Custom visualizations in AI/BI dashboards

Preview

Custom visualizations let you customize charts in AI/BI dashboards beyond the built-in visualization types. Custom visualizations use the Vega-Lite library to render charts from a JSON specification.

Create a custom visualization

Select a dataset.
In the visualization configuration pane, select Custom Viz under the Advanced visualization section.
In the Fields section, add the fields you want to use. Each field has a unique Name. Reference fields in your Vega-Lite specification using these names.
Enter your Vega-Lite JSON specification in the Vega-Lite specification editor.

Example: Layered chart with a rolling mean

This example creates a layered chart that plots raw temperature data points with a rolling mean line overlay, using weather data from the Databricks sample datasets.

Custom visualization chart example.

Create a dataset with the following query:

SQL
SELECT date, temperature AS temp_max
FROM samples.accuweather.historical_hourly_imperial
WHERE city_name = 'singapore'
ORDER BY date

In the visualization configuration pane, under Advanced, select Custom Viz.
Select the dataset you created in the previous step.
In the Fields section, add a field for the date column and set its Name to date.
Add a field for the temperature column and set its Name to temp_max.

Copy the following specification into the Vega-Lite specification editor. If the x-axis is clipped, see Make a chart resize automatically.

JSON specification

JSON
{
  "$schema": "https://vega.github.io/schema/vega-lite/v5.json",
  "width": "container",
  "height": "container",
  "config": {
    "autosize": { "type": "fit", "contains": "padding" }
  },
  "data": { "name": "databricks_query" },
  "transform": [
    {
      "window": [{ "field": "temp_max", "op": "mean", "as": "rolling_mean" }],
      "frame": [-15, 15]
    }
  ],
  "encoding": {
    "x": { "field": "date", "type": "temporal", "title": "Date" },
    "y": {
      "type": "quantitative",
      "scale": { "zero": false },
      "axis": { "title": "Max temperature and rolling mean" }
    }
  },
  "layer": [
    {
      "mark": { "type": "point", "opacity": 0.3 },
      "encoding": { "y": { "field": "temp_max", "title": "Max temperature" } }
    },
    {
      "mark": { "type": "line", "color": "red", "size": 3 },
      "encoding": { "y": { "field": "rolling_mean", "title": "Rolling mean of max temperature" } }
    }
  ]
}

Reference dataset columns

There are two ways to reference columns in a Vega-Lite specification:

Use "field": "{columnName}". The following example assigns the xField column to the x-axis:
JSON
```
"encoding": {
  "x": { "field": "xField", "type": "quantitative" }
}
```
In expressions, use datum["{columnName}"] or datum.{columnName}. The following example defines a new x column from the r and angle columns:
JSON
```
{ "calculate": "datum.r * cos(datum.angle)", "as": "x" }
```

For more information, see datum in the Vega expressions documentation.

Inherit the dashboard theme

Custom visualizations automatically adapt to the dashboard's theme, including light and dark mode. The following chart elements inherit theme values without any changes to your specification:

Axis, legend, title, and header fonts match the dashboard's configured fonts.
Axis grid lines match the dashboard's grid line color for the active mode.
The chart background is transparent over the widget's themed background.

Settings you define in your specification's config block take precedence over the inherited defaults.

Reference theme values in expressions

To style marks with theme-aware values, reference the following signals inside a Vega-Lite expression ({ "expr": "..." }):

Signal	Description
`colors`	Pre-resolved color tokens for the active mode. Use these for common values such as `colors.textPrimary.default`, `colors.gridColor`, and `colors.markHighlightColor`. Do not index these with `[mode]`; they are already resolved.
`mode`	The active color mode, either `'light'` or `'dark'`. Use it to index `dashboardTheme` fields that provide per-mode variants.
`dashboardTheme`	The full theme configured by the dashboard owner, including fonts (`resolvedFontSettings`), the categorical palette (`visualizationColors`), and per-mode colors (`gridLineColor`). Fields with per-mode variants require the `[mode]` index.

The colors and dashboardTheme signals are independent. colors provides convenience tokens resolved for the active mode, while dashboardTheme exposes the complete owner-configured theme. Use colors first, and use dashboardTheme for fonts, the full palette, or any value colors doesn't provide.

The following examples show how to reference theme values:

Use the dashboard's body font for a text mark:

JSON
{ "expr": "dashboardTheme.resolvedFontSettings.fieldValue.fontFamily" }

Use the dashboard's title color for the active mode:

JSON
{ "expr": "dashboardTheme.resolvedFontSettings.fieldTitle.fontColor[mode]" }

Use a color from the dashboard's categorical palette:
JSON
```
{ "expr": "dashboardTheme.visualizationColors[0]" }
```

note

Because colors tokens are already resolved for the active mode, do not index them with [mode]. Use colors.markHighlightColor, not colors.markHighlightColor[mode]. Fields under dashboardTheme that provide per-mode variants, such as dashboardTheme.gridLineColor[mode], require the [mode] index.

Filter other widgets on selection

A custom visualization can act as a cross-filter source: when a user clicks a mark, the selection filters the other widgets on the dashboard. To enable this, add a point selection parameter with the reserved name databricks_mark_selection. The renderer detects this name and connects the selection to the dashboard's cross-filter state.

JSON
"params": [
  {
    "name": "databricks_mark_selection",
    "select": { "type": "point", "fields": ["categoryName"] }
  }
]

The following requirements apply:

The parameter name must be exactly databricks_mark_selection. Any other name is treated as a regular parameter and does not drive cross-filtering.
select.type must be point. Interval (brush) selections are not supported as cross-filter sources.
select.fields must list the field names defined in the Fields section of the widget configuration, not the raw column names.
List only dimension (grouping) fields in select.fields. Aggregated measures, such as SUM(...) or AVG(...), cannot drive a cross-filter.
To select on multiple fields, list them together: "fields": ["categoryName", "regionName"].

Highlight selected marks

To highlight selected marks, use stroke and strokeWidth conditions on a fill-based mark (such as bar, arc, or rect) and keep the color encoding bound to your field. Use { "expr": "colors.markHighlightColor" } for the stroke so the highlight stays legible in both light and dark modes.

The following example filters the rest of the dashboard by categoryName when a bar is clicked. Selected bars get a themed stroke, and unselected bars dim while keeping their color encoding.

JSON specification

JSON
{
  "$schema": "https://vega.github.io/schema/vega-lite/v6.json",
  "data": { "name": "databricks_query" },
  "width": "container",
  "height": "container",
  "config": { "autosize": { "type": "fit", "contains": "padding" } },
  "params": [
    {
      "name": "databricks_mark_selection",
      "select": { "type": "point", "fields": ["categoryName"] }
    }
  ],
  "mark": { "type": "bar", "stroke": null },
  "encoding": {
    "x": { "field": "categoryName", "type": "nominal" },
    "y": { "field": "salesValue", "type": "quantitative" },
    "color": { "field": "categoryName", "type": "nominal" },
    "fillOpacity": {
      "condition": { "param": "databricks_mark_selection", "value": 1 },
      "value": 0.3
    },
    "stroke": {
      "condition": {
        "param": "databricks_mark_selection",
        "empty": false,
        "value": { "expr": "colors.markHighlightColor" }
      },
      "value": null
    },
    "strokeWidth": {
      "condition": { "param": "databricks_mark_selection", "empty": false, "value": 2 },
      "value": 0
    }
  }
}

Resize a chart automatically

To make a chart resize to fit its container, add the following settings at the top level of your specification:

JSON
"width": "container",
"height": "container",
"config": {
  "autosize": {
    "type": "fit",
    "contains": "padding"
  }
}

Example chart specifications

The following specifications show charts that aren't available as built-in visualization types. For more examples, see the Vega-Lite example galleries.

Bullet chart

Define categoryField, currentField, paceField, and targetField in the Fields section.

JSON specification

JSON
{
  "$schema": "https://vega.github.io/schema/vega-lite/v5.json",
  "width": "container",
  "height": "container",
  "data": { "name": "databricks_query" },
  "config": {
    "autosize": { "type": "fit", "contains": "padding" }
  },
  "transform": [
    {
      "fold": ["targetField", "paceField", "currentField"],
      "as": ["measure_name", "measure_value"]
    },
    {
      "calculate": "toNumber(datum.measure_value)",
      "as": "measure_value"
    },
    {
      "calculate": "{ \"targetField\": \"Target\", \"paceField\": \"Pace\", \"currentField\": \"Current\" }[datum.measure_name]",
      "as": "measure_label"
    },
    {
      "calculate": "indexof([\"Target\", \"Pace\", \"Current\"], datum.measure_label)",
      "as": "measure_order"
    }
  ],
  "layer": [
    {
      "mark": "bar",
      "params": [
        {
          "name": "legend_click",
          "select": { "type": "point", "fields": ["measure_label"] },
          "bind": "legend"
        }
      ],
      "encoding": {
        "color": { "field": "measure_label" },
        "opacity": { "value": 0 }
      }
    },
    {
      "transform": [{ "filter": { "param": "legend_click" } }],
      "layer": [
        {
          "layer": [
            {
              "mark": { "type": "bar", "tooltip": true },
              "encoding": { "color": { "field": "measure_label", "legend": null } },
              "transform": [{ "filter": { "field": "measure_label", "oneOf": ["Pace"] } }]
            },
            {
              "mark": { "type": "bar", "height": 7, "tooltip": true },
              "encoding": { "color": { "field": "measure_label", "legend": null } },
              "transform": [{ "filter": { "field": "measure_label", "oneOf": ["Current"] } }]
            },
            {
              "mark": { "type": "tick", "tooltip": true, "thickness": 3 },
              "encoding": { "color": { "field": "measure_label", "legend": null } },
              "transform": [{ "filter": { "field": "measure_label", "oneOf": ["Target"] } }]
            }
          ],
          "encoding": {
            "x": {
              "field": "measure_value",
              "type": "quantitative",
              "stack": null,
              "title": "Value",
              "axis": { "orient": "bottom" }
            },
            "color": {
              "scale": {
                "domain": ["Target", "Pace", "Current"],
                "range": ["#000000", "#bcbcbc", "#A66BBF"]
              }
            },
            "order": {
              "field": "measure_order",
              "type": "quantitative",
              "sort": "descending"
            }
          }
        }
      ],
      "encoding": {
        "y": {
          "field": "categoryField",
          "type": "ordinal",
          "title": "Category",
          "axis": { "labelOverlap": true }
        },
        "tooltip": [
          { "field": "categoryField", "type": "nominal", "title": "Category" },
          { "field": "currentField", "type": "quantitative", "title": "Current" },
          { "field": "paceField", "type": "quantitative", "title": "Pace" },
          { "field": "targetField", "type": "quantitative", "title": "Target" }
        ]
      }
    }
  ]
}

Gauge

Gauge chart example.

Define $valueField and $totalField in the Fields section.

JSON specification

JSON
{
  "$schema": "https://vega.github.io/schema/vega-lite/v6.json",
  "width": "container",
  "height": "container",
  "data": { "name": "databricks_query" },
  "config": {
    "concat": { "spacing": 0 },
    "autosize": { "type": "fit", "contains": "padding" }
  },
  "params": [
    { "name": "ring_max", "expr": "min(width, height) / 2 - 16" },
    { "name": "ring_width", "expr": "max(12, (min(width, height) / 2) * 0.12)" },
    { "name": "ring_gap", "expr": "max(4, (min(width, height) / 2) * 0.03)" },
    { "name": "label_color", "value": "#000000" },
    { "name": "ring_background_opacity", "value": 0.3 },
    { "name": "ring0_percent", "value": 100 },
    { "name": "ring0_outer", "expr": "ring_max + 2" },
    { "name": "ring0_inner", "expr": "ring_max + 1" },
    { "name": "ring1_outer", "expr": "ring0_inner - ring_gap" },
    { "name": "ring1_inner", "expr": "ring1_outer - ring_width" },
    { "name": "ring1_middle", "expr": "(ring1_outer + ring1_inner) / 2" },
    { "name": "arc_size", "expr": "220" }
  ],
  "transform": [
    { "as": "ratio", "calculate": "datum['$valueField'] / datum['$totalField']" },
    { "as": "_arc_start_degrees", "calculate": "360 - ( arc_size / 2 )" },
    { "as": "_arc_end_degrees", "calculate": "0 + ( arc_size / 2 )" },
    { "as": "_arc_start_radians", "calculate": "2 * 3.14 * ( datum['_arc_start_degrees'] - 360 ) / 360" },
    { "as": "_arc_end_radians", "calculate": "2 * 3.14 * datum['_arc_end_degrees'] / 360" },
    { "as": "_arc_total_radians", "calculate": "datum['_arc_end_radians'] - datum['_arc_start_radians']" },
    { "as": "_ring_start_radians", "calculate": "datum['_arc_start_radians']" },
    {
      "as": "_ring_end_radians",
      "calculate": "datum['_arc_start_radians'] + ( datum['_arc_total_radians'] * datum['ratio'] )"
    }
  ],
  "layer": [
    {
      "mark": {
        "type": "arc",
        "color": "lightgrey",
        "theta": { "expr": "datum['_arc_start_radians']" },
        "radius": { "expr": "ring1_outer" },
        "theta2": { "expr": "datum['_arc_end_radians']" },
        "radius2": { "expr": "ring1_inner" },
        "cornerRadius": 10
      }
    },
    {
      "name": "RING",
      "mark": {
        "type": "arc",
        "theta": { "expr": "datum['_ring_start_radians']" },
        "radius": { "expr": "ring1_outer" },
        "theta2": { "expr": "datum['_ring_end_radians']" },
        "radius2": { "expr": "ring1_inner" },
        "cornerRadius": 10
      },
      "encoding": {
        "color": {
          "value": "#307E31",
          "condition": [
            { "test": "datum['ratio'] < 0.33", "value": "#880808" },
            { "test": "datum['ratio'] < 0.66", "value": "#E49B0F" }
          ]
        }
      }
    },
    {
      "mark": { "type": "text", "fontSize": 40 },
      "encoding": {
        "text": { "field": "$valueField" },
        "color": {
          "value": "#307E31",
          "condition": [
            { "test": "datum['ratio'] < 0.33", "value": "#880808" },
            { "test": "datum['ratio'] < 0.66", "value": "#E49B0F" }
          ]
        }
      }
    }
  ]
}

Radar chart

Radar chart example.

Define $key and $value in the Fields section.

JSON specification

JSON
{
  "$schema": "https://vega.github.io/schema/vega-lite/v5.json",
  "width": "container",
  "height": "container",
  "config": {
    "autosize": { "type": "fit", "contains": "padding" }
  },
  "data": { "name": "databricks_query" },
  "transform": [
    { "window": [{ "op": "row_number", "as": "category" }] },
    { "calculate": "datum.category - 1", "as": "category" },
    {
      "joinaggregate": [
        { "op": "count", "as": "numCategories" },
        { "op": "max", "field": "$value", "as": "maxValue" }
      ]
    },
    { "calculate": "2 * PI * datum.category / datum.numCategories", "as": "angle" },
    { "calculate": "100 * datum['$value'] / datum.maxValue", "as": "r" },
    { "calculate": "datum.r * cos(datum.angle)", "as": "x" },
    { "calculate": "datum.r * sin(datum.angle)", "as": "y" },
    { "calculate": "110 * cos(datum.angle)", "as": "label_x" },
    { "calculate": "110 * sin(datum.angle)", "as": "label_y" }
  ],
  "layer": [
    {
      "transform": [
        { "joinaggregate": [{ "op": "count", "as": "numCategories" }] },
        { "aggregate": [{ "op": "max", "field": "numCategories", "as": "numCategories" }] },
        { "calculate": "[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20]", "as": "cats" },
        { "flatten": ["cats"], "as": ["cat"] },
        { "filter": "datum.cat <= datum.numCategories" },
        { "calculate": "2 * PI * datum.cat / datum.numCategories", "as": "angle" },
        { "calculate": "100 * cos(datum.angle)", "as": "x" },
        { "calculate": "100 * sin(datum.angle)", "as": "y" }
      ],
      "mark": { "type": "line", "color": "#ddd", "strokeWidth": 1 },
      "encoding": {
        "x": { "field": "x", "type": "quantitative", "scale": { "domain": [-120, 120] }, "axis": null },
        "y": { "field": "y", "type": "quantitative", "scale": { "domain": [-120, 120] }, "axis": null },
        "order": { "field": "cat" }
      }
    },
    {
      "transform": [
        { "joinaggregate": [{ "op": "count", "as": "numCategories" }] },
        { "aggregate": [{ "op": "max", "field": "numCategories", "as": "numCategories" }] },
        { "calculate": "[20,40,60,80,100]", "as": "levels" },
        { "flatten": ["levels"], "as": ["level"] },
        { "calculate": "[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20]", "as": "cats" },
        { "flatten": ["cats"], "as": ["cat"] },
        { "filter": "datum.cat <= datum.numCategories" },
        { "calculate": "2 * PI * datum.cat / datum.numCategories", "as": "angle" },
        { "calculate": "datum.level", "as": "r" },
        { "calculate": "datum.r * cos(datum.angle)", "as": "x" },
        { "calculate": "datum.r * sin(datum.angle)", "as": "y" }
      ],
      "mark": { "type": "line", "color": "#ddd", "strokeWidth": 1 },
      "encoding": {
        "x": { "field": "x", "type": "quantitative" },
        "y": { "field": "y", "type": "quantitative" },
        "detail": { "field": "level" },
        "order": { "field": "cat" }
      }
    },
    {
      "mark": { "type": "line", "color": "#9467bd", "strokeWidth": 2, "interpolate": "linear-closed" },
      "encoding": {
        "x": { "field": "x", "type": "quantitative" },
        "y": { "field": "y", "type": "quantitative" },
        "order": { "field": "category" }
      }
    },
    {
      "mark": { "type": "point", "filled": true, "size": 50, "color": "#9467bd" },
      "encoding": {
        "x": { "field": "x", "type": "quantitative" },
        "y": { "field": "y", "type": "quantitative" }
      }
    },
    {
      "mark": { "type": "text", "fontSize": 14, "fontWeight": "bold" },
      "encoding": {
        "x": { "field": "label_x", "type": "quantitative" },
        "y": { "field": "label_y", "type": "quantitative" },
        "text": { "field": "$key", "type": "nominal" }
      }
    }
  ],
  "view": { "stroke": null }
}

Radial chart

Radial chart example.

Define $valueField and $colorField in the Fields section.

JSON specification

JSON
{
  "$schema": "https://vega.github.io/schema/vega-lite/v6.json",
  "width": "container",
  "height": "container",
  "config": {
    "autosize": { "type": "fit", "contains": "padding" }
  },
  "data": { "name": "databricks_query" },
  "transform": [
    {
      "aggregate": [{ "op": "sum", "field": "$valueField", "as": "total" }],
      "groupby": ["$colorField"]
    },
    {
      "window": [{ "op": "rank", "as": "rank" }],
      "sort": [{ "field": "total", "order": "descending" }]
    }
  ],
  "layer": [
    {
      "mark": { "type": "arc", "innerRadius": 20, "stroke": "#fff" }
    }
  ],
  "encoding": {
    "theta": {
      "field": "total",
      "type": "quantitative",
      "scale": { "type": "sqrt" },
      "stack": true,
      "sort": "descending"
    },
    "radius": { "field": "total", "scale": { "type": "sqrt", "zero": true } },
    "color": {
      "field": "$colorField",
      "type": "nominal",
      "title": "Sub-Category",
      "sort": { "field": "total", "order": "descending" },
      "legend": { "orient": "right" }
    },
    "tooltip": [
      { "field": "$colorField", "type": "nominal", "title": "Sub-Category" },
      { "field": "total", "type": "quantitative", "title": "Sales" }
    ]
  },
  "view": { "stroke": null }
}

Sunburst chart

Sunburst chart example.

Define outerGroupField, innerGroupField, and sizeField in the Fields section.

JSON specification

JSON
{
  "$schema": "https://vega.github.io/schema/vega-lite/v5.json",
  "width": "container",
  "height": "container",
  "data": { "name": "databricks_query" },
  "config": {
    "autosize": { "type": "fit", "contains": "padding" }
  },
  "transform": [
    { "calculate": "datum['outerGroupField']", "as": "OUTSIDE" },
    { "calculate": "datum['innerGroupField']", "as": "INSIDE" },
    { "calculate": "datum.OUTSIDE + '-' + datum.INSIDE", "as": "OUT_IN" },
    { "calculate": "toNumber(datum['sizeField'])", "as": "SIZE" }
  ],
  "resolve": {
    "scale": { "color": "independent" },
    "legend": { "color": "independent" }
  },
  "layer": [
    {
      "mark": {
        "type": "arc",
        "tooltip": true,
        "innerRadius": { "expr": "min(width, height)/9" },
        "outerRadius": { "expr": "min(width, height)/3" }
      },
      "encoding": {
        "theta": { "field": "SIZE", "type": "quantitative", "stack": true },
        "color": {
          "field": "OUT_IN",
          "type": "ordinal",
          "sort": "ascending",
          "title": "Inner Grouping",
          "scale": {
            "range": [
              "#1DF9B9",
              "#1DE5B9",
              "#1DD1B9",
              "#1DBDB9",
              "#1DA9B9",
              "#3DF23B",
              "#3DDA3B",
              "#3DC23B",
              "#3DAA3B",
              "#3D923B"
            ]
          }
        },
        "order": { "field": "OUT_IN", "sort": "ascending" },
        "tooltip": [
          { "field": "OUTSIDE", "type": "nominal", "title": "Outer Grouping" },
          { "field": "INSIDE", "type": "nominal", "title": "Inner Grouping" },
          { "field": "SIZE", "type": "quantitative", "title": "Count" }
        ]
      }
    },
    {
      "transform": [
        {
          "aggregate": [{ "op": "sum", "field": "SIZE", "as": "total_users" }],
          "groupby": ["OUTSIDE"]
        }
      ],
      "mark": {
        "type": "arc",
        "tooltip": true,
        "innerRadius": { "expr": "min(width, height)/3" }
      },
      "encoding": {
        "theta": {
          "field": "total_users",
          "type": "quantitative",
          "stack": true,
          "sort": "ascending",
          "title": "Users Count"
        },
        "color": {
          "field": "OUTSIDE",
          "type": "ordinal",
          "sort": "ascending",
          "title": "Outer Grouping",
          "scale": { "range": ["#1DD1B9", "#3DC23B"] }
        },
        "order": { "field": "OUTSIDE", "sort": "ascending" },
        "tooltip": [
          { "field": "OUTSIDE", "type": "nominal", "title": "Outer Grouping" },
          { "field": "total_users", "type": "quantitative", "title": "Count" }
        ]
      }
    }
  ]
}

Limitations

Treemap charts aren't supported. Vega-Lite doesn't support treemaps.
Images aren't currently supported in custom Vega visualizations.

Create a custom visualization​

Example: Layered chart with a rolling mean​

Reference dataset columns​

Inherit the dashboard theme​

Reference theme values in expressions​

Filter other widgets on selection​

Highlight selected marks​

Resize a chart automatically​

Example chart specifications​

Bullet chart​

Gauge​

Radar chart​

Radial chart​

Sunburst chart​

Limitations​

Create a custom visualization

Example: Layered chart with a rolling mean

Reference dataset columns

Inherit the dashboard theme

Reference theme values in expressions

Filter other widgets on selection

Highlight selected marks

Resize a chart automatically

Example chart specifications

Bullet chart

Gauge

Radar chart

Radial chart

Sunburst chart

Limitations