6 months later, DataKit has taught me what a modern data studio should actually be:
It should meet you where your data is
Started local-only with DuckDB-wasm (no servers, data never leaves your machine). Then users wanted cloud connections, so we added MotherDuck and PostgreSQL. Your workflow shouldn't change based on where your data lives.
AI should feel natural, not forced
I learned this the hard way with a previous text-to-SQL project. In DataKit, AI isn't the main feature - it's just there when you need it. Works with whatever provider user want to work with. OpenAI, Anthropic, or your local Ollama models.
The biggest surprise maybe for me?
People at my company started using it for daily workflows. What began as a "quick CSV tool" now has Python notebooks, chart generation, workspace persistence. Users taught us what it needed to become.
Privacy should be the default
Running everything client-side means your data stays put. Even AI features work with local models if you prefer. This wasn't a marketing decision - it just made sense.
The tool evolved from user feedback, not a grand vision. Analysts use it for exploration, non-technical folks ask questions in plain English.
You can also try it here: https://datakit.page/
What would your ideal data studio look like? There are things missing here but I'm curious what "main" workflows/features I'm missing.