SQL to Dashboard#

Use when (1) user provides a SQL query and asks to generate a chart, dashboard, or data visualization from the results. (2) user says "show this as a graph", "make a dashboard from this query", "visualize the results", or "plot this data". (3) user has a query result and wants a visual summary rather than raw rows.

Core Position#

This skill solves the specific problem of: SQL query results are rows and columns — a visual chart makes trends, comparisons, and anomalies immediately visible.

This skill IS NOT:

• A SQL writing or debugging tool — it works with queries the user provides
• A database management tool — it does not modify schema or data
• A BI platform guide — it generates chart specs, not platform-specific dashboards

This skill IS activated ONLY when: SQL query or result set + visualization/dashboard intent are both present.

Modes#

/sql-to-dashboard#

Default mode. Takes a SQL query or result set and outputs a chart specification.

When to use: User provides SQL and wants to see the data as a chart.

/sql-to-dashboard/multi#

Combines multiple queries into a single dashboard with multiple panels.

When to use: User has several related metrics they want on one screen.

Execution Steps#

Step 1 — Analyze the Query Result#

1. Receive SQL query (pasted text or query result as CSV/table)
2. If only query provided, analyze it to infer result structure:
   • SELECT columns → X-axis candidates (time, category) or Y-axis (numeric values)
   • GROUP BY → X-axis or legend groupings
   • ORDER BY with LIMIT → top-N pattern, suggest horizontal bar chart
   • COUNT, SUM, AVG aggregates → Y-axis values
   • Date/time columns → time series → line chart
   • String/category columns → bar chart or pie chart
3. Identify column types:
   • Numeric (int, float) → value axis
   • Date/datetime → time axis or category axis
   • String/category → label axis, legend, or grouping

Step 2 — Suggest Chart Type#

Query PatternChart TypeReasonTime series (GROUP BY date)Line chartShows trend over timeTop-N with COUNT/SUMHorizontal barEasy to compare magnitudesPie/sum by categoryPie or donutShows proportionTwo numeric columnsScatter plotShows correlationMultiple aggregatesGrouped barCompares categories across groupsCumulative sumArea chartShows accumulationYes/No countDonut chartBinary proportion

Step 3 — Generate Dashboard Spec#

Output a specification in a format appropriate to context:

Plotly JSON spec:

json

{
  "data": [{
    "x": ["Jan","Feb","Mar"],
    "y": [120, 80, 150],
    "type": "scatter",
    "mode": "lines+markers",
    "name": "Revenue"
  }],
  "layout": {
    "title": "Monthly Revenue",
    "xaxis": {"title": "Month"},
    "yaxis": {"title": "Revenue ($)"}
  }
}

Grafana panel JSON:

json

{
  "title": "Monthly Revenue",
  "targets": [{"expr": "sum(revenue)"}],
  "fieldConfig": {"defaults": {"unit": "currencyUSD"}}
}

Mermaid (for simple data):

Rendering diagram...

Step 4 — Validate#

• Column mapping to axes is correct
• Chart type matches the query's data shape
• Axes are labeled with column names or derived names
• No data values are invented or truncated

Mandatory Rules#

Do not#

• Do not change the query — work with exactly what the user provides
• Do not invent data points not in the result set
• Do not use pie charts for >7 categories
• Do not render a line chart for non-time-series data

Do#

• Preserve exact column names as axis labels
• State which column is mapped to which axis
• Suggest a chart type appropriate for the data shape, even if user suggested a different one
• If the result set is empty, report "No data returned" rather than rendering an empty chart

Quality Bar#

A good output:

• Chart type matches query result shape
• Axes are correctly mapped to the query's columns
• Labels and titles are derived from column names (not generic)
• Spec is valid and renderable in the target tool (Plotly, Grafana, etc.)

A bad output:

• Line chart for non-sequential categories
• Pie chart with 20 slices
• X-axis shows numeric values as categories (should be scatter or bar)
• Empty chart when query returns no data (should explicitly say "no data")

Good vs. Bad Examples#

ScenarioBad OutputGood OutputSELECT date, COUNT(*) grouped by monthBar chartLine chart (time series = trend)SELECT status, COUNT(*) 3 statuses3D pie chartDonut chart with legendEmpty result setEmpty chart rendered"Query returned 0 rows — no chart to display"Very large result (10k rows)Renders all 10k pointsAggregates or samples — states "showing top 100"

References#

• references/ — Chart type decision tree, Plotly/Grafana/Redash spec templates, dashboard best practices