API Shield’s schema learning can now be invoked manually, letting you run a full‑zone analysis whenever you need it rather than on the default weekly cadence. Practitioners who are iterating on API contracts or validating new traffic patterns can see the learned schema immediately, which shortens feedback loops and helps keep protection rules in sync with actual usage.
How to launch an ad‑hoc schema learning run
Navigate to Web Assets > Operations in the Cloudflare dashboard. Open the More options menu on the operations table and choose Ad‑hoc learn schema. In the dialog that appears, click Run learning. The interface will display the current status and the timestamp of the last completed run for the zone.
Operational impact
The ad‑hoc run processes the entire zone and relies on traffic that has already been observed. It does not introduce new synthetic traffic; instead it evaluates the requests that have passed through the zone since the last learning cycle. All existing learning thresholds remain in effect, so the run will only adopt schema elements that meet those criteria. While the zone‑wide run is active, the Learned schema view updates per operation, allowing you to inspect individual results without waiting for the overall process to finish.
Security considerations
Since the learned schema feeds into API Shield’s request validation, a timely run can reduce false positives or missed anomalies that stem from outdated models. However, the quality of the output depends on the representativeness of the traffic captured before the run. Engineers should ensure that the traffic profile reflects the intended usage patterns, otherwise the resulting schema may be incomplete or overly permissive. The unchanged thresholds act as a safeguard against over‑learning from sparse data.
Related CloudNinjas coverage: security.
What This Means For Practitioners
- Incorporate an ad‑hoc learning step into API testing pipelines after deploying new endpoints or changing request formats.
- Monitor the status dialog and per‑operation results to verify that the schema aligns with expectations before promoting changes to production.
- Validate that the traffic used for learning is representative; consider generating realistic load if natural traffic is insufficient.
- Keep an eye on the learning thresholds and adjust them only if you have a clear justification, as they continue to govern what schema elements are accepted.