Story-Story is an engaging storytelling app designed for children. It takes the rich histories of Nigerian and African figures and turns them into child-friendly, factual stories. With vibrant illustrations and captivating narratives, Story-Story aims to make history come alive for young minds. Stories are generated using an LLM agentic workflow, see demo here.

Story-Story uses an ensemble of AI LLM agents implemented with LangGraph to generate entire 10-page stories about a hero from a single input (the hero's name).

Below is the architecture

1. Planner Agent
   • Responsible for generating the outline / chapter titles for the story
   • input: historical_figure
   • output: chapter_title
   • model: GPT-4o
2. Storyteller Agent (GPT-4o)
   • Responsible for generating accurate content for the chapter
   • input: historical_figure chapter_title
   • output: chapter_content
   • model: GPT-4o
3. Illustrator Agent
   • Responsible for generating prompts that will be used to generate images
   • input: chapter_title historical_figure
   • output: prompt
   • model: GPT-4o
4. Image Creator
   • Responsible for generating the illustrations given a prompt
   • input: historical_figure prompt
   • outputs: image
   • model: Flux.1 Schnell
5. Publisher Agent
   • Responsible for verifying validity of outputs, making sure it's okay for child consumption
   • inputs: historical_figure chapter_content image
   • outputs: score
   • model: GPT-4v

• Install requirements

pip install -r requirements.txt

• Add OpenAI API Key to environment

export OPENAI_API_KEY="sk-xxxxxxxxxxxxxxxx"  

• Add Supabase credentials to environment

export SUPABASE_URL="https://your-project-url.supabase.co"
export SUPABASE_KEY="your-supabase-key"

• Add Hugging Face token to environment

export HF_TOKEN="hf_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"

python main.py --historical_figure "Ahmadu Bello"

This will create the 10-page content including illustrations about Ahmadu Bello and upload to a Supabase database

The app uses Supabase for storing story-related data. Here's a brief schema overview:

1. Stories:

• id: Primary Key
• title: Title of the story
• historical_figure: Name of the figure
• description: Brief description
• created_at: Timestamp

2. StoryPages:

• id: Primary Key
• story_id: Foreign Key (References Stories)
• page_number: Page number in the story
• chapter: Chapter content
• illustration_url: URL for the illustration
• prompt: Prompt used to generate illustrations

Contributions are highly welcome!

Here are ideas I am currently working on

• Illustrations
• Add reference store so storyteller agent can perform RAG
• Narrator