.body, request.headers["X-VideoFetch-Signature"], wh.secret)

Concurrency limits and 429s

Your account may run a limited number of jobs at once (queued + processing). When the cap is reached the API answers 429 with a typed RateLimitError instead of silently queueing forever:

from videofetch import RateLimitError
try:
    client.downloads.create(url=url, format="1080p")
except RateLimitError as e:
    e.code        # concurrency_limit_exceeded | queue_limit_exceeded | platform_at_capacity
    e.limit, e.active, e.scope
    e.retry_after # seconds, from the Retry-After header

GET /v1/usage returns active_jobs / concurrency_limit so you can schedule work before hitting the cap. A suspended account returns 403 with code=account_suspended (surface it to your own operators rather than retrying).