Hosted appGenerative UIHosted, in placesucceeded

Public demos of OpenUI by Thesys, an open standard for generative UI: prompt to interface, output tabs and benchmarks. Tested in place.

Tested in place byDeepQA Teamatwww.openui.com/demosonSep 20, 2026

Run #1model gemini-balanced (vertex)took 7m

All 9 scenarios passed, no issues after the audit.

Share on X
OpenUI demos in the browser during the run

By the numbers

9 of 9
scenarios passed
136
browser actions
32
screenshots
140
model calls
6.9
minutes
9
scenarios
9
passed
0
failed
0
blocked
0
issues

Walkthrough

Every scenario DeepQA drove in the browser, in plan order, with the 28 screenshots it captured along the way. A passing scenario is evidence too.

  1. S1
    Generate UI from built-in preset

    4 steps, 4 screenshots

    pass
    S1-1.png
    S1 · Generate UI from built-in preset
    S1-3.png
    S1 · Generate UI from built-in preset
    S1-9.png
    S1 · Generate UI from built-in preset
    S1-11.png
    S1 · Generate UI from built-in preset
    • Navigated to /demos and viewed the initial state with preset buttons.
    • Selected the 'Weather dashboard' preset, populating the input textbox and enabling the Generate button.
    • Clicked the Parsed JSON tab and verified the structured JSON output is displayed along with token [redacted] metrics.
    • Navigated to /demos where the UI generation form was displayed with presets.
    • Selected the 'Weather dashboard' preset button, which filled the prompt input and enabled the Generate button.
    • Clicked the Generate button, observing real-time streaming output in the code pane.
    • Generation completed successfully, rendering the preview UI and populating Raw Output (1,129 tokens) and Parsed JSON (4,412 tokens) with token [redacted] metrics (74% fewer tokens).
    • Switched between Raw Output and Parsed JSON tabs to verify both views rendered correctly.
  2. S2
    Generate UI from custom prompt and model selection

    5 steps, 4 screenshots

    pass
    S2-1.png
    S2 · Generate UI from custom prompt and model selection
    S2-4.png
    S2 · Generate UI from custom prompt and model selection
    S2-7.png
    S2 · Generate UI from custom prompt and model selection
    S2-10.png
    S2 · Generate UI from custom prompt and model selection
    • Navigated to /demos and viewed the initial UI generator interface.
    • Selected model openai/gpt-5.2 from the model dropdown.
    • Entered custom prompt in the textarea and verified the Generate button became enabled.
    • Clicked Generate and observed the model stream raw OpenUI code, token [redacted] statistics, and the rendered UI preview for the user profile card.
    • Successfully navigated to /demos, switched the model to openai/gpt-5.2, entered a custom UI prompt for a user profile card, and clicked Generate.
    • The application processed the custom prompt and rendered the generated UI in the preview pane alongside the raw OpenUI output and token [redacted]
  3. S3
    Empty prompt validation

    3 steps, 3 screenshots

    pass
    S3-1.png
    S3 · Empty prompt validation
    S3-4.png
    S3 · Empty prompt validation
    S3-9.png
    S3 · Empty prompt validation
    • Initial state on /demos shows the 'Describe the UI you want to generate…' textarea is empty and the 'Generate' button is disabled.
    • Entering only whitespace into the textarea keeps the 'Generate' button disabled.
    • Clearing the prompt disables the 'Generate' button, preventing empty prompt submission.
    • Navigated to /demos and verified that the 'Describe the UI you want to generate…' textarea is empty by default.
    • Confirmed the 'Generate' button is disabled when the prompt textarea is empty.
    • Entered whitespace (' ') into the textarea and verified the 'Generate' button remained disabled.
    • Entered text into the textarea to observe the 'Generate' button enable, then cleared the textarea back to empty and verified the 'Generate' button became disabled again.
    • Pressed Enter in the empty textarea and verified that no generation occurred and the button remained disabled.
  4. S4
    Inspect generated outputs via tabs and fullscreen

    5 steps, 2 screenshots

    pass
    S4-1.png
    S4 · Inspect generated outputs via tabs and fullscreen
    S4-14.png
    S4 · Inspect generated outputs via tabs and fullscreen
    • Initial view of /demos with preset buttons and empty output areas.
    • Generated UI for Pricing cards preset; Raw Output tab displays OpenUI code with token [redacted]
    • Navigated to /demos where the UI generation demo is hosted.
    • Selected the 'Pricing cards' preset and triggered UI generation with the 'Generate' button.
    • Observed the streaming generation complete successfully, displaying the generated OpenUI code under the Raw Output tab along with token [redacted] metrics (737 tokens vs 3,411 tokens, 78% fewer tokens).
    • Switched to the 'Parsed JSON' tab and observed structured JSON syntax representing the generated UI tree.
    • Both raw text and structured JSON tabs switch dynamically and render expected content without errors.
  5. S5
    Whitespace-only prompt validation

    3 steps, 2 screenshots

    pass
    S5-1.png
    S5 · Whitespace-only prompt validation
    S5-5.png
    S5 · Whitespace-only prompt validation
    • Entered multiple spaces into the prompt textarea; the Generate button remained disabled, preventing submission of whitespace-only input.
    • Navigated to /demos and verified the prompt input validation with whitespace characters.
    • Typing spaces into the 'Describe the UI you want to generate…' input keeps the 'Generate' button disabled.
    • No UI generation request is initiated for whitespace-only input.
  6. S6
    Documentation search functionality

    4 steps, 4 screenshots

    pass
    S6-2.png
    S6 · Documentation search functionality
    S6-4.png
    S6 · Documentation search functionality
    S6-7.png
    S6 · Documentation search functionality
    S6-10.png
    S6 · Documentation search functionality
    • Navigated to /docs/openui-lang documentation page.
    • Typed query 'agent' into search input; modal displayed rich list of clickable documentation search results instantaneously.
    • Selected search result navigated successfully to the target documentation page /docs/agent/reference/agentinterface-props.
    • The documentation search modal opened promptly when clicking the 'Search documentation' button on /docs/openui-lang.
    • Typing 'agent' into the search input instantly displayed relevant search results across various documentation categories.
    • Clicking/activating a search result cleanly navigated to the corresponding documentation page (/docs/agent/reference/agentinterface-props).
  7. S7
    Benchmark format comparison tab switching

    3 steps, 2 screenshots

    pass
    S7-2.png
    S7 · Benchmark format comparison tab switching
    S7-4.png
    S7 · Benchmark format comparison tab switching
    • Navigated to /benchmarks where the default active tab is 'Model comparison'.
    • Clicked the 'Format comparison' tab; URL updated to ?view=formats and content updated to display format comparison chart and statistics.
    • Navigating to /benchmarks opens the Generative UI Benchmark page with 'Model comparison' active by default.
    • Clicking the 'Format comparison' tab updates the URL to /benchmarks?view=formats and swaps the chart and data to format comparison seamlessly without a full page reload.
    • Switching between tabs is smooth, reactive, and retains valid UI state.
  8. S8
    Page load and React hydration integrity

    2 steps, 3 screenshots

    pass
    S8-1.png
    S8 · Page load and React hydration integrity
    S8-3.png
    S8 · Page load and React hydration integrity
    S8-5.png
    S8 · Page load and React hydration integrity
    • Navigated to /demos; page rendered completely and React hydration succeeded with full interactivity when clicking preset buttons.
    • The /demos page loaded cleanly without hydration errors or rendering crashes.
    • Interactive controls such as preset buttons hydrated properly and updated component state seamlessly upon user interaction.
  9. S9
    Navigation and link integrity for product pages

    5 steps, 4 screenshots

    pass
    S9-2.png
    S9 · Navigation and link integrity for product pages
    S9-4.png
    S9 · Navigation and link integrity for product pages
    S9-6.png
    S9 · Navigation and link integrity for product pages
    S9-8.png
    S9 · Navigation and link integrity for product pages
    • Navigated to home page (/) and verified that it loaded successfully with hero section, framework highlights, and testimonials.
    • Navigated to /cloud/gateway and verified that it loaded successfully with feature details, pricing tiers, and FAQ.
    • Navigated to /cloud/observability and verified that it loaded successfully with feature details, session replay preview, and waitlist form.
    • Navigated to /pricing and verified that it loaded successfully with open source, gateway tiers, observability information, and FAQ.
    • All top-level product routes (/, /cloud/gateway, /cloud/observability, /pricing) loaded cleanly with their expected contents and headers without errors or broken layouts.

Issues

No finding survived the audit. Nothing to fix from this run.

Environment limitations

These failures came from the test environment, not from the application: a credential the sandbox does not hold, a demo nobody may write to, a resource it cannot reach. They are not counted as issues.

  • S2 could not exercise this: Resource fails to resolve during custom generation. A network error occurs because a domain name cannot be resolved. The audit recorded the test environment as the cause, so it is not counted as an issue.

Critic audit

An adversarial second pass over every finding before it reaches the report.

0
findings reviewed
0
re-verified live
0
withdrawn
    • A console error regarding a failed domain resolution in S2 points to a test environment restriction where the sandbox could not reach a required external asset.

    Report

    QA report: external/www.openui.com at hosted

    All nine tested scenarios passed with zero application defects identified across the demo workflows.

    Testing covered core features on the OpenUI demos platform, including generating UI from presets, custom prompt submission, model selection, prompt validation handling, output inspection through tabs and fullscreen views, documentation search, benchmark comparisons, and general navigation integrity.

    All scenarios passed their verification checks without breaking application state or raising defects. During custom generation testing in scenario S2, a domain resolution failure occurred due to test environment sandbox network restrictions, preventing full access to an external resource without impacting application logic.

    With no application issues identified and all standard demo interactions functioning properly, the tested surface area remains stable.

    Run summary
    MetricCount
    Scenarios executed9
    Passed9
    Failed0
    Blocked0
    Findings raised0
    Issues after the audit0
    Withdrawn by the audit0
    Critical / high / medium / low0 / 0 / 0 / 0

    Target: https://www.openui.com/demos · Testing level: deep_feature · Stack: unknown

    Issues

    No issues survived the audit.

    Environment limitations

    These failures came from the test environment, not from the application: a credential the sandbox does not hold, a demo nobody may write to, a resource it cannot reach. They are not counted as issues. They record what this run could not exercise.

    • S2 could not exercise this: Resource fails to resolve during custom generation. A network error occurs because a domain name cannot be resolved. The audit recorded the test environment as the cause, so it is not counted as an issue.
    Scenario results
    ScenarioPriorityResultIssues
    S1 Generate UI from built-in presethighpassnone
    S2 Generate UI from custom prompt and model selectionhighpassnone
    S3 Empty prompt validationhighpassnone
    S4 Inspect generated outputs via tabs and fullscreenhighpassnone
    S5 Whitespace-only prompt validationmediumpassnone
    S6 Documentation search functionalitymediumpassnone
    S7 Benchmark format comparison tab switchingmediumpassnone
    S8 Page load and React hydration integritylowpassnone
    S9 Navigation and link integrity for product pageslowpassnone
    The audit

    The Critic reviewed 0 findings and re-verified 0 of them live in the browser, replaying the reported steps on a fresh page.

    • A console error regarding a failed domain resolution in S2 points to a test environment restriction where the sandbox could not reach a required external asset.
    Coverage and caveats

    In scope: Generative UI prompt submission and validation on /demos; Preset UI generation flows; Output viewing tabs (Raw, Parsed, Fullscreen preview); Documentation search modal functionality; Benchmark page tab switching; Page load integrity and console error monitoring across mapped routes.

    Not covered: External links to GitHub and thesys.dev (origin boundary constraints); Account creation or saving functionality (if any exist, out of scope for unauthenticated hosted run).

    • UI generation may involve network latency; assuming standard timeouts or observable loading states apply during generation.
    • Form validation for required fields uses standard HTML5 constraints or visual error indicators on the page.
    By the numbers
    MetricValue
    Scenarios9 passed, 0 failed, 0 blocked of 9 (34 planned steps)
    Browser actions136 (20 clicks, 11 inputs, 26 navigations, 79 snapshots)
    Screenshots32 (4 explore, 28 scenario, 0 critic), 28 captioned
    Coverage10 pages, 2 forms, 3 flows, 1 console errors
    Audit0 findings, 0 re-verified live, 0 confirmed, 0 promoted, 0 withdrawn
    Model calls140
    Tokens885,579 input, 7,098 output, 8,995 thinking
    Time7 min
    StageCallsInputOutputThinkingSeconds
    explore26203,7132,2171,19375
    plan14,2161,5562,12729
    test111672,2172,9374,021294
    critique14,1162241,28012
    report11,3171643745

    Run log

    stagecallstokenstime
    Explore26207.1k1m 15s
    Plan17.9k29s
    Test111679.2k4m 54s
    Critique15.6k12s
    Report11.9k5s
    Total140901.7k6m 55s
    Intake
    Explore
    Plan
    Test
    Critique
    Report
    • 00:29:12Zexploreexplore started
    • 00:36:07ZexploreExplored /demos (18 controls, 0 forms)
    • 00:36:07ZexploreExplored / (78 controls, 0 forms)
    • 00:36:07ZexploreExplored /benchmarks (24 controls, 0 forms)
    • 00:36:07ZexploreExplored /benchmarks/methodology (22 controls, 0 forms)
    • 00:36:07ZexploreExplored /docs (27 controls, 0 forms)
    • 00:36:07ZexploreExplored /chat (3 controls, 0 forms)
    • 00:36:07ZexploreExplored /cloud/gateway (42 controls, 0 forms)
    • 00:36:07ZexploreExplored /cloud/observability (25 controls, 2 forms)
    • 00:36:07ZexploreExplored /pricing (22 controls, 0 forms)
    • 00:36:07ZexploreExplored /docs/openui-lang (50 controls, 0 forms)
    • 00:36:07ZexploreMapped 10 pages, 2 forms, 3 flows in 26 turns.
    • 00:36:07Zexploreexplore completed in 75s.
    • 00:36:07Zplanplan started
    • 00:36:07ZplanPlanned 9 scenarios (4 high, 3 medium, 2 low).
    • 00:36:07Zplanplan completed in 29s.
    • 00:36:07Ztesttest started
    • 00:36:07ZtestS1 executed (pass)
    • 00:36:07ZtestS2 executed (pass)
    • 00:36:07ZtestS3 executed (pass)
    • 00:36:07ZtestS4 executed (pass)
    • 00:36:07ZtestS5 executed (pass)
    • 00:36:07ZtestS6 executed (pass)
    • 00:36:07ZtestS7 executed (pass)
    • 00:36:07ZtestS8 executed (pass)
    • 00:36:07ZtestS9 executed (pass)
    • 00:36:07ZtestExecuted 9 scenarios: 9 passed, 0 failed, 0 blocked, 0 findings.
    • 00:36:07Ztesttest completed in 294s.
    • 00:36:07Zcritiquecritique started
    • 00:36:07ZcritiqueReviewed 0 findings; 1 possible defect spotted in passed scenarios.
    • 00:36:07ZcritiqueAudit complete: 0 confirmed, 0 withdrawn, 0 promoted, 0 re-verified live.
    • 00:36:07Zcritique1 failure came from the test environment rather than the application. It is reported as an environment limitation, not an issue.
    • 00:36:07Zcritiquecritique completed in 12s.
    • 00:36:07Zreportreport started
    • 00:36:07ZreportReported 0 issues (0 critical, 0 high, 0 medium, 0 low) from 0 findings.
    • 00:36:07Zreportreport completed in 5s.

    Put an agent team on your next pull request.

    Connect a repo, dispatch a Run, and read an audited, evidence-backed report the same day.