feat: capture requests - in-app screenshots, screen recordings and logs

[boehlerlukas]

What this adds

iOS support for capture requests. When a workflow, an AI agent or a teammate asks for a screenshot or a screen recording, the customer answers from the widget card and the SDK captures the app. Logs requests are collected in the background.

Capture, in the app only, with no prompt

• No ReplayKit and no new permission prompt. The SDK draws the app's own windows (drawViewHierarchyInRect).
• Screenshots:
  • every visible window of the foreground scene, in window order;
  • Gleap's own windows and the keyboard are left out;
  • JPEG, long edge ≤ 2560.
• Recordings: the same window pipeline at an adaptive ~4 fps, encoded with AVAssetWriter as H.264 MP4.
  • Long edge ≤ 1280, letterboxed on rotation, a keyframe every 2 s.
  • Background time is cut out: parts are joined without the gap.
  • Stops cleanly on a memory warning; hard cap 180 s.
• Masking: secure text fields, one-time-code and card-number fields, and views passed to maskView:, blacked out before anything leaves the device.
  • Masks follow views where they are on screen, also mid-animation. Each mask covers the area the view swept while the window was drawn, plus its final position and 3 pt padding.
  • If the masks can't be placed, the whole window is blacked out.
• Gleap's own launcher, banners, modals and notifications are left out of captures.
• Capture bar: a small dark pill that can be dragged anywhere.
  • Ready: grip, then Capture or Start as a pill in the widget colour, then a round ×.
  • Recording: only the red dot, mm:ss / max and a round red Stop.
  • Instruction: a caption above the bar for about 4 s.
  • After a drag: it springs back into the safe area and snaps to the centre. Its position is remembered for the app session and re-fitted on rotation and keyboard changes.
  • Touches outside the pill reach the app.
• Preview:
  • dark full screen, rounded video, tap to play or pause, thin scrubber;
  • × at the top right, which during an upload cancels it;
  • Retake and Send ➤ pills.
• Text: every word in the bar and preview comes from the widget's translated capture-start.labels; English is only a fallback.
• Widget during a capture: minimized, then restored on the conversation, without widgetOpened / widgetClosed callbacks for that.
• Robustness:
  • Widget open, minimize, restore and close run one after another, each with a watchdog, so a cancel or close mid-animation can't strand the widget.
  • The widget's JavaScript alerts always complete; this fixes a crash with an alert during a capture.
  • If WebKit's web content process dies, the widget page reloads and gets the pending capture back.
  • Leftover capture files are deleted at launch.
  • Every entry point of the capture flow catches exceptions.

Logs requests

• Taken from the SDK websocket (capture-request) or the ping answer (cr), with caps announced on both.
• Flow: claim → flush handler (≤ 500 ms) → build the bundle from the existing collectors → gzip → /logs. Once per request id.
• setRemoteLogCollectionEnabled:NO also stops logs attached to captures.
• A logs request that fails for network reasons is tried three times (now, +2 min, +8 min), then reported failed.

Public API (Gleap class)

typedef void (^GleapLogFlushHandler)(void (^done)(void));
+ (void)setCaptureEnabled:(BOOL)enabled;             // NO: the widget offers upload only
+ (void)setRemoteLogCollectionEnabled:(BOOL)enabled; // NO: logs requests answered "unsupported"
+ (void)maskView:(UIView *)view;                     // weak reference
+ (void)unmaskView:(UIView *)view;
+ (void)setLogFlushHandler:(nullable GleapLogFlushHandler)handler; // done from any thread, 500 ms cap

Also in this PR

• Inline video: videos play inline in the widget.
• Replay frames are kept as small JPEGs (about 13 KB instead of about 12 MB each). The upload format is unchanged.
• Flutter screenshots: Flutter views are drawn at their real frame.
• Build: Package.swift and the podspec link libz; the podspec adds AVFoundation, CoreMedia and CoreVideo.
• Privacy manifest: declares Customer Support data. Please confirm it should be marked linked to the user; the existing entries are "not linked".

Verification

• Unit tests: 124/124 on iPhone 17 Pro, including a new mask test that fails with the old model-only logic. pod lib lint passes with no warnings. The package and the example app build for iPhone and iPad simulators.
• iPhone simulator end to end (against a local stub):
  • Screenshot: 1178×2560 with masked views and the secure field black; no keyboard, bar or Gleap button in it.
  • Recording: H.264 High 588×1280, keyframe ~2 s, rotation letterboxed, 6 s auto-stop exactly 6.000 s, background gap removed.
  • Retake, Cancel (bar and preview), capture-cancel, Gleap.close() mid-recording, background/foreground and a simulated memory warning all clean up.
  • Logs over the websocket and the ping answer, including deduplication, expired and invalid ids, and the kill switch.
  • Memory stays flat over 5 cycles.
• iPad simulator: the bar in portrait and landscape, a 2560×1920 screenshot, a pillarboxed 1280×960 recording, dark mode, German labels.
• UI round, iPhone and iPad, portrait and landscape:
  • every bar and preview state, with screenshots;
  • drag to a corner, then the next capture starts there;
  • rotation with a remembered position;
  • keyboard up;
  • neither the bar nor the caption is ever in a capture.
• Stability round, iPhone and iPad simulators:
  • Masks during a stress animation: before the fix, 40/40 screenshots leaked. After it, 0/40 screenshots and 0 of 46–47 recorded frames leak, including a rotation.
  • Cancel at 0–600 ms and close during the minimize and restore animations leave the widget usable.
  • An alert() during a capture: no crash.
  • Killing WebContent mid-capture: the image arrives after the reload.
  • Logs retries end in failed.

Not covered

• No real devices.
• Split View / Stage Manager windows smaller than the screen.
• Black content: video players and DRM content come out black.
• Not auto-masked: password fields inside Flutter or web views can't be detected (maskView: covers native views). The customer reviews every capture before sending.
• Snapshot transitions: content inside UIKit's snapshot-based transitions (e.g. a cross-dissolve) has no live view for the masks to follow.
• [CATransaction flush]: the mask fix calls it up to twice per window per frame. It commits pending UI changes slightly earlier in the run loop; no visible effect was seen.
• Slow devices: a WebContent reload took about 10 s on the simulator, close to the existing 15 s widget timeout.

Release

Version bump and CHANGELOG at release. The wrappers (React Native, Flutter, Capacitor) need this release; their draft PRs bump the pin.

---

The whole program: capture requests ("Show me the issue")

Workflows, AI agents (Kai, Kai Resolve, custom agents) and teammates can now ask a customer for a screenshot (with annotation) or a screen recording in the middle of a conversation, and can collect logs (console, network, custom data, environment data, events) from the customer's app in the background.

One server object, the capture request, backs all three. The customer sees a card in the widget; the SDK in their app does the capture; the result lands in the thread and resumes the workflow or wakes the agent. Background log requests travel over the SDK websocket every SDK already keeps open.

How each platform captures:

• Web, desktop Chrome/Edge: screenshot via one-click "Share this tab" (exact pixels), in-browser render as fallback; recording via tab capture, "Record this page" fallback when sharing is blocked.
• Web, Firefox/Safari/phones/tablets: in-browser screenshot; window/screen recording on desktop, page recording (rrweb) on phones.
• iOS / Android apps (incl. React Native, Flutter, Capacitor): in-app capture of the app's own windows: no OS prompt and no new permissions (no ReplayKit, no MediaProjection). Recordings are H.264 MP4 built from frames.
• Older SDKs, email, WhatsApp & co.: upload / reply-with-attachment fallback, so a request never dead-ends.

Design: https://claude.ai/artifact/5i88BeWfcsWUFhhJmPwK5V (private)

PRs in this program (one per repo):

• Server · Messenger-App · Frontend (dashboard)
• JavaScript-SDK · Gleap-iOS-SDK · Android-SDK
• ReactNative-SDK · Flutter-SDK · Capacitor-SDK (merge after the native release)
• gleap-api-docs (merge after the SDK releases)

🤖 Generated with Claude Code