Explain this Map

Explain this Map sends what's on your map to a frontier model on your Databricks workspace and streams back an explanation: the area, the patterns, hotspots and outliers, and what's worth investigating next. It uses your existing Databricks sign-in, so there's no API key and no extra subscription.

One click

1. Set up the map you want explained: layers, styling and zoom.
2. Click the Explain this Map toolbar button (a red map with a sparkle), or Plugins → Databricks DBSQL Connector → Explain this Map.
3. The explanation appears in a few seconds and fills in as it's written. The left side shows the image that was sent.

The first time, it uses your last-used saved connection and picks the best available model automatically. After that it remembers your choices.

It works on any map. Here it explains groundwater prohibition areas around Adelaide, a Unity Catalog table added with Discover Tables, shown in QGIS's dark theme:

You need a saved connection first. If you don't have one, the plugin opens the connection dialog for you.

What the model sees

Alongside a sharp image of the current view, the plugin sends map facts, so answers use your real names and numbers instead of guessing from colours: - coordinate system, scale and area (longitude/latitude) - each visible layer: name, geometry type, feature count, and how many are in view - what each layer's style encodes, for example graduated by 'crashes' in 6 classes: 20-35 (#fee5d9) … 288-4396 (#67000d) - labels, selections, and the Databricks table or SQL behind each layer

Optional knobs

Everything below is optional and remembered between sessions.

Control	What it does
Model	Pick any model on your workspace that can read images (Claude, GPT, Gemini, Llama and more). Only image-capable models are listed
Ask for	Prompt presets: Explain this map · Summary for a report (one polished paragraph) · Patterns and outliers · Title, caption and alt text (for layouts and accessibility) · Next analysis steps (QGIS or Databricks SQL)
Options → Custom prompt	Write your own instruction, e.g. "Explain this map for a road safety committee in Victoria". Leave it empty to use the preset
Options → Include layer and style details	Turn the map facts off to send only the image
Options → Image size	Small (1024 px), Medium (1600 px, default) or Large (2048 px). Larger shows more detail and uses more tokens. Busy maps (for example a detailed basemap) are sent as JPEG so they always fit the model's image limit
Explain	Capture the map again (after panning, restyling or changing options) and explain it

After the answer

• Ask a follow-up about the same map, e.g. "Which areas should we prioritise?". The conversation continues with the same image.
• Copy the explanation, or Save... it as Markdown with the map image beside it (.md + .png, or .jpg for busy maps), ready for a report or wiki.
• The status line shows the model, tokens used and time taken.

Tips

• Style first, then explain. A clear legend (graduated or categorised) gives the model much better material than a single colour.
• Zoom to what matters. The image and the facts describe the current view.
• Add a basemap (see Basemaps) so the model can recognise places.
• Treat the answer like a colleague's first read of the map: useful, and worth checking against your data.

Where it runs

The image and map facts go to the model's serving endpoint on your Databricks workspace, using your saved connection's sign-in and your workspace's model access and governance (including AI Gateway settings, if your admins use them). Model availability depends on your workspace and region. If a model can't read images, or isn't available, you'll see a short message and can pick another in Model.