PPTX

Working with PowerPoint (.pptx)

You build and edit presentations by writing Python code, not by filling in a fixed schema. The docgen tool's run_python runs your code in the app's bundled interpreter, which has these libraries pre-installed:

• pptx (python-pptx) — create/edit decks, load templates, add slides, text, images, tables, shapes.
• fitz (PyMuPDF) and PIL (Pillow) — rasterize PDFs to images, image ops.
• docx, openpyxl — Word / Excel, if the task drifts there.

run_python is real execution with full filesystem access. print(...) is captured and returned to you. Iterate: write code → render → inspect → fix → re-render.

Reading / analyzing a deck

To extract text, use the files read tool on the .pptx (it returns per-slide text), or write a quick script:

from pptx import Presentation
prs = Presentation("/path/to/deck.pptx")
for i, slide in enumerate(prs.slides, 1):
    print(f"## Slide {i}")
    for shape in slide.shapes:
        if shape.has_text_frame:
            print(shape.text_frame.text)

Creating / editing

Start from the user's template when they have one (Presentation("template.pptx")) — it preserves their masters, fonts, and branding. Otherwise Presentation() gives a blank 4:3 deck; set 16:9 with prs.slide_width = Inches(13.333); prs.slide_height = Inches(7.5). Add content with python-pptx (add_slide, add_textbox, add_picture, add_table, add_shape). Save to a real path, and pass that path to run_python's output_files argument — the saved deck then comes back as a downloadable tile the user can open directly, so you don't have to make them hunt for the path. Still mention where it landed.

Visual QA (required)

Your first render usually has real issues — overflow, overlap, misalignment. Render and look before declaring done:

import os, subprocess, tempfile, fitz
soffice = os.environ.get("COSMON_SOFFICE_BIN") or "soffice"
out = tempfile.mkdtemp()
subprocess.run([soffice, "--headless", "--convert-to", "pdf", "--outdir", out, "deck.pptx"], check=True)
pdf = os.path.join(out, "deck.pdf")
doc = fitz.open(pdf)
paths = []
for i, page in enumerate(doc):
    p = os.path.join(out, f"slide-{i+1}.png")
    page.get_pixmap(dpi=150).save(p)
    paths.append(p)
print("\n".join(paths))