Choose which sensitive image types the AI Agent reads
You can now choose which types of sensitive image your AI Agent withholds. Two toggles under Image redactions cover government ID documents and other credential documents. Both are enabled by default, which matches the previous behavior. Disable a type when your end users must send that document to complete a flow, such as a bank-transfer confirmation.
Payment cards are always withheld, and card numbers read from any image are always redacted.
For more information, see Redaction settings.
Updating a Knowledge API source needs the write scope
To update a knowledge source with PATCH /v2/knowledge/sources/{id}, your API key now needs the knowledge_sources:write scope. This is the scope that already creates and deletes sources. Until now, a key with only knowledge_sources:read could update a source too. If your integration updates sources with a read-only key, it now gets a 401 response. Add the write scope to the key, or to the scopes your App registers.
The same request is refused, and nothing is saved, when it:
- gives the source a name that another source already uses
- renames the Created in Ada source, or gives its name to another source
- sets a sync frequency on a source that is not a website import
- targets a source that is being deleted
Before this change, the sync frequency and the availability rules in a refused request were saved before the name was rejected. Creating a source named Created in Ada is refused on every API.
For more information, see Sources in the API reference.
Mobile WebView recovery event
The mobile Messaging SDKs report ada.webview.contentProcessTerminated when the WebView content process ends and automatic recovery stops.
On this event, create a new view to restore chat on iOS, Android, and React Native.
On iOS, use a new AdaWebHost.
See the iOS, Android, and React Native event references.