API keys, rate limits, and building a real pipeline
The REST API is the same one the web UI uses. The difference is that a script can be polite about it, and the platform will treat it accordingly.
Keys
- Create and revoke keys under Settings → API keys.
- Scope each key to the task: uploads, reads, or full access.
- Rotate without downtime by running two keys during a deploy.
curl -H "Authorization: Bearer $KEY" \
-F "[email protected]" \
https://zerostorage.net/api/files/uploadRate limits
Limits are per key and per IP, with response headers showing the remaining budget. Back off on 429 using those headers instead of a fixed sleep — the budget refills continuously.
Retry pattern
- Retry 429 and 5xx with exponential backoff and jitter.
- Never retry a 401 or 403 — rotate the key instead.
- Idempotency comes from hashes: re-sending the same file is free.
A pipeline that survives contact
Bulk imports go wrong in predictable ways: the network blips, a key rotates mid-run, and a naive loop hammers a server that is still recovering. The pattern that holds is boring — queue the work, run a handful of workers, and let the response headers pace them. Persist the hash of each finished item so a restart resumes instead of repeating. When a run ends with errors, the useful artefact is a list of hashes and status codes, not a log file you have to read like a novel.
- Cap concurrency first; raise it only while the headers show budget left.
- Checkpoint progress every batch, not every file.
- Alert on the ratio of 429s to successes — that ratio is the real health signal.
None of this is exotic. It is the difference between a script that runs once on your laptop and a pipeline someone else can operate without asking you how it works. Write the runbook while the failure modes are still fresh — which key scopes exist, what the steady-state concurrency is, and what a healthy day of responses looks like — because the person paged at 3am will not read the source first.
Browse the API reference and the playground before wiring anything in production.
Continue reading
Why HLS still wins for video that has to just play
Adaptive bitrate is not a buzzword — it is the reason a viewer on hotel Wi-Fi and a viewer on fiber both reach the end of your video. Here is how HLS does it on ZeroStorage.
Upload without an account: how anonymous sessions work
No sign-up, no quota anxiety, no lost files. Anonymous sessions tie uploads to a browser instead of a user — this is what that means for you and for us.
SFTP for bulk uploads: a practical field guide
Dragging ten thousand files into a browser is a bad time. SFTP is the boring, reliable path — keys, folders, partial transfers, and what breaks first.