Data & files

Your work needs data going in and produces files coming out. Both live in the project's workspace, the sandbox filesystem, which you manage from the Files tab in the sidebar.

Uploading data

Upload a dataset from the Files tab and it lands in the workspace, where any cell can read it from disk. Then point the agent at it by name:

Load transactions.parquet and show me the date range and row count.

For anything sizeable, uploading beats pasting a sample into chat. The code reads the real file, and you only move the data once.

There's a difference between attaching a file to a chat message and uploading it to the workspace. An attachment gives the agent the file as context (good for a data dictionary or a spec). An upload puts it on disk where code can open it. For anything you'll actually load, upload it.

Browsing the workspace

The Files tab shows what's on disk: datasets you uploaded, files your code or the agent wrote, and intermediate artifacts. You can browse the tree and download anything. When the project is connected to GitHub, this view also reflects each file's git status.

The terminal

Sometimes a shell is just faster. The notebook has a terminal panel into the sandbox, good for installing a package, unzipping an archive, peeking at a file with head, or checking nvidia-smi on a GPU runtime. It's there when you want it and out of the way when you don't.

The variable inspector

The variable inspector lists what's currently in the kernel: your variables, their types, and a peek at their values. It's a quick way to confirm a DataFrame loaded with the shape you expected, without writing a cell to print it.

Getting output back out

Anything the notebook produces (a cleaned dataset, a model file, a chart, a report) is a file in the workspace you can download. From there it can also go further: commit it to GitHub, or publish a notebook to Hugging Face or Kaggle. You can also download the whole notebook as .ipynb, output included.

The workspace has a storage limit that depends on your plan. If you're working with large datasets or saving big checkpoints, keep an eye on it.