Skip to main content
This notebook shows you how to automatically generate and assemble professional PowerPoint decks using CAMEL-AI’s PPTXToolkit. You’ll learn how to:
  • Prompt an LLM to produce fully structured JSON for every slide
  • Turn that JSON into a polished .pptx with titles, bullets, step diagrams, tables, and images
  • Leverage Markdown styling (**bold**, *italic*) and Pexels image search via img_keywords
  • Plug in your own .pptx templates (modern, boardroom, minimalist, etc.)
  • Enjoy auto-layout selection for text, diagrams, and tables

​
🚥 Pipeline Overview

  1. Single Agent: Content → JSON
    • You send one prompt to the LLM
    • It returns a JSON list with:
      • A title slide (title, subtitle)
      • At least one step-by-step slide (all bullets start with >>)
      • At least one table slide (table: {headers, rows})
      • Two or more slides with meaningful img_keywords
      • All bullet slides using Markdown formatting
  2. PPTXToolkit: JSON → .pptx
    • Pass that JSON into PPTXToolkit.create_presentation(...)
    • Renders slides with your chosen template, images via img_keywords, chevrons/pentagons, and tables
    • Outputs a ready-to-share PowerPoint file
Ready to build your next deck? Let’s get started! 🎉 You can also check this cookbook in colab here 
%pip install "camel-ai[all]==0.2.66"

​
⚙️ Configuration

Set your API keys securely: 
from getpass import getpass
import os

# Prompt the user securely (input is hidden)
os.environ["OPENAI_API_KEY"] = getpass("🔑 Enter your OpenAI API key: ")
os.environ["PEXELS_API_KEY"] = getpass("🔑 Enter your Pexels API key (leave blank if not using images): ")

​
✏️ Agent #1: Generate Structured JSON

One single LLM call to produce exactly the JSON your toolkit expects. 
import os, json
from camel.agents import ChatAgent
from camel.models import ModelFactory
from camel.types import ModelPlatformType, ModelType

def generate_slides_json(topic: str, slide_count: int) -> list:
    # Build our strict prompt
    prompt = f"""
You are a world-class PowerPoint generator for CAMEL-AI's PPTXToolkit.

Create a presentation about "{topic}" with exactly {slide_count + 1} slides (1 title + {slide_count} content).
MANDATORY:
1. First slide: title & subtitle.
2. ≥1 step-by-step slide (all bullets start with ">>").
3. ≥1 table slide (with headers & rows).
4. ≥2 slides with non-empty img_keywords (search terms only).
5. Bullet slides use Markdown (**bold**, *italic*) in bullet_points.
6. Output MUST BE raw JSON (list of dicts), no markdown fences or commentary.

Example:
[
  {{"title":"...","subtitle":"..."}},
  {{"heading":"...","bullet_points":["...","..."],"img_keywords":"..."}},
  {{"heading":"...","bullet_points":[">> Step 1..."],"img_keywords":"..."}},
  {{"heading":"...","table":{{"headers":["A","B"],"rows":[["1","2"],["3","4"]]}},"img_keywords":"..."}}
]
"""
    agent = ChatAgent(
        system_message="You strictly output valid JSON for PowerPoint slides.",
        message_window_size=5,
        model=ModelFactory.create(
            model_platform=ModelPlatformType.OPENAI,
            model_type=ModelType.GPT_4_1,
            model_config_dict={"temperature": 0.0},
        )
    )
    resp = agent.step(prompt)
    slides = json.loads(resp.msgs[0].content)
    return slides

​
📦 Build & Save PPTX

Local code uses PPTXToolkit to turn that JSON into a .pptx file: 
from camel.toolkits.pptx_toolkit import PPTXToolkit

def build_pptx(slides_json: list, out_name: str = "presentation.pptx") -> str:
    kit = PPTXToolkit(output_dir=".")
    kit.create_presentation(json.dumps(slides_json), out_name)
    return out_name

​
🔗 Run the Pipeline

if __name__ == "__main__":
    topic = input("🎯 Topic: ")
    n     = int(input("📄 # of content slides: "))
    slides = generate_slides_json(topic, n)
    filename = f"{topic.replace(' ','_')}.pptx"
    path = build_pptx(slides, out_name=filename)
    print(f"✅ Deck saved to: {path}")