Getting started with SciStudio
SciStudio is a workspace for building workflows over scientific data: you wire together blocks — load, process, analyze, save — on a canvas, run them, and inspect the results, with an AI assistant on hand the whole time. This page takes you from an empty project to your first run.
Choose how to use SciStudio
Install the SciStudio desktop app from the Releases page, then choose a mode when you launch it:
- Desktop opens the SciStudio window. To use its embedded AI assistant, install and sign in to one supported CLI provider — see Install a provider. You can use the canvas without an AI provider.
- External AI lets you use SciStudio in the built-in browser of an AI app such as ChatGPT. The browser must support WebMCP. Follow Using SciStudio in your AI app to connect. This path uses the AI app's own conversation and does not require a local agent CLI for that conversation.
1. Create a project
Everything lives in a project — a folder that holds your workflows, your custom blocks, your notes, and the record of every run. Create one from the start screen (or open an existing one). Inside, you will find:
| Area | What it is |
|---|---|
| Canvas | The center, where you build workflows by placing and wiring blocks. |
| Block palette | The list of available blocks (built-in + any installed packages) you drag onto the canvas. |
| Parameter panel | Appears when you select a block; shows that block's settings. |
| Bottom panel | Tabbed: run logs, plots, run history, and git branches. |
| Preview | Shows the data on any port you click. |
| AI chat | In Desktop mode, the embedded assistant. In an AI app, use that app's conversation to ask it to build, fix, or explain things. |
The AI browser layout places the sidebar on the right and shows Preview as a separate card. The workflow steps below apply in either layout.
If you are new to SciStudio, the start screen can also show Run Your First SciStudio Workflow. It creates a real tutorial project with a small fluorescence table, then walks you through creating a normalization custom block, building and running the workflow, creating a plot card, and reviewing run history. You can dismiss the prompt, hide it permanently, or restart the tutorial later from the start screen when the prompt is available.
2. Build a small workflow
A workflow is a graph of blocks. Let's build the simplest useful one: load a file → transform it → save the result.
- Add a Load block. Drag Load from the palette onto the canvas. Select
it; in the parameter panel pick the data type (e.g.
DataFrame) and the file to read. - Add a transform. Drag in a processing block — a built-in one, a block from an installed package, or one you wrote yourself (see writing-blocks.md). Wire Load's output port to its input port. The canvas only lets you connect ports whose types are compatible, so you cannot wire a table into an image block by mistake.
- Add a Save block. Drag in Save, wire the transform's output into it, and set where to write.
That three-block pipeline is a complete workflow. The full mechanics of the canvas — palette, wiring, parameters, variadic ports — are in using-the-gui.md.
3. Run it
Run the workflow. The bottom panel's logs tab streams progress; each node shows its status on the canvas (running, done, failed). When it finishes, the output files are written and every port is filled with data you can inspect.
If something fails, the logs and the node status tell you where; the AI assistant can read the same logs and help you fix it.
4. Look at the data
Click any port — input or output — to open its preview. SciStudio shows a bounded view of the data without loading the whole thing into memory, so even a very large object previews instantly. Tables show as tables, images as images, spectra as plots. See previewing and, to make your own quick figures, writing-plots.md.
5. It is all recorded
Every run is saved. The run history remembers what you ran, with which parameters, and lets you restore any of it later; branches let you keep several variants of a pipeline side by side. See history-and-branches.md.
Where to go next
| You want to… | Read |
|---|---|
| Learn the canvas, running, and previews in depth | using-the-gui.md |
| Learn SciStudio hands-on, step by step | learning-center.md |
| See every built-in block and what it does | built-in-blocks.md |
| Go back to past work; keep pipeline variants on branches | history-and-branches.md |
| Get the AI assistant to do the work | ai-assistant.md |
| Connect from an AI app's built-in browser | using-with-ai-apps.md |
| Make a quick plot of a result | writing-plots.md |
| Explore data interactively | miniapps.md |
| Write your own block | writing-blocks.md |
| Make your own data type | custom-types.md |
You rarely have to do any of this by hand — the AI assistant can build workflows, write blocks and plots, and tune parameters for you. These pages are here for when you want to understand or do it yourself.