Sample prompts

Use these in the chat to test different flows. For Jupyter-first prompts, ensure Jupyter MCP is running (npm run jupyter-mcp) and work/notebook.ipynb is open with a kernel started.

Jupyter-first (files in work folder)

Use when you have uploaded CSVs in the Jupyter file browser (work/ or root), without attaching files in chat.

Single file — load and summarize

I uploaded my data file in the Jupyter file browser (work folder). Add code in the notebook to load it (e.g. pd.read_csv('work/filename.csv') or the filename I used), then run a quick summary and show the first few rows. Run the cells there so I see the result in the notebook.

Two files — summarize and relate

I have 2 CSV files in the Jupyter file browser (e.g. work/holdings.csv and work/trades.csv). Add notebook code to load both (pd.read_csv), summarize structure (columns, dtypes), and create a relation or join if there are common columns. Run the cells in the notebook so I see the result.

Short one-liners

Summarize holdings.csv and trades.csv from the Jupyter work folder.
Load work/holdings.csv and work/trades.csv in the notebook, then merge them on the common column.

With file attached in chat

Use when you attach a CSV in the chat input (paperclip or drag-and-drop).

Summarize attached dataset

Summarize the attached dataset: show columns, types, sample rows, and basic stats. Also output Python code in a ```python block that computes this summary so I can run it.

Guide and code

Guide me through this dataset: (1) summarize and assess data quality, (2) run exploratory data analysis, (3) suggest the best visualizations. For each step provide runnable Python in ```python blocks.

General

Suggest next steps

Suggest 2–4 next analysis steps for this dataset (EDA, visualization, SQL, or grouping). For each step provide runnable Python in a ```python block so I can execute them.

Data quality

Assess data quality: missing values, duplicates, outliers; suggest cleaning or imputation. Provide runnable ```python code that performs these checks on df.

Built-in prompts

Type / in the chat to see built-in prompts (e.g. “I uploaded a file in Jupyter (work folder)”, “Summarize or relate 2 datasets in Jupyter”) and click one to insert it.