Running Python

The AI generates Python; HISAB runs it sandboxed; results land in your sheet.

Pre-installed packages

HISAB ships with CPython 3.11 and a curated set of ~40 pre-installed packages:

• Data: pandas, numpy
• Excel / Office: openpyxl, python-docx, python-pptx, lxml
• PDF: reportlab, pypdf, pdfplumber, pdfminer, pypdfium2
• Charts: plotly (interactive HTML; static PNG needs kaleido, not bundled)
• Images: pillow
• Utilities: rapidfuzz, dateutil, pytz, charset-normalizer
• Standard library: json, math, re, datetime, decimal, collections, itertools, functools, statistics, csv, io, hashlib, base64

Sandbox guarantees

The Python interpreter runs in an isolated CPython process with these restrictions:

• Filesystem read/write limited to a per-run temp folder and the active workbook range API
• Network access disabled by default (skills can opt-in for specific hosts)
• Subprocess spawning blocked
• 30-second wall-clock limit per execution (configurable per skill)
• 2 GB memory limit

Excel I/O from Python

HISAB injects a hisab helper module into every Python run:

import hisab as h

# read a range as a pandas DataFrame
df = h.read_range("Bank Statement", "A1:E1000")

# write back to a (new or existing) sheet
h.write_range("Mismatches", "A1", df_unmatched)

# create a sheet
h.create_sheet("Reconciliation Summary")

# refresh formulas
h.calculate()