e2e-walkthrough

Recording interactive browser walkthroughs with step-level screenshots uses a guided agent-browser session that executes an observe-and-continue loop, capturing per-step screenshots and trace data for visual QA. You simply select a mapping file to define the exploration scope, and the pipeline handles the capture automatically.

Human-in-the-loop UI exploration is a structured walkthrough process where a user guides browser interactions step-by-step to validate application features. You need it for PR-driven UI reviews when automated tests fail, requiring human validation to capture anomalies and generate reproducible trace artifacts.

You can auto-generate flow YAML and dual markdown reports after web application feature exploration by stopping the agent-browser trace and running analysis. The pipeline outputs report.md and pr-summary.md alongside the flow YAML, consolidating step logs and anomaly correlations into actionable review artifacts.

Agent-browser trace capture works with multi-site mapping discovery by enforcing mapping selection to reliably locate pages and elements across different sites before exploration. This ensures cross-site scenarios maintain consistent trace analysis and step-log.json generation throughout the walkthrough.

The best way to handle authentication when recording a guided UI walkthrough is using the pipeline's built-in pre-flight checks, which support both auto and manual auth handling. This ensures the agent-browser session securely establishes access before starting the step-level screenshot capture.

A visual QA walkthrough stops when an anomaly is detected because the observe-and-continue loop logs anomalies for trace analysis and triggers retry or debug flows. This mechanism ensures errors are captured with step-level screenshots before proceeding, maintaining reproducible artifact generation.