1.0.0-beta.1 (2026-09-18)
Features Added
- Initial beta release of the modular
@azure/ai-content-safetypackage, generated from API version2026-09-01-preview. (#39966) - Added
ContentSafetyClientoperations for Unified Moderate, prompt shielding, and protected-material detection. - Added
ContentProvenanceClientwith long-running content provenance detection and poller restoration. - Added regex matching support for blocklist items.
Breaking Changes
- Migrating from
@azure-rest/ai-content-safetyrequires installing@azure/ai-content-safetyand updating package imports. - Replaced the default REST factory with named
ContentSafetyClient,ContentProvenanceClient, andBlocklistClientclasses. - Added dedicated request models for blocklist item and blocklist update payloads so service-generated identifiers are not required as input.
- Changed
ImageData.contentfrom a base64-encoded string to rawUint8Arraybytes.
Bugs Fixed
- Honor
byPage({ maxPageSize })for initial, continuation, and resumed blocklist-item pages. The operation-levelmaxpagesizeoption remains supported but is deprecated. - Preserve the custom response type in
restorePollerdeserialization callbacks.