Mobile sessions now report the operating system name and version. Previously the OS information was missing, so Android sessions never had an OS recorded. The SDK now sends the Android release alongside the API level (Android).
Survey widgets and the screenshot editor now respect the status bar, display cutout, and gesture bar. The SDK passes the system bar insets to the survey, so content no longer sits under the clock and the editor toolbar clears the gesture bar (Android).
Added Screeb.setAnonymousId() to adopt the anonymous id your CDP already uses (e.g. Amplitude device id, Segment/RudderStack anonymous id). Call it right after initSdk; the respondent stays anonymous, or is switched to an existing one tied to that id if it isn't already identified (Android).
Added Screeb.setAnonymousId(anonymousId:) to link the current respondent to the anonymous id your CDP already uses (e.g. Amplitude device id, Segment or RudderStack anonymous id). The respondent stays anonymous, and if that id already belongs to an existing respondent, Screeb switches to it (unless the current one is already identified). Call it right after initSdk (iOS).
Fixed a race where picking a file attachment could occasionally fail or drop the picked file, caused by cache cleanup running concurrently with the copy of a newly picked file (Android).
Session replay no longer gets stuck recording a stale screen after a burst of capture failures (e.g. under heavy GPU load); it now backs off and retries instead of disabling itself for the rest of the session, and recovers any changes made while capture was paused (Android).
SDK error reports are no longer sent over the network before the SDK is initialized (when consent may not yet be collected), and no longer include command arguments, message contents, or opened links that could carry personal data (Android).
SDK error reports are no longer sent over the network before initSdk runs, since the host app may not have collected consent yet, they now stay local until then. Error messages also no longer include command arguments or other data that could carry visitor identities or properties (iOS).
Session replay adapts more smoothly under sustained memory pressure: pressure now eases gradually rather than staying elevated (and capture quality/cadence reduced) until the next app resume (Android).
Slightly smoother session replay capture rate under normal conditions (Android).
Screenshot answers no longer capture the iOS screen-sharing "Stop Sharing" overlay — the SDK now waits for that system sheet to be dismissed before grabbing the frame (iOS).
Session replay recovers on its own if an update is ever lost in transit: a changing screen now gets a fresh keyframe at least every 30 seconds instead of staying wrong for the rest of the session (iOS).
Session replay no longer misses the last change made right as a screen settles (iOS).
A brief memory warning no longer keeps session replay (quality, resolution, capture rate) throttled for the rest of the time the app is in the foreground — the throttling now eases back automatically a few seconds after the warning stops recurring (iOS).
Session replay capture does less work for hidden looping animations, duplicate change regions, and paused capture passes, lowering CPU and battery impact with no change in replay fidelity (iOS).
Session replay recovers automatically from frames lost due to internal capture limits or delivery hiccups, instead of leaving the replay stuck until an unrelated event forced a resync (Android).
Reduced main-thread work during session replay capture on static screens, accessibility scans, and hidden views, lowering CPU overhead for apps with session replay enabled (Android).
Exit-intent surveys now work in native apps: leaving a tracked screen can trigger a survey, as on web. Requires screen tracking (TrackScreen) instrumentation.
New "inactivity" targeting rule: trigger a survey after the visitor has stopped interacting with the app for a configured time.
Time-based triggers (delay, session duration, inactivity) now fire on time: the native SDKs arm their own timer instead of relying on webview timers, which the OS throttles while the webview is hidden.
A device without a usable network connection (airplane mode, captive portal, blocked DNS) no longer reports device-side failures as errors; genuine problems are still reported and now identify the workspace they belong to.
Respondents can attach a picture to an answer again on iOS: the attach button opened nothing, and the photo library could end up behind the survey (iOS).
An audio answer no longer asks for the camera, and refusing one permission no longer takes the whole answer down (Android).
Recording works on the first attempt instead of failing until the respondent tried again (Android).