python-visuals

Add the visual with `pbir add visual pythonVisual`, write a matplotlib or seaborn script that ends with `plt.show()`, then inject it using `pbir visuals python --script-file chart.py`. Validate bindings afterward with `pbir validate`.

The most common cause is a missing `plt.show()` call, which is mandatory as the final line of the script. Also verify the script references columns by their nativeQueryRef display names and that only one figure is rendered.

The Service supports matplotlib 3.8.4, seaborn 0.13.2, numpy 2.0.0, pandas 2.2.2, scipy 1.13.1, scikit-learn 1.5.0, statsmodels 0.14.2, and pillow 10.4.0 on Python 3.11. Plotly, bokeh, and altair are not supported because networking is blocked.

Python visuals are capped at 150,000 rows, applied to the deduplicated set of bound Values columns, not the raw fact table. Input is also capped at 250 MB, and individual string values over 32,766 characters are silently truncated.

Use a Python visual only when the chart requires render-time statistical computation like model fits or forecast bands and has no native or Deneb equivalent. If interactivity or cross-filtering matters, use Deneb; for small inline marks, use an SVG measure.

Bind a guaranteed-unique column, such as a transaction key, to the Values role alongside your other fields. Do not use a measure as the key, since that changes the projection kind and prevents distinct-row expansion.