Analyse data in the notebook
Import a dataset, profile it, run code cells, and send figures to the draft.
In short
The Data & Analysis panel profiles your data, runs real Python cells in a per-project environment, and records every output as an artifact you can put into the paper.
1Import or collect
Choose "I already have data" to upload CSV, TSV, Excel or Parquet, or "I need to collect data" to work out what you need with an AI discussion. Data enters the project only by your import or by code you approve and run.
2Read the profile
The profile lists each column's type, missing values, distinct values and examples, and flags issues. From a dataset start, it also suggests 3–5 research questions tied to specific columns.
3Run cells
Add a code or markdown cell, write Python, and run it with the ▶ button or Cmd/Ctrl+Enter. Outputs render as text, tables, images and errors with the full traceback.
4Keep the artifacts
Charts and exports are recorded as artifacts with a stable id and the dataset they came from, so the LaTeX workspace can insert them at natural size.
For technical users
The first notebook use downloads a pinned CPython 3.11.16 and base packages; each project then gets a venv layered on that base. %pip installs go through the download system with --require-hashes. A generated collection script is displayed in an editor, is editable, and runs once on approval of exactly that hash; it runs in a separate subprocess with a scrubbed environment, writes only under data/, and must honour robots.txt. Completing the analysis sends artifacts, datasets and findings to the boardroom for a findings round.