ToolStore Agent
feat: 4 new toolsets — docx-toolkit, pptx-toolkit, text-gen, batch-ops (15 functions total)
088cdb2
|
Raw
History Blame Contribute Delete
2.48 kB

pptx‑toolkit

Create, inspect, and extract content from PowerPoint (.pptx) presentations — slide‑level operations with code bindings for reading and generation.


When to Use This Toolset

  • Extracting text from slides for search or conversion
  • Inspecting presentation structure (slide count, layouts, dimensions)
  • Extracting speaker notes
  • Creating presentations from structured slide data
  • Generating pitch decks or report slides programmatically

Process

Extracting Content

Open → Inspect → Extract → Review
  1. Open with pptx_info to get slide count and layouts
  2. Inspect specific slides with pptx_read
  3. Extract content — text, tables, notes
  4. Review the extraction for completeness

Creating a Presentation

Plan → Build → Format → Save
  1. Plan the slide structure (title slide + content slides)
  2. Build with pptx_create using structured data
  3. Format is auto‑handled — title slide uses layout 0, content uses layout 1
  4. Save to the target path

Function Reference

pptx_read

Extract all text slide‑by‑slide, including speaker notes.

When to use: Content extraction, slide indexing, note retrieval.

Args: filepath (str)

Returns: {slides: [{index, title, body_text, notes, shape_count}], count}

pptx_info

Quick structural inspection.

When to use: Before editing — check layouts and dimensions.

Args: filepath (str)

Returns: {slides, slide_width, slide_height, layouts, filename}

pptx_create

Create a new presentation from structured data.

When to use: Report generation, pitch deck creation, automated slide output.

Args: filepath (str), title (str), slides_data (list)

Slides data format: [{title, bullets: [...]}, ...] — each dict becomes a slide.


Guidelines

Do

  • Use pptx_info first to understand the presentation structure
  • Extract speaker notes with pptx_read — they often contain key context
  • Verify output by reading back with pptx_read

Don't

  • Don't assume placeholder indices are consistent across templates
  • Don't extract charts as text — they render as images
  • Don't overwrite without confirming

Limitations

  • Charts and media: Text‑only extraction — charts, images not readable
  • Custom layouts: pptx_create uses standard layouts 0 and 1
  • Animations: Not preserved in programmatic creation