I kept rebuilding the same CSV workflows (reconciling exports, creating aging schedules, stitching months together), so I built a small browser-only experiment.
This lets you build full, node-based data workflows powered by DuckDB — all running entirely in the browser.
You can:
- Upload CSVs
- Create and connect nodes
- Reconcile datasets
- Build aging schedules
- Ask questions in plain English and see the DuckDB SQL generated live with full workflows
- Inspect and modify the SQL at every step
Think “Base44-style UX,” but for data workflows instead of apps.
This is intentionally early:
- CSV-only
- No persistence
- Usage is capped
But I’m curious whether this workflow-first approach is useful or a dead end.
https://www.gptbeyond.com
Feedback very welcome — especially missing primitives, UX issues, or workflows you wish this supported.
This isn’t just “explore a CSV” — you can build full, interactive workflows (with AI) by connecting nodes (joins, transforms, reconciliations, agings, etc.), and inspect/modify the DuckDB SQL at every step.
Still very early and intentionally constrained (CSV-only, no persistence), but I’m especially interested in: - workflows people repeat monthly - primitives that feel missing - points where the node model breaks down
Appreciate the thoughtful feedback so far.