Notebook setup and bringing data in
Every notebook on your Mac starts from one setup that you keep in Settings: the folders it can read, the Python packages it has, and the hosts it can reach. Then you bring in data in one of three ways. This page covers both. It applies to the Mac app on a Mac that can run notebooks (see The notebook sandbox).
One setup for every notebook
Open Settings and choose Setup. Everything there applies to every notebook, with nothing to choose per notebook. Folder locations stay on your Mac. Retain stores no credentials, so keep short-lived tokens in your own code or files.
A change applies the next time a notebook's Python starts. A notebook that is already running shows "Setup changed: restart Python to use it", with a Restart button. Restarting clears variables and keeps the files in the notebook's work folder.
Folders
Choose Add Folder to let every notebook read a folder of yours. It appears inside the notebook at /data/ followed by the folder's name, and it is read-only: a notebook can never change it. You can rename or remove a folder. If a folder has moved, Setup says "Folder not found" and a notebook that opens says which folder it skipped.
Packages
Notebooks start with pandas, NumPy, DuckDB, Altair and PyArrow. Add anything else by name, such as polars, or with a version, such as statsmodels>=0.14. Packages are installed when a notebook's Python starts and downloaded once for your Mac. Versions are not pinned unless you type one.
If a notebook installs a package of its own, the notebook's "..." menu lists it under Only in This Notebook, with Add to My Setup to keep it for every notebook. Reset Python Environment rebuilds Python from the base and your setup.
Hosts and tunnels
A notebook can reach PyPI and the hosts you list under Access, over HTTPS, and nothing else. A refused host says to add it to your setup. For a database or another service that does not go through a web proxy, add a tunnel: a name, the host and port, and a local port. In the notebook it appears at 127.0.0.1 on that local port and connects to the host from your Mac.
Every notebook, and the AI working in any notebook, can use every host and tunnel you add.
The notebook's work folder
Each notebook has its own work folder, where its code reads and writes at /work. It lives on your Mac at Retain, Notebooks, in a folder named for the notebook, and it is renamed if the notebook is. Choose Reveal Work Folder in Finder from the "..." menu to open it. Its files survive restarts and resets.
Ways to bring data in
- A declared folder. Add it in Setup and read it at /data. Use this for files you already keep, or large ones. Nothing is copied or uploaded.
- A file attached in chat. In a notebook's chat on the Mac, attach a CSV, TSV, Parquet, JSON or JSON Lines file of up to 500 MB. It is uploaded to that notebook, and a data cell is added with its name, size, columns, row count and first rows. The count reads "about" until a run has counted every row. Code reads the whole file by the cell's name. Spreadsheet workbooks are refused: save the sheet as CSV first.
- A file dropped into the work folder. Drag files onto the notebook, or put them in the folder from Finder. They stay on your Mac.
What the AI sees
The AI in a notebook gets a list of what is in /data and /work: names, sizes, types and dates, not contents. It reads a file by running code over it, and it quotes numbers that code computed instead of typing a file's rows into a cell. A result's caption names the files it read, with their size and date. See Notebooks: cells, Run and outputs.
Need help?
Can't find what you're looking for? Contact our support team for a quick reply.
Contact Support