Preserve legend selections across metric changes

[cquil11]

Summary

• Preserve the user's active legend subset when switching the Y-axis metric or X-axis mode.
• Refresh the automatic Best per SKU winners when the selected metric changes.
• Keep official and ?unofficialrun= overlay behavior aligned.
• Continue resetting legend selections when model, scenario, precision, or official/unofficial-run scope changes.

Root cause

The legend reset key included axis controls, so every metric change reset manual selections. Removing those controls exposed a separate automatic-mode requirement: Best per SKU is metric-aware and must still refresh when the axes change. The implementation now treats manual and automatic selections independently.

Verification

• bun run fmt
• bun run lint
• bun run typecheck
• bun run test:unit (3,922 tests)
• Focused Cypress coverage in Chrome (38 tests), including unofficial-run overlays

---

Note

Medium Risk
Changes inference legend state logic across context and overlay mode; behavior is nuanced but covered by expanded Cypress and unit tests rather than security-critical paths.

Overview
Legend hardware selection no longer resets when you change the Y-axis metric or X-axis mode. The reset key in InferenceContext now only tracks model, sequence, precision, and official vs unofficial scope—not axis controls—so manual hide/show subsets survive metric switches.

When Best per SKU stays enabled, winners still recompute for the new axes (metric-aware Pareto), including official series and unofficial overlay series in ScatterGraph. If no series is scoreable, selection falls back to the full in-scope set.

Overlay legend edits (toggle or remove while ?unofficialrun= is loaded) turn off Best per SKU via setBestPerSku(false, { applySelection: false }) so the restored official activeHwTypes when the overlay is dismissed are not overwritten.

Cypress and unit tests cover metric-change persistence, Best per SKU refresh, and overlay behavior.

^{Reviewed by Cursor Bugbot for commit 4868850. Bugbot is set up for automated code reviews on this repo. Configure here.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
inferencemax-app	Ready	Preview	Aug 16, 2026 8:15pm

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 1600fdd. Configure here.}