Create Illustrated Storybooks Instantly with Crew AI Agents! (Groq)

---

---

AI Summary

AI-Generated Storybook Creation Summary

Overview

• AI agents can now be used to generate storybooks with illustrations and convert them to PDF.
• The process involves creating agents, tasks, custom tools for image generation and PDF conversion, and making them work together.

Steps to Create a Storybook

1. Set Up Environment
   • Install necessary packages with pip.
   • Export API keys for Gro and OpenAI.
2. Code Preparation
   • Import required packages and define tools and agents in app.py.
   • Create three tools: file reader, image generator, and PDF converter.
   • Define five agents: story outliner, writer, image generator, content formatter, and PDF converter.
3. Task Assignment
   • Assign tasks to agents for outlining, writing, image generation, content formatting, and PDF conversion.
4. Execution
   • Define a crew with agents and tasks, set to process sequentially.
   • Run the code to generate the storybook.

Tools and Agents

• Tools: File reader, image generator (using DALL-E 3), and markdown to PDF converter.
• Agents: Story outliner, story writer, image generator, content formatter, and markdown to PDF converter.

Process

1. Story Outlining
   • Generate a document outline with chapter titles, character descriptions, and main plot points.
2. Story Writing
   • Write the complete storybook with specified chapters and word count.
3. Image Generation
   • Create one image per chapter.
4. Content Formatting
   • Merge the written story and generated images into a markdown file.
5. PDF Conversion
   • Convert the markdown file into a PDF.

Results

• The storybook is generated quickly using Gro, with the potential for even faster generation if Gro implements image generation.
• The same process is tested with OpenAI for comparison.

Additional Information

• The author encourages subscribing to their YouTube channel for more AI-related content.
• The process can be fine-tuned by adjusting the prompts in the script.
• The entire book creation takes approximately 1 minute, potentially less with faster image generahe agents

CrewAI Groq Create Story Books

conda create -n crewai python=3.11  
conda activate crewai  
pip install -U 'crewai[tools]' mdpdf  
export OPENAI_API_KEY=xxxxxxxxxx  
export GROQ_API_KEY=xxxxxxxxxxxx  

from crewai import Agent, Task, Crew, Process  
from crewai_tools import tool  
from langchain_openai import ChatOpenAI  
from crewai_tools.tools import FileReadTool  
import os, requests, re, mdpdf, subprocess  
from openai import OpenAI  
  
llm = ChatOpenAI(  
    openai_api_base="https://api.groq.com/openai/v1", # https://api.openai.com/v1 or https://api.groq.com/openai/v1   
    openai_api_key=os.getenv("GROQ_API_KEY"), # os.getenv("OPENAI_API_KEY") or os.getenv("GROQ_API_KEY")  
    model_name="mixtral-8x7b-32768" #  gpt-4-turbo-preview or mixtral-8x7b-32768   
)  
  
file_read_tool = FileReadTool(  
	file_path='template.md',  
	description='A tool to read the Story Template file and understand the expected output format.'  
)  
  
@tool  
def generateimage(chapter_content_and_character_details: str) -> str:  
    """  
    Generates an image for a given chapter number, chapter content, detailed location details and character details.  
    Using the OpenAI image generation API,  
    saves it in the current folder, and returns the image path.  
    """  
    client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))  
  
    response = client.images.generate(  
        model="dall-e-3",  
        prompt=f"Image is about: {chapter_content_and_character_details}. Style: Illustration. Create an illustration incorporating a vivid palette with an emphasis on shades of azure and emerald, augmented by splashes of gold for contrast and visual interest. The style should evoke the intricate detail and whimsy of early 20th-century storybook illustrations, blending realism with fantastical elements to create a sense of wonder and enchantment. The composition should be rich in texture, with a soft, luminous lighting that enhances the magical atmosphere. Attention to the interplay of light and shadow will add depth and dimensionality, inviting the viewer to delve into the scene. DON'T include ANY text in this image. DON'T include colour palettes in this image.",  
        size="1024x1024",  
        quality="standard",  
        n=1,  
    )  
    image_url = response.data[0].url  
    words = chapter_content_and_character_details.split()[:5]   
    safe_words = [re.sub(r'[^a-zA-Z0-9_]', '', word) for word in words]    
    filename = "_".join(safe_words).lower() + ".png"  
    filepath = os.path.join(os.getcwd(), filename)  
  
    # Download the image from the URL  
    image_response = requests.get(image_url)  
    if image_response.status_code == 200:  
        with open(filepath, 'wb') as file:  
            file.write(image_response.content)  
    else:  
        print("Failed to download the image.")  
        return ""  
  
    return filepath  
  
@tool  
def convermarkdowntopdf(markdownfile_name: str) -> str:  
    """  
    Converts a Markdown file to a PDF document using the mdpdf command line application.  
  
    Args:  
        markdownfile_name (str): Path to the input Markdown file.  
  
    Returns:  
        str: Path to the generated PDF file.  
    """  
    output_file = os.path.splitext(markdownfile_name)[0] + '.pdf'  
      
    # Command to convert markdown to PDF using mdpdf  
    cmd = ['mdpdf', '--output', output_file, markdownfile_name]  
      
    # Execute the command  
    subprocess.run(cmd, check=True)  
      
    return output_file  
  
story_outliner = Agent(  
  role='Story Outliner',  
  goal='Develop an outline for a children\'s storybook about Animals, including chapter titles and characters for 5 chapters.',  
  backstory="An imaginative creator who lays the foundation of captivating stories for children.",  
  verbose=True,  
  llm=llm,  
  allow_delegation=False  
)  
  
story_writer = Agent(  
  role='Story Writer',  
  goal='Write the full content of the story for all 5 chapters, each chapter 100 words, weaving together the narratives and characters outlined.',  
  backstory="A talented storyteller who brings to life the world and characters outlined, crafting engaging and imaginative tales for children.",  
  verbose=True,  
  llm=llm,  
  allow_delegation=False  
)  
  
image_generator = Agent(  
  role='Image Generator',  
  goal='Generate one image per chapter content provided by the story outliner. Start with Chapter number, chapter content, character details, detailed location information and detailed items in the location where the activity happens. Generate totally 5 images one by one. Final output should contain all the 5 images in json format.',  
  backstory="A creative AI specialized in visual storytelling, bringing each chapter to life through imaginative imagery.",  
  verbose=True,  
  llm=llm,  
  tools=[generateimage],  
  allow_delegation=False  
)  
  
content_formatter = Agent(  
    role='Content Formatter',  
    goal='Format the written story content in markdown, including images at the beginning of each chapter.',  
    backstory='A meticulous formatter who enhances the readability and presentation of the storybook.',  
    verbose=True,  
    llm=llm,  
    tools=[file_read_tool],  
    allow_delegation=False  
)  
  
markdown_to_pdf_creator = Agent(  
    role='PDF Converter',  
    goal='Convert the Markdown file to a PDF document. story.md is the markdown file name.',  
    backstory='An efficient converter that transforms Markdown files into professionally formatted PDF documents.',  
    verbose=True,  
    llm=llm,  
    tools=[convermarkdowntopdf],  
    allow_delegation=False  
)  
  
  
# Create tasks for the agents  
task_outline = Task(  
    description='Create an outline for the children\'s storybook about Animals, detailing chapter titles and character descriptions for 5 chapters.',  
    agent=story_outliner,  
    expected_output='A structured outline document containing 5 chapter titles, with detailed character descriptions and the main plot points for each chapter.'  
)  
  
task_write = Task(  
    description='Using the outline provided, write the full story content for all chapters, ensuring a cohesive and engaging narrative for children. Each Chapter 100 words. Include Title of the story at the top.',  
    agent=story_writer,  
    expected_output='A complete manuscript of the children\'s storybook about Animals with 5 chapters. Each chapter should contain approximately 100 words, following the provided outline and integrating the characters and plot points into a cohesive narrative.'  
)  
  
task_image_generate = Task(  
    description='Generate 5 images that captures the essence of the children\'s storybook about Animals, aligning with the themes, characters, and narrative outlined for the chapters. Do it one by one.',  
    agent=image_generator,  
    expected_output='A digital image file that visually represents the overarching theme of the children\'s storybook, incorporating elements from the characters and plot as described in the outline. The image should be suitable for inclusion in the storybook as an illustration.',  
)  
  
task_format_content = Task(  
    description='Format the story content in markdown, including an image at the beginning of each chapter.',  
    agent=content_formatter,  
    expected_output='The entire storybook content formatted in markdown, with each chapter title followed by the corresponding image and the chapter content.',  
    context=[task_write, task_image_generate],  
    output_file="story.md"  
)  
  
task_markdown_to_pdf = Task(  
    description='Convert a Markdown file to a PDF document, ensuring the preservation of formatting, structure, and embedded images using the mdpdf library.',  
    agent=markdown_to_pdf_creator,  
    expected_output='A PDF file generated from the Markdown input, accurately reflecting the content with proper formatting. The PDF should be ready for sharing or printing.'  
)  
  
crew = Crew(  
  agents=[story_outliner, story_writer, image_generator, content_formatter, markdown_to_pdf_creator],  
  tasks=[task_outline, task_write, task_image_generate, task_format_content, task_markdown_to_pdf],  
  verbose=True,  
  process=Process.sequential  
)  
  
result = crew.kickoff()  
  
print(result)  

Ollama

llm = ChatOpenAI(  
    openai_api_base="http://localhost:11434/v1", # https://api.openai.com/v1 or https://api.groq.com/openai/v1   
    openai_api_key="Null", # os.getenv("OPENAI_API_KEY") or os.getenv("GROQ_API_KEY")  
    model_name="mistral" #  gpt-4-turbo-preview or mixtral-8x7b-32768  
)  
  
  
  
---