Isolate a chart display issue with AI
Turn a chart that looks wrong in one notebook into a short example somebody else can run and recognise.
Isolate a chart display issue with AI when a colleague can open your notebook but cannot see what went wrong. Imagine that the labels on a heatmap disappear only after its plot narrows. The minimal-example guide helps an agent keep the data, the specified rendering backend and the resize action needed to show that symptom in a small file.
Start with a handful of invented values, then try the example from a clean Python session. An interactive notebook often remembers imports or settings absent from the script you send. For a browser-only problem, a local preview and two screenshots can show which labels were visible before and after the resize. A sentence saying what should remain legible makes the difference easier to inspect.
What it does
- Reduce a chart to the one visual feature that disappears.
- Replace a private data file with a small repeatable set of values.
- Select the rendering backend explicitly so another reader sees the same code path.
- Keep a screenshot for each action that changes the display.
- Record the library versions alongside a plain description of the expected result.
Install and use
- Use Python and the chart’s HoloViz packages. For automated browser checks, add Playwright and its matching Chromium.
- Write down the visible symptom and remove parts of the example that do not influence it.
- Run the small file from a fresh session and compare the display with the expected result.
FAQ
Source
Made by HoloViz Dev. Original skill: minimal-example, published on GitHub under the BSD-3-Clause licence.
View on GitHub