FE-1685: Draw every sweep study's objective by step under the Parameters sliders

[kube]

Important

Experimental
Behind the Parameter sweeps and In-browser optimization feature flags.

Summary

Before this PR, a study started from a sweep's Parameters card showed its progress as one line under the sliders: the step count while it ran, the outcome once it settled. The objective's path over the steps, the chart the Optimizations tab draws for every study, was nowhere in the sweep drawer.

This PR draws it at the bottom of the Parameters card. From the first Optimize on, a full-width purple strip sits under the sliders: a disclosure row naming the metric, the steps run and the best value so far, and beneath it a compact chart of the objective per step with the best-so-far line, the x axis pinned to the steps to come so dots fill left to right while the study streams. The strip stays after the study settles and folds from its row in both states. A second Optimize appends its steps after a dashed divider, the numbering continuing. Before any study the card is unchanged.

9660.mp4

Links

• Ticket FE-1685
• Docs The Optimizations tab › Studies started from a sweep
• Docs Experiments › Optimizing a sweep

Changes

Shared chart

• ObjectiveHistoryChart moves to SimulateView/shared

  Takes points, an ObjectiveHistoryStyle, an optional pinned xMax and dividers drawn dashed. The default style keeps the study drawer's look; ObjectiveHistoryCard renders it unchanged.

• ObjectiveHistoryCard moves to study-results/objective-history-card.tsx

  The card around the shared chart in the study results, its look unchanged.

• objective-history-data.ts is shared too

  One buildObjectiveHistory for the study card and the strip; the infeasible colour lives in shared/infeasible-color.ts.

• Fold exported through drawer-frame.tsx

  The strip folds behind the frame kit's Fold; FoldProps stays private to the folder.

Sweep drawer

• useSweepOptimizer exposes every study of the sweep

  studies in creation order; study is the last of them.

• ResultsBand.below slot

  Rendered after a band's content; the Parameters band fills it once a study exists, null before.

• SweepObjectiveStrip

  A full-row disclosure button with a rotating chevron, a dot that breathes while a study drives, the title Objective by step and a tabular summary.
  The summary reads like Infected · 47 steps in 2 optimizations · best 650.500.
  Under it a Fold holds the 120 px chart in the purple ramp.
  Canvas colours ignore the dark theme, as the study chart's do.

• Studies concatenate

  buildSweepObjectiveHistory offsets each study's steps by the ones before, restarts the best-so-far line per study, draws a divider at each later study's first step, and pins the axis to the last study's requested steps only while it is active.
  Studies of different metrics share one y axis; the row summarises the last study that drew a point.

• Zero layout shift within a state

  Collapsed and open heights are constants; the chart instance stays alive while folded, so reopening shows the current picture.
  The open state resets when the drawer reopens.

Review fixes

• Best-so-far line breaks at each divider

  uPlot drew the stepped line as one series, so the earlier study's best held across the divider into the next.
  toObjectiveHistoryData takes the dividers and lays a null gap sample half a step before each, which the dots skip and the axis never labels.

• Sweep whose only study failed at start gets the empty summary

  buildSweepObjectiveHistory fell back to the last study for the metric name and best, so the row read Metric · 0 steps against its own contract.
  Summary now comes from the last counted study only and the row omits the metric segment when there is none, reading 0 steps.

• Strip reads "No steps run" when its every study failed at start

  The shared chart said "Waiting for the first step" for any empty points, so a sweep whose only study failed before a step showed a wait that never ends under a row reading 0 steps.
  ObjectiveHistoryChart takes an emptyLabel, the strip sets it from history.studyCount, and view-experiment-drawer.test.tsx asserts both labels.

Next steps

• Push a changed xMax through setScale if recreating the plot on settle ever flickers

Test coverage

• sweep-objective-history.test.ts:

  Global step numbering across studies, dividers, best-so-far restarting per study, the pinned axis only while active, an empty running study keeping its divider, a study that failed at start left out of the count and the summary, a sole failed study giving the empty summary.

• objective-history-data.test.ts:

  Aligned data with a null gap sample before each divider, no gap without dividers or for a divider no step has reached.

• experiment-results.test.tsx, view-experiment-drawer.test.tsx:

  The band's below slot null before a study and the strip after; the disclosure's aria-expanded and aria-controls; the row summary.
  The row reading 0 steps with no metric when the sweep's only study failed at start.

• sweep-optimizer coverage through view-experiment-drawer.test.tsx:

  Two sweep studies passed newest-first sort into creation order.
  The row reads 7 steps in 2 optimizations with the later study's best, the divider and the axis edge follow it, and Stop cancels its id.

• drawer-frame.test.tsx:

  Gains a jsdom ResizeObserver stub for the computing list's popover; nothing in the suite depends on a measurement.

• Storybook:

  Simulate / ViewExperimentDrawer gains Sweep, optimized twice and Sweep, optimization stopped; the optimizable and optimizing stories show the strip.

How to test

• Open Petrinaut preview on Vercel
• Viewport controls > Settings > Simulation > Parameter sweeps, In-browser optimization
• Menu > Load example > SIR Model
• Simulate > Experiments > Create
• Max runs per selection 100, Scenario Seasonal Flu, Sweep on population and infected_ratio
• Add metric > Place tokens, Place Infected
• Create sweep

  Expect no strip under the sliders

• Parameters > Optimize > Minimize, Steps 12 > Start

  Expect a purple Objective by step row under the sliders with a breathing dot, a chart filling left to right as steps land, the axis spanning 12 steps

• Click the row

  Expect the chart to fold away, the row to stay, the cards below to move as one block

• Click the row again, wait for the last step

  Expect the chart back, the dot still, the row reading Infected · 12 steps · best …

• Parameters > Optimize > Steps 6 > Start

  Expect a dashed divider after step 12 and the new steps numbered 13 onward, the row reading 18 steps in 2 optimizations

[vercel]

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

Project	Deployment	Actions	Updated
hash	Ready	Preview	Sep 13, 2026 2:53pm UTC
petrinaut	Ready	Preview	Sep 13, 2026 2:53pm UTC
petrinaut-docs	Ready	Preview	Sep 13, 2026 2:53pm UTC

1 Skipped Deployment

Project	Deployment	Actions	Updated
hashdotdesign-tokens	Ignored	Preview	Sep 13, 2026 2:53pm UTC

[cursor]

PR Summary

Low Risk
Simulate UI and chart refactor only; sweep optimization behavior is unchanged aside from surfacing study history in the drawer.

Overview
Adds an Objective by step strip to the sweep experiment drawer: after the first Optimize, a foldable purple row under the Parameters sliders shows the same objective-over-steps chart as the Optimizations tab, streaming while the study runs and staying after it settles. Further optimizations append on one axis with dashed study dividers and a per-study best-so-far line.

Shared chart refactor: ObjectiveHistoryChart and objective history data move to SimulateView/shared, with optional xMax, dividers, styling, and gap samples so the best-so-far line does not carry across studies. The study drawer keeps ObjectiveHistoryCard as a thin wrapper.

Sweep wiring: useSweepOptimizer exposes all sweep-origin studies (creation order); ResultsBand.below renders SweepObjectiveStrip, backed by buildSweepObjectiveHistory. Docs and Storybook/tests cover multi-study ordering, empty/failed studies, and fold behavior.

^{Reviewed by Cursor Bugbot for commit 76d4a9e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[Copilot]

🟡 Changes recommended

The chart incorrectly connects studies, drops infeasible-point rendering, mishandles failed-only summaries, and lacks the required changeset.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

Adds a persistent objective-history chart beneath sweep parameter sliders, reusing the optimization history chart infrastructure.

Changes:

• Extracts reusable objective-history data and chart components.
• Adds concatenated sweep-study histories, dividers, summaries, and folding UI.
• Updates tests, stories, architecture docs, and user documentation.

File summaries

File	Description
libs/@local/petrinaut-arch-docs/content/ui/optimizations-tab.mdx	Documents shared chart and sweep strip architecture.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/shared/results-view.tsx	Renders content below result-band controls.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/shared/results-model.ts	Adds the result-band below slot.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/shared/objective-history-data.ts	Provides shared objective-history transformation.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/shared/objective-history-data.test.ts	Tests objective-history generation.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/shared/objective-history-chart.tsx	Generalizes chart styling, axis pinning, and dividers.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/shared/infeasible-color.ts	Shares the infeasible-step color.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/shared/drawer-frame.tsx	Exports the shared fold component.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/shared/drawer-frame.test.tsx	Adds a ResizeObserver test stub.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/optimizations/view-optimization-drawer.test.tsx	Updates the objective-card mock path.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/optimizations/study-results/steps-table.tsx	Uses the shared infeasible color.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/optimizations/study-results/objective-history-card.tsx	Wraps the shared chart for study results.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/optimizations/study-results.tsx	Uses the extracted card and initializes below.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/optimizations/optimization-full-view.test.tsx	Updates the objective-card mock path.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/experiments/view-experiment-drawer.test.tsx	Tests sweep-strip behavior and multiple studies.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/experiments/view-experiment-drawer.stories.tsx	Adds optimized and stopped sweep stories.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/experiments/sweep-optimizer.ts	Exposes all sweep studies in chronological order.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/experiments/sweep-objective-strip/sweep-objective-history.ts	Builds concatenated sweep histories and summaries.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/experiments/sweep-objective-strip/sweep-objective-history.test.ts	Tests history concatenation and axis behavior.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/experiments/sweep-objective-strip.tsx	Implements the folding objective strip.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/experiments/experiment-results.tsx	Places the strip beneath sweep sliders.
libs/@hashintel/petrinaut/src/ui/views/Editor/panels/SimulateView/experiments/experiment-results.test.tsx	Tests result-model strip integration.
libs/@hashintel/petrinaut/docs/optimization.md	References the sweep objective chart.
libs/@hashintel/petrinaut/docs/experiments.md	Documents the new sweep-strip workflow.

Review details

• Files reviewed: 21/24 changed files
• Comments generated: 3
• Review effort level: Balanced

---

💡 Configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[cursor]

Stale Bugbot comment from a previous run.

The merge-base changed after approval.