---

name: databricks-dashboards description: | Create Databricks AI/BI dashboard JSON files in lvdash.json format. Triggers: ダッシュボード, 可視化, visualize, create dashboard, build dashboard, dashboard JSON, lvdash.json, BI dashboard, visualization, chart, graph, データ可視化, テーブル可視化, analyze data, data analysis, sales dashboard, KPI dashboard, metrics dashboard, report dashboard.

Databricks Dashboards

Overview

AI/BI dashboards (formerly Lakeview dashboards) are defined in .lvdash.json files.

File Format

Dashboard files use .lvdash.json extension. The structure:

Copy{
  "datasets": [...],
  "pages": [
    {
      "name": "<8-char-id>",
      "displayName": "Page Title",
      "layout": [...],
      "pageType": "PAGE_TYPE_CANVAS"
    }
  ],
  "uiSettings": {
    "genieSpace": {
      "isEnabled": true,
      "enablementMode": "ENABLED"
    }
  }
}

Note: The uiSettings block with genieSpace is required. The pageType: "PAGE_TYPE_CANVAS" is required for each page.

Naming Conventions

Query name formats:

• Simple (recommended): main_query - use for most widget queries
• Full format: dashboards/{dashboard_id}/datasets/{dataset_id}_{description} - use for filters with cross-dataset associations

Parameter queries use parameter_ prefix:

• parameter_dashboards/{dashboard_id}/datasets/{dataset_id}_{param_name}

Generate hex IDs: crypto.randomBytes(4).toString('hex') or openssl rand -hex 4

Quick Start

Minimal Dashboard

Copy{
  "datasets": [
    {
      "name": "01f0ac3f",
      "displayName": "Sales Data",
      "queryLines": [
        "SELECT\n",
        "  date,\n",
        "  revenue\n",
        "FROM\n",
        "  catalog.schema.sales"
      ]
    }
  ],
  "pages": [
    {
      "name": "a1b2c3d4",
      "displayName": "Sales Overview",
      "layout": [
        {
          "widget": {
            "name": "e5f6a7b8",
            "queries": [
              {
                "name": "main_query",
                "query": {
                  "datasetName": "01f0ac3f",
                  "fields": [
                    {"name": "date", "expression": "`date`"},
                    {"name": "revenue", "expression": "SUM(`revenue`)"}
                  ],
                  "disaggregated": false
                }
              }
            ],
            "spec": {
              "version": 3,
              "widgetType": "bar",
              "encodings": {
                "x": {"fieldName": "date", "scale": {"type": "categorical"}, "displayName": "Date"},
                "y": {"fieldName": "revenue", "scale": {"type": "quantitative"}, "displayName": "Revenue"}
              },
              "frame": {"showTitle": true, "title": "Daily Revenue"}
            }
          },
          "position": {"x": 0, "y": 0, "width": 6, "height": 4}
        }
      ],
      "pageType": "PAGE_TYPE_CANVAS"
    }
  ],
  "uiSettings": {
    "genieSpace": {
      "isEnabled": true,
      "enablementMode": "ENABLED"
    }
  }
}

Structure Reference

Datasets

Define SQL queries that power visualizations. Use 8-character hex ID for name:

Copy{
  "name": "01f0ac3f",
  "displayName": "Human Readable Name",
  "queryLines": [
    "SELECT\n",
    "  col1,\n",
    "  col2\n",
    "FROM\n",
    "  catalog.schema.table\n",
    "WHERE\n",
    "  condition = 'value'"
  ]
}

Note: Use queryLines (array of strings) instead of query (single string). Each line should end with \n for proper formatting.

Pages

Pages contain layouts with widgets. Use 8-character hex IDs and include pageType:

Copy{
  "name": "a1b2c3d4",
  "displayName": "Dashboard Page Title",
  "layout": [...],
  "pageType": "PAGE_TYPE_CANVAS"
}

Note: pageType: "PAGE_TYPE_CANVAS" is required for each page.

Widget Position

Grid-based layout (12 columns total):

Copy{
  "position": {
    "x": 0,      // 0-11 (column position)
    "y": 0,      // Row position (starts at 0)
    "width": 6,  // 1-12 (columns to span)
    "height": 4  // Rows to span
  }
}

Widget Types

Important: Most chart widgets use spec.version 3. Filter and Counter require version 2, Table requires version 1. Using the wrong version will cause errors.

For detailed widget specifications: See Widget Reference

Common Patterns

Counter Widget (spec.version: 2)

Copy{
  "widget": {
    "name": "c1d2e3f4",
    "queries": [{
      "name": "main_query",
      "query": {
        "datasetName": "01f0ab72",
        "fields": [{"name": "total", "expression": "SUM(`amount`)"}],
        "disaggregated": false
      }
    }],
    "spec": {
      "version": 2,
      "widgetType": "counter",
      "encodings": {
        "value": {"fieldName": "total", "displayName": "Total Amount"}
      },
      "frame": {"showTitle": true, "title": "Total Amount"}
    }
  },
  "position": {"x": 0, "y": 0, "width": 3, "height": 2}
}

Date Range Filter (spec.version: 2)

Copy{
  "widget": {
    "name": "f1a2b3c4",
    "queries": [{
      "name": "parameter_dashboards/01f0a403a6891cc1b5cf06c4960354b8/datasets/01f0ab724630173fb05934717b44e451_param_date_range",
      "query": {
        "datasetName": "01f0ab72",
        "fields": [{"name": "date", "expression": "`date`"}],
        "disaggregated": true
      }
    }],
    "spec": {
      "version": 2,
      "widgetType": "filter-date-range-picker",
      "encodings": {
        "fields": {"fieldName": "date", "displayName": "Date Range"}
      }
    }
  },
  "position": {"x": 0, "y": 0, "width": 3, "height": 1}
}

Text/Markdown Widget

Copy{
  "widget": {
    "name": "d4e5f6a7",
    "multilineTextboxSpec": {
      "lines": [
        "# Dashboard Title",
        "",
        "Description text here."
      ]
    }
  },
  "position": {"x": 0, "y": 0, "width": 12, "height": 1}
}

Note: Use multilineTextboxSpec with lines array. Each element is a line of markdown.

Encodings Structure

Quantitative Fields (measures)

Copy{
  "y": {
    "fieldName": "revenue",
    "scale": {"type": "quantitative"},
    "displayName": "Revenue ($)"
  }
}

Categorical Fields (dimensions)

Copy{
  "x": {
    "fieldName": "category",
    "scale": {"type": "categorical"},
    "displayName": "Category"
  }
}

Color Encoding

Copy{
  "color": {
    "fieldName": "region",
    "scale": {"type": "categorical"},
    "displayName": "Region"
  }
}

Validation

Validate dashboard JSON to catch errors early:

Copy# Syntax check
jq empty dashboard.lvdash.json

# Verify only allowed top-level keys exist (datasets, pages, uiSettings)
jq -e 'keys - ["datasets", "pages", "uiSettings"] | length == 0' dashboard.lvdash.json

# Verify required structure
jq '{
  datasets: (.datasets | length),
  pages: [.pages[] | {name, displayName, widgets: (.layout | length)}],
  hasUiSettings: (.uiSettings != null)
}' dashboard.lvdash.json

# List all dataset names (for debugging references)
jq -r '.datasets[].name' dashboard.lvdash.json

# Check widget-dataset references
jq -r '.pages[].layout[].widget.queries[]?.query.datasetName // empty' dashboard.lvdash.json | sort -u

Common errors:

• Unexpected top-level keys (only datasets, pages, uiSettings are allowed)
• Missing datasets array
• Missing uiSettings block
• Widget referencing non-existent dataset name (must match 8-char hex ID in datasets)
• Invalid page/widget/dataset name format (must be 8-char hex, e.g., a1b2c3d4)
• Missing required encoding fields for widget type

Required Rules

1. 8-char hex IDs - Required for page name, widget name, and dataset name/datasetName
2. queryLines format - Datasets must use queryLines (array of strings), not query (single string)
3. pageType required - Each page must include "pageType": "PAGE_TYPE_CANVAS"
4. Query name format - Use main_query for simple widget queries; use full format dashboards/{dashboard_id}/datasets/{dataset_id}_{description} only for filters with cross-dataset associations
5. Parameter prefix - Filter widget queries must use parameter_ prefix
6. spec.version - Filter and Counter require version 2, Table requires version 1, all other chart widgets use version 3
7. Frame with title - Always set spec.frame with showTitle: true and title. Widgets without titles severely degrade dashboard readability
8. uiSettings required - Every dashboard must include the uiSettings block with genieSpace configuration
9. Text widgets - Use multilineTextboxSpec with lines array, not textbox_spec

Tips

• Grid layout - 12 columns total, plan widget positions carefully
• Test incrementally - Import after adding each component to catch errors early
• Use disaggregated: false for aggregations, true for raw field access

References

• Widget Reference: Complete widget specifications
• Examples: Full dashboard examples