A chart should explain the values it displays. WebVision Canvas helps you work on units, missing data and the accessible alternative to the visual view.
const series = normalizePoints(input);
if (!series.length) return renderEmptyChart();
renderChart(series, { unit: 'ms' });
renderAccessibleTable(series);| Review before applying | Compare actual behavior with the expected result. If a check fails, investigate it before applying more changes. |
|---|---|
| Connection | Base URL, credentials and a reachable server. |
| Project | Distinguish a missing point from a true zero and provide the same values in an accessible table. |
const series = normalizePoints(input);
if (!series.length) return renderEmptyChart();
renderChart(series, { unit: 'ms' });
renderAccessibleTable(series);WebVision Canvas is a Windows coding workspace for developing software with a connected AI model. Bring a presentation task into the project conversation by naming the data source and the screen that uses it. Ask the model to inspect that route before suggesting changes, so the proposal follows the application's existing representation of values.
A chart revision illustrates why this matters. You might need a unit to remain visible and a table to present the same values when the visual view is unsuitable. Discuss missing observations separately from real zeros, and identify where those distinctions could be lost. These are tasks for your own chart implementation, not claims of a chart builder included in the workspace.
Once the expected presentation is clear, request a focused change and review both consumers of the data. Exercise empty input, a genuine zero and an absent observation with your project's tools. Compare the displayed labels and table values rather than accepting a screenshot as proof that the underlying interpretation is correct.
Choose a supported OpenAI-compatible or Anthropic endpoint for the coding conversation. If it offers an unrestricted AI model, you can consider that selection while checking the provider's actual conditions. The phrase does not remove endpoint limits; available tools and context remain dependent on the model and server you configure.
Name the source fields, units and meaning of missing values, then identify the visual component and alternate table. State which values must agree across both views. This gives the proposed edit an observable contract beyond colors and spacing.
Only if your data contract explicitly defines that conversion. Otherwise it may imply a measurement that never occurred. Ask the model to trace normalization and display code, then review a missing observation alongside a genuine zero in your tests.
The workspace supports coding discussion and review of proposed changes. Accessibility still requires checking your application's rendered result with appropriate tools and interaction methods. A table example alone is not evidence that the complete interface is accessible.
Use a small synthetic series preserving the relevant shape and edge cases. Requests go to your configured model endpoint, and other tools may contact separate services. Review those destinations before including confidential observations in the conversation.
WebVision Canvas · Updated 2026-10-25
