Zillusion · warehouse & monitor
A dataset is a liability until something keeps it fresh.
Scrapers rot: site redesigns silently break selectors, sources move, schemas drift. Zillusion treats the first successful crawl as the beginning — every dataset is versioned, diffed, and watched.
Versioned datasets, not loose files
When a run completes, its output is saved into your warehouse as a versioned dataset with a manifest. Subsequent runs produce incremental diffs against the previous version, so you can see exactly what changed between crawls instead of re-downloading and re-comparing by hand. Datasets are private to your account, previewable in the workbench, and deletable by you at any time.
Schedules, drift, and notifications
Any workflow that passed the verification gate can be promoted to a robot and put on a schedule. Scheduled runs re-execute the robot, version the new output, and diff it against the last version. When something drifts — or a run fails outright — you’re notified by webhook or email, so a broken source surfaces as an alert, not as a quietly stale dashboard three weeks later.
In the recorded demo run on our homepage, this is the closing beat: “dataset v1 + manifest saved · monitor armed → diff & notify.” That sequence — version, arm, diff, notify — is the default lifecycle, not an add-on.
Whose data is it?
Yours. You retain ownership of your content, subject to third-party rights in the underlying sources. Datasets are account-scoped; deleting a dataset — or your whole account, which is self-service — removes it. There is no automatic expiry of your service content while your account is open, and no third-party analytics or advertising trackers anywhere on the platform.
Frequently asked questions
How does Zillusion keep a dataset up to date?
Verified workflows are saved as robots and re-run on schedules. Each run versions the new output into the warehouse and computes an incremental diff against the previous version; drift and failures trigger notifications by webhook or email.
Can I see what changed between two crawls?
Yes. Datasets are versioned with manifests, and scheduled runs produce incremental diffs against the previous version, so changes are visible run over run instead of requiring manual comparison.
What formats does the output come in?
Runs produce structured tabular output — the recorded demo run, for example, produced 63 records across 3 CSV files. Datasets are also previewable directly in the workbench.
Who can see my datasets?
Datasets are private to your account. You can delete individual datasets, and account deletion is self-service with asynchronous cleanup of your stored content.