Dashboards with DuckDB, written in SQL and Markdown

A Dash dashboard is a Markdown file with SQL blocks. Each block shows its result as a chart, a table or a dropdown that filters other blocks. The queries run on DuckDB in your browser, so your data stays on your machine.

1. 01

   Add your data

   Drop a CSV, Parquet, JSON or Excel file into Dash, or open a DuckDB file. You can also query a file by its URL.

   SELECT * FROM 'https://example.org/services.parquet'

2. 02

   Write a block for each chart

   The view line of a sql block sets how its result is shown. Pages have bar, line, area, scatter, pie and radar charts, tables and single numbers.

   ```sql busiest_stations
   -- view: bar
   -- x: StationName
   -- y: stops
   SELECT StationName, count(*) AS stops
   FROM 'services.parquet'
   GROUP BY ALL ORDER BY stops DESC LIMIT 10
   ```

3. 03

   Filter with a dropdown

   A block with the select view shows a dropdown. Other blocks read the picked values with refs.<name>() and run again when the pick changes.

   ```sql service_type
   -- view: select
   -- column: ServiceType
   SELECT DISTINCT ServiceType FROM 'services.parquet'
   ```
   
   ... WHERE ServiceType IN
         (SELECT ServiceType FROM refs.service_type())

4. 04

   Arrange the blocks

   Write text between the blocks to explain the numbers, and use columns to put charts side by side. The dashboard stays one Markdown file, so you can keep it in git.

   <columngroup>
   <column span="7">
     ...chart...
   </column>
   <column span="5">
     Utrecht Centraal is the busiest station.
   </column>
   </columngroup>