Reconstruct what happened in your app

Connect what the user saw with captured logs, telemetry, and app state to investigate a failure and preserve the moments that matter.

The local host player is built around reconstruction: show what happened, in order, with enough evidence to explain a failure without relying on memory or guesswork.

Select a session

The session explorer lists connected and historical captures. Select one to load its metadata and evidence. Live sessions refresh as their counts and latest timestamp change; recorded sessions remain stable until you annotate, trim, import, or otherwise manage them.

Timeline review

The player aligns:

  • screenshots and visual tree snapshots;
  • logs, exceptions, lifecycle, and network events;
  • touch input and inferred gestures;
  • trends and custom telemetry;
  • annotations and artifact snapshots.

Scrub to a moment to compare what the user saw with what the app reported. Scrubbing away from the newest point pauses live following; choose Go live to resume at the latest evidence.

For SDK sessions with Apple Pencil or Android stylus input, pen contact uses a distinct marker and eraser input appears pink. Marker size reflects available pressure or normalized Pencil force. Tilt changes its shape; azimuth or orientation rotates it with a direction dot. A dashed outline marks hover. Hover over a marker for the exact reported fields. Ordinary finger touches and older recordings without pen details retain their usual display. Estimated Pencil corrections remain in session data without drawing another marker.

For an eligible connected simulator or emulator, switch between Simulator for the current controllable target and Playback for the captured timeline.

Select, extract, and trim a range

Enable range selection and drag across the interval that matters. Depending on session state, the range menu can offer:

  • Add annotation to mark a problem or area of interest without changing evidence;
  • Extract new session to preserve a focused copy while keeping the source;
  • Remove selected range to permanently delete evidence inside the range;
  • Keep only selected range to permanently delete evidence outside it.

Trim operations cannot be undone and are unavailable for a connected live session. Prefer extraction for review and handoff.

Use annotations to guide an agent’s investigation or direct developers and testers to a moment or screen area that needs review. Include what happened, what you expected, and what to check next.

Trees, files, and artifacts

Visual trees appear as searchable hierarchies with node properties. Selecting a bounded node can overlay it on the corresponding screenshot. Artifacts and approved live files open through the host’s bounded visualization service.

Tests shows persisted test batches and model/tool execution audits. When a session has a linked test run, Trace opens that exact run so its decisions, tool calls, evidence, and outcome can be reviewed beside the capture.

For runs captured with --trace, select a TypeScript task in the execution trace to view its API call results table. Each row shows the call’s order, start time, duration, and outcome. Select a row to inspect its input arguments and output or error, including calls made by composed tasks. Older traces show the available metadata and identify payloads that were not captured. Large payloads are bounded and marked when truncated.

Trends groups monitored metrics by registered app and shows their current value, reference, delta, decision, and history.

Agents reconstruct sessions with ansight session show, logs, metrics, images, touches, trees, artifacts, and annotations, using --json for structured evidence. Use session extract for a focused copy and session trim only for an intentional destructive edit. See CLI Commands.