spreadsheet-xlsx

name	spreadsheet-xlsx
description	Generate real Excel .xlsx workbooks from data. Trigger when the user asks to "export to Excel", "make a spreadsheet", "give me an xlsx", "download as Excel", or wants query/CSV/JSON results as a formatted workbook with headers, number formats, and multiple sheets. Uses openpyxl in the Python sandbox.

spreadsheet-xlsx

Produce a genuine .xlsx (not a CSV renamed) with proper types, number formats, a frozen/bold header row, and auto-sized columns. Use openpyxl — it's pure Python, has no system dependencies, and installs in one step.

Reusable helpers are in snippets.py — copy them in rather than rewriting boilerplate each time.

Setup

uv pip install openpyxl          # pure-python, no apt packages needed

If the data is already in a pandas DataFrame, df.to_excel(path, index=False, engine="openpyxl") is the fast path — but you still want the formatting pass below, so openpyxl directly is usually clearer.

Rules for a workbook that doesn't look amateur

1. Write typed cells, not strings. Put real numbers, dates, and booleans in cells so Excel can sum/sort them. Never write "1,234" or "$1,234" as text — write 1234 and set a number format.
2. Number formats, not pre-formatted strings. Currency '"$"#,##0.00', thousands '#,##0', percent '0.0%' (store 0.25, not 25), dates 'yyyy-mm-dd'.
3. Header row: bold, frozen (ws.freeze_panes = "A2"), and turn on an autofilter over the data range.
4. Column widths: size to the longest value in each column (see snippet) — default widths clip everything.
5. One sheet per logical table. Name sheets meaningfully (ws.title), <= 31 chars, no []/\*?:. Add a small summary sheet first when there are many.
6. Totals go in a labeled row using real =SUM(B2:B100) formulas, not a hardcoded number, so they stay correct if the user edits cells.

Minimal example

from openpyxl import Workbook
from openpyxl.styles import Font
# from snippets.py:
from snippets import write_table, autosize, apply_number_format

wb = Workbook()
ws = wb.active
ws.title = "Revenue"
rows = [
    {"Region": "EMEA", "Revenue": 128400.0, "Growth": 0.12},
    {"Region": "APAC", "Revenue": 98200.0,  "Growth": -0.03},
]
write_table(ws, rows)                       # bold+frozen header, autofilter
apply_number_format(ws, "Revenue", '"$"#,##0')
apply_number_format(ws, "Growth", "0.0%")
autosize(ws)
wb.save("/workspace/revenue.xlsx")

Verify before handing off

• Reopen it: python -c "import openpyxl; wb=openpyxl.load_workbook('/workspace/revenue.xlsx'); ws=wb.active; print(ws.max_row, ws.max_column, ws['B2'].value, ws['B2'].number_format)" — confirm the count, that numeric cells hold numbers (not strings), and the format stuck.
• Tell the user the path and sheet names. Don't paste the binary into chat.