Skip to main content

QueryFlow — Visual Query Builder

QueryFlow is a drag-and-drop canvas for building SQL-like queries across multiple DocTypes — without writing SQL. Queries run client-side using DuckDB WASM, making them fast even for large datasets.

Opening QueryFlow​

  1. Click [+] in the sheet tab bar
  2. Select Query Sheet
  3. The QueryFlow canvas opens

Canvas Elements​

DocType Nodes​

Drag any DocType from the left panel onto the canvas. Each node represents a table.

  • Click a node to expand its field list
  • Check fields to include them in the output

Join Connections​

Drag from a field on one node to a matching field on another node to create a JOIN.

  • The join type (INNER, LEFT, RIGHT) can be set by right-clicking the connection line
  • Multiple joins can be chained

Filter Nodes​

Drag a Filter node from the sidebar onto the canvas. Connect it to a DocType node. Configure:

  • Field to filter
  • Operator
  • Value (can be a static value or a reference to another node's output)

Aggregate Nodes​

Drag an Aggregate node to add GROUP BY + aggregation:

  • Choose GROUP BY fields
  • Choose aggregate fields (SUM, COUNT, AVG, MIN, MAX)

Sort Node​

Drag a Sort node to add ORDER BY.

Running the Query​

Click the Run button (play icon, top-right of the canvas). The query executes in DuckDB WASM in the browser. Results appear in the sheet grid below.

The generated SQL is shown in the SQL Preview panel (bottom of the canvas) — useful for debugging or copying to run elsewhere.

Performance​

DuckDB WASM runs in your browser. For datasets up to ~100,000 rows, queries typically complete in under 2 seconds. For larger datasets, increase the sample size in settings or filter down your DocType nodes first.

Saving a Query​

Query sheet configurations (node positions, join definitions, filter values) are saved in the workbook. Reloading the workbook restores the canvas exactly.

Limitations​

  • QueryFlow queries are read-only — results cannot be saved back to Frappe
  • Complex recursive CTEs are not supported in the visual builder (use the SQL editor mode for advanced queries)
  • DuckDB WASM loads only data that has been fetched to the client — filters on DocType nodes limit the loaded data before the query runs