Run calculations with run_python

Ask for the calculation#

Loupist uses its built-in calculation tools for supported operations and run_python for custom numerical work. You ask in Chat; you do not need a shell or a separate Python installation.

Example prompt using invented inputs:

Use run_python to calculate the margin from revenue of 100
and operating profit of 20. Show the code, assumptions and result.
Save a small CSV in outputs/ so I can inspect it.

The arithmetic below is an example, not a company figure:

from pathlib import Path
import pandas as pd

revenue = 100
operating_profit = 20
margin_pct = operating_profit / revenue * 100
Path("outputs").mkdir(exist_ok=True)
pd.DataFrame([{"margin_pct": margin_pct}]).to_csv(
    "outputs/margin.csv", index=False
)
print(margin_pct)

Inspect what ran#

1. Expand the answer's work receipt and open Ran Python.
2. Read the code and output. Check units, period and inputs.
3. Open the resulting workspace file or Work item beside Chat.
4. For source-based figures, inspect the linked input evidence too. Running code alone does not retain an original as a Source.

A code block in an answer is not evidence it ran. Look for the actual completed step and saved output.

Offline runtime#

Python runs in a pinned offline runtime. It has no shell, network, DNS, process spawning or host environment. The numerical packages include numpy, pandas, scipy, matplotlib, statsmodels and scikit-learn. It cannot download a missing package.

The default timeout is 60 seconds, up to 120 seconds. Guest memory is bounded at 512 MiB. Large workspaces can produce a partial file snapshot; the result reports omissions. Split work or provide a smaller input if needed.

Files and cancellation#

Attachments are read-only. Top-level workspace files are read-only to Python; allowed writes include outputs/ and admitted workspace subfolders. The app checks copy-back destinations and avoids overwriting a host file changed during the call.