pptx

name	pptx
description	Extract text and tables from PowerPoint (.pptx) files. Use when a user uploads a PPT or PPTX document that needs to be processed.

PowerPoint Processing for Text Extraction

Purpose

Extract readable text and structured tables from PPTX files so the content can be summarized and turned into a new PPT.

When to Use

• User uploads a .pptx file
• User uploads a legacy .ppt file that has already been converted or can be converted to .pptx
• User references an existing PPT/PPTX generated earlier in the same conversation
• Need to extract presentation content for summarization, analysis, regeneration, speaker notes, or scripts

Approach

Step 1: Try python-pptx First

from pptx import Presentation

prs = Presentation("presentation.pptx")
text_parts = []
tables = []

for slide in prs.slides:
    for shape in slide.shapes:
        if shape.has_text_frame:
            for paragraph in shape.text_frame.paragraphs:
                line = paragraph.text.strip()
                if line:
                    text_parts.append(line)

        if shape.has_table:
            table_data = []
            for row in shape.table.rows:
                cells = [cell.text.strip() for cell in row.cells]
                table_data.append(cells)
            if table_data:
                tables.append(table_data)

Step 2: Fallback to markitdown

python -m markitdown presentation.pptx

Use this when python-pptx is unavailable or extraction quality is better with markdown output.

Step 3: Handle Legacy .ppt

If the original file is .ppt, convert it first:

soffice --headless --convert-to pptx file.ppt

Then process the generated .pptx.

Handling Common Issues

Problem	Solution
Legacy .ppt file	Convert to .pptx first
python-pptx missing	Use markitdown fallback
Empty extraction	Try markdown conversion or inspect whether slides are image-heavy
Tables not found	Check shape.has_table; some "tables" are actually pasted images

Integration Notes

For the upload-to-PPT workflow:

1. Extract slide text and tables from the uploaded PPT/PPTX
2. Preserve slide count as a page estimate
3. Pass the extracted content to document-summarizer
4. Then call generate_ppt(content=...)

General follow-up rule for an existing deck:

• Treat the uploaded PPT as source material first.
• If the user wants any derived text output, analysis, notes, translation, or explanation, stop after extraction and answer that request directly.
• Only call generate_ppt when the requested final artifact is a PPT and the user explicitly wants a new deck or slide revisions/regeneration.

Output Format

Return (raw_text: str, tables: list[list[list[str]]], slide_count: int):

• raw_text: text extracted from all slides
• tables: tables extracted from slide shapes
• slide_count: total slide count