DataForge 1.0

DataForge is a VS Code analysis workbench that puts a Power Query-style interface on top of DuckDB.

Supported sources

• XLSX (DuckDB excel extension)
• CSV
• TXT / TSV
• Parquet
• JSON / JSONL

XLS and XLSB are intentionally excluded from 1.0.

Core workflow

1. Add files.
2. For XLSX, choose one or more detected sheets.
3. Preview data without loading the complete dataset into the dashboard.
4. Apply visual transformations; DataForge compiles them to DuckDB SQL.
5. Export results to CSV, Parquet, or XLSX.

Quick analysis shortcuts

• Combine Files / Sheets
• XLOOKUP / Merge
• SUMIF / SUMIFS
• IF / TRUE-FALSE
• Compare actual vs expected with difference and difference %
• Reconcile two datasets
• Duplicate Finder
• Rate Variance and potential excess
• Ageing buckets
• Partial Header Match
• Data Profile
• DuckDB SQL
• AI Assistant

Keyboard shortcuts

• Ctrl+Alt+O — Add files
• Ctrl+Alt+M — Combine
• Ctrl+Alt+L — Lookup / Merge
• Ctrl+Alt+R — Reconcile
• Ctrl+Alt+D — Duplicates
• Ctrl+Alt+P — Profile
• Ctrl+Alt+E — Audit / Exceptions

DuckDB runtime

DataForge 1.0 uses the DuckDB CLI as its execution engine. It searches in this order:

1. runtime/win32-x64/duckdb.exe inside the extension
2. dataforge.duckdbPath
3. common user locations
4. duckdb.exe on PATH

If DuckDB already works from CMD/PowerShell, DataForge should normally detect it automatically.

The source ZIP intentionally does not pretend to include a Windows DuckDB binary if one was not available at build time. Copy a trusted portable duckdb.exe into runtime/win32-x64/ before packaging if you want a fully self-contained build.

XLSX behavior

DataForge reads XLSX workbook metadata itself to discover sheet names. Selected sheets are then queried using DuckDB read_xlsx(); Microsoft Excel and PowerShell automation are not used.

On first XLSX use, DuckDB may need its official excel extension. If corporate network policy blocks extension downloads, install/cache the extension through your approved DuckDB setup or bundle the matching excel.duckdb_extension under runtime/.

Python

Python is optional. DataForge detects existing Python installations (py -0p, python --version) but does not require or automatically modify them.

AI

• GitHub Copilot / VS Code Language Model API: uses the model access available in VS Code.
• Gemini API: optional API key stored in VS Code SecretStorage.
• DataForge sends table/schema metadata by default; DuckDB remains responsible for actual data processing.

Source safety

Source files are queried read-only. Visual transformations are represented as SQL expressions/views in the DataForge session. Export creates a new output file.