datadevquest-challenges

DDQ2026-10: Image Exports

A picture is worth a thousand words. We only need 150.


Code	`#DDQ2026-10`
Levels	Beginner, Intermediate, Bonus
Tools	REST API or Tableau Server Client (TSC), Tableau MCP (optional), any AI model that can read images
Presented	DataDev Tableau User Group, October 2026
Solution	Walked through at the November 2026 TUG and added to this folder
Author	Pablo Gomez

Open the live demo to see what a finished result looks like: switch between the original Tableau export and the accessible version, change the language, and hear what a screen reader reads.

One dashboard, two files

Tableau can export the same view as a PNG or an SVG. They look the same, but they are very different files.

PNG	SVG
Pixels: a grid of coloured dots	Code: XML text you can open in any editor
Blurs when you zoom in	Sharp at any size
You can look at it, not edit it	You can search it, edit it and add to it
Best for AI models that read images, email and Slack	Best for web pages, print and accessibility

Try it: open both files from the sample folder in a text editor.

What’s new

SVG export now works through the API, not only in Tableau Desktop.

REST API

The Query View Image endpoint accepts format=SVG: GET /api/<api-version>/sites/<site-id>/views/<view-id>/image?format=SVG
SVG needs Tableau Server 2026.2 or later (REST API 3.29 or later). PNG is still the default and works on every version.
Filter the image with vf_<field>=<value>, for example ?vf_Region=West.
Tableau Server Client is adding the same option through ImageRequestOptions. Check your version; if it isn’t there yet, a plain requests call works.

Tableau MCP

The get-view-image tool returns PNG or SVG.
Its documentation recommends PNG when the image will be analysed, and SVG when it will be shown to people. That advice is the design of this challenge.

Why it matters

Screen readers cannot read pixels.

A dashboard shared as an image is silent to a screen reader.
Alt text is usually missing, or just says “dashboard”.
Nobody writes it by hand for 50 views.

The fix:

AI drafts the description from the PNG.
The SVG carries the text, so it travels with the image.
A person checks it before it ships.

The challenge

1. Export	2. Describe	3. Ship
Download the same view twice from the REST API: once as PNG, once as SVG.	Send the PNG to an AI that can read images. Ask for alt text (150 characters max) and a short summary.	Write the text into the SVG as `<title>` and `<desc>`, by hand or with code. Screen readers read it aloud.

The PNG goes to the AI. The SVG goes to the people.

Your quest

Pick your level. Every level starts with the same dashboard.

Beginner

Export one view as PNG and SVG
Ask an AI to describe the PNG
Print the alt text and summary

Intermediate

All three steps: Export, Describe, Ship
Repeat it for each value of a filter, e.g. one accessible SVG per Region (?vf_Region=West, ?vf_Region=East, …)

Bonus: make it reusable

Package the process so anyone can run it
Build a Claude Skill, a Gemini Gem, a Custom GPT or your own app

What a finished result looks like

What a screen reader announces before and after, for the sample infographic. The infographic is in Spanish; with a language switch, someone who doesn’t read Spanish can understand it in English. These values were read from Chrome’s accessibility tree, which is what screen readers use.

What a screen reader hears, before and after, in English

The only lines that change in the SVG. The other ~970 lines of drawing stay exactly as Tableau wrote them.

The top of the SVG, before and after

Get started

You need

A Tableau Cloud or Server 2026.2+ site (free with the Developer Program)
A Personal Access Token
Python with tableauserverclient or requests
An AI that can read images: Claude, Gemini, ChatGPT or a local model
A sample workbook such as Superstore

No site handy? Start at step 2 with the files in sample: an infographic exported from Tableau as PNG and SVG.

On an older Server? PNG export works for steps 1 and 2. For step 3, export the SVG from Tableau Desktop and start there.

Tips

Look at the top of the SVG. Tableau already writes a <title> and a <desc>, but they are placeholders.
Ask the AI for JSON (for example {"alt": "...", "summary": "..."}) so your code can read the answer.
Use AI for the words, code for the edit. Tableau SVGs are large; an AI rewriting the whole file may change or cut parts of the drawing.
Mind the language. Write the text in the infographic’s language and set lang on the <svg> tag so screen readers pick the right voice.
<img> tags ignore the title inside an SVG. The text works when the SVG is placed inside the page (inline) or opened on its own. If you embed it with <img src="chart.svg">, also add alt="..." to the tag.
Use sample data. For real dashboards, follow your organisation’s AI policy before sending images to an external model.

Post your solution in #datadev-quest-challenges on the Tableau #DataDev Slack
Challenge details at datadevquest.com
A solution walkthrough at the November 2026 TUG, added here afterwards. Star or watch the repo so you don’t miss it

Credits

Sample infographic: Océanos de Plástico by Pablo Gomez, Data Monster, Earth Day 2023. Data source: Our World in Data.

Back to all challenges