golden_lens 0.2.0

Agent-grade golden tests for Flutter: attribute visual diffs to the exact widget and source line, score perceptual parity, and emit agent-readable reports.

Changelog

0.2.0

• Z-order-aware attribution: overlapping same-size widgets now attribute to the topmost (latest-painted) one.
• HTML report adds numbered offender overlays on the "New" image, plus a highlighted Diff panel (changed regions tinted + offenders boxed) via the shared renderDiffPng.
• Report JSON now includes candidate.devicePixelRatio.
• README + pub.dev screenshots generated from a real pipeline run.

0.1.0

Initial release.

• Attribution engine — maps a changed-pixel region to the innermost owning widget and resolves framework-internal render objects up to the user's own file:line:column (via Flutter's public widget-creation tracking).
• Image pipeline — anti-alias-tolerant pixel diff, connected-component region clustering, and a summed-area-table SSIM perceptual parity score (global + per region). Zero dependencies beyond the Flutter SDK.
• Drop-in comparator — installGoldenLens() wraps the active golden comparator so every matchesGoldenFile and alchemist golden gains attribution on failure, with no test changes. Never alters pass/fail.
• Reports — a machine-readable agent JSON payload (schema v1.0) and a self-contained human HTML report.