Data analysis

gen_data_analysis · slug data-analysis · icon 📊 · outputs markdown, image, data

Paste a CSV (or describe the data you have) and ask a question. A data scientist profiles it, analyses it in Python, produces charts as PNG, and an analyst writes a findings report that says what the numbers mean and what to do next.

Inputs

Field	Type	Notes
Data	text	required, 10-200,000 chars — CSV with a header row, or a description (a realistic sample is synthesised and clearly labelled as such)
What do you want to know?	text	required, 5-800 chars — the decision or question the analysis should answer
Context	string	optional, 400 chars max
Maximum charts	number	default 4, 1-8
Chart style	pick	Clean minimal · Dark editorial · Colourful presentation (advanced)
Report for	pick	A manager who needs the answer · An analyst who wants the detail · A general reader (advanced)

Pipeline

Step	Role	Skills	Tier	Depends on
profile	data-scientist	code.execute, files.write, files.read	cheap	—
analyze	data-scientist	code.execute, files.write, files.read, files.list	cheap	profile
findings	analyst	docs.write_markdown, files.write, llm.generate	standard	profile, analyze

Both profile and analyze run real Python via code.execute — this is the one seed generator where the numbers and charts in the final report are computed by an actual interpreter (pandas/numpy/matplotlib in the sandbox image) rather than described by a language model.

Example run

{
  "generatorId": "gen_data_analysis",
  "inputs": {
    "data": "date,region,orders,revenue\n2026-01-03,North,42,1830.50\n2026-01-03,South,31,1290.00\n...",
    "question": "Which region is growing fastest, and is revenue per order changing?"
  }
}