Quick start
This page builds a first dataflow in the hosted instance: a small table made in a Data Loading node, drawn as a bar chart by a Vega-Lite node. It takes a few minutes and needs nothing installed. If you run Curio on your own computer instead (see Install), skip the sign-in step; the rest is the same.
Sign in
Open the hosted instance and choose Create account on the sign-in page. A name, a username and a password are enough; the email is optional. A new account opens on the Projects page, which already holds its own copies of the example dataflows for you to open and change.
Continue as Guest lets you look around without an account. A guest cannot save dataflows, but what they change in an example, a new dataflow once they add an agent or a dataset to it, and the datasets they import or download stay theirs alone until they sign out or the server restarts.
Create a dataflow
On the Projects page, click + New Dataflow. The canvas opens with the built-in nodes in the rail on the left (hover a tile to see its name) and the top bar along the top: the File, View and Share menus, the save state, and a button for each catalog. Click the dataflow's name at the top to rename it.
Load a small dataset
Drag the Data Loading tile onto the canvas. The node opens on its Data box; click its Code tab and replace the code with this Python, which builds a small table and returns it to the next node.
import pandas as pd
d = {'a': ["A", "B", "C", "D", "E", "F", "G", "H", "I"], 'b': [28, 55, 43, 91, 81, 53, 19, 87, 52]}
df = pd.DataFrame(data=d)
return dfClick the orange play button at the left of the node's header, or press Ctrl+Enter (Cmd+Enter on a Mac) while you edit it. The node shows Done when the run succeeds and Error when it fails.

Chart it with Vega-Lite
Drag the Vega-Lite tile onto the canvas, then drag from the handle on the right edge of the Data Loading node to the handle on the left edge of the Vega-Lite node to connect them. The Vega-Lite node opens on its grammar editor. It may already hold a starter chart picked from the table's columns; replace it with this spec:
{
"mark": "bar",
"encoding": {
"x": {"field": "a", "type": "nominal", "axis": {"labelAngle": 0}},
"y": {"field": "b", "type": "quantitative", "stack": null}
}
}Click the Vega-Lite node's play button, and the chart appears in the node with one bar per letter. Playing a node first runs any node upstream of it that has not run yet or whose code changed, and Run all nodes, at the bottom of the rail, runs the whole dataflow.
Save it and bring your own data
Save with File > Save dataflow, or click the save state next to the Share menu, which reads Saved once everything on the canvas is saved. After the first save, Curio also saves your changes automatically as you work. File > Save dataflow as downloads the dataflow as a file, which File > Load dataflow opens as a new dataflow.
To chart a file of your own, drop a CSV, GeoJSON or other supported file on a fresh Data Loading node's Data box, or click Choose file there and pick it: Curio adds it to your Data Catalog and the node loads it. A dataset already in the Data Catalog can be dragged from the rail's Data Catalog list onto the canvas instead. Data Catalog covers datasets in depth, and Charts and maps covers what Vega-Lite and Autark can draw.

