Create amazing adventures with AI magic! ✨
StoryForge is a modern, child-safe AI story generation platform that empowers young minds to create interactive adventures, mysteries, fantasies, and educational stories. Built with cutting-edge AI technology and designed with children's safety and creativity in mind.
- Interactive storytelling with branching narratives
- 200-300 word continuations for each choice made
- Unlimited story branches based on user decisions
- Real-time story progression with seamless UI
- Integrated character builder within story creation
- Multiple character roles: Protagonist, Sidekick, Mentor, Villain, Helper
- Character consistency maintained across story branches
- Visual character management with add/remove functionality
- 5 Story Types: Adventure, Mystery, Fantasy, Friendship, Educational
- 3 Age Groups: 5-8, 9-12, 13-16 years (age-appropriate content)
- 5 Tone Options: Exciting, Mysterious, Funny, Heartwarming, Educational
- Custom Word Limits: 100-3000 words with age-appropriate ranges
- Custom Instructions: Add educational facts, themes, or special requirements
- Multi-layer content filtering with balanced safety approach
- Age-appropriate vocabulary and themes
- COPPA compliant database design
- No inappropriate content generation
- Parental oversight capabilities
- Custom Fine-Tuned Model - Qwen2.5-0.5B specialized for children's stories
- Local AI processing with Ollama integration
- Advanced prompt engineering for consistent, quality output
- Rate limiting and connection pooling for stability
- LoRA fine-tuning on 3,685 children's stories for improved quality
- Material-UI design with responsive layout
- Intuitive story creation with accordion-based organization
- Real-time feedback and loading states
- Mobile-friendly design for all devices
- Accessibility features built-in
A quick visual tour of StoryForge.
| Landing Page | Build Characters | Example Story |
|---|---|---|
 |
 |
 |
| Generated Story (options) | Generated Story (next part) |
|---|---|
 |
 |
Images above:
- StoryForge.png — Landing page
- Characters.png — Build Characters page
- Story.png — Example story screen
- Story1.png — Generated story (3 choices + open option)
- Story2.png — Next part of story with more options
- Node.js 18+ and npm
- Ollama installed locally
- Git for version control
-
Clone the repository
git clone https://github.com/prateekvishnu/StoryForge.git cd StoryForge -
Install dependencies
npm install
-
Set up Ollama and load the custom model
# Install Ollama (if not already installed) curl -fsSL https://ollama.ai/install.sh | sh # Load the custom StoryForge model (already included in the repo) ollama create storyforge-qwen-fine-tuned -f training/models/Modelfile.storyforge
-
Initialize the database
npm run db:init
-
Start the development server
npm run dev
-
Open your browser Navigate to http://localhost:3000
- Visit the Create Page (
/create) - Enter your story prompt - describe the adventure you want to create
- Select age group and story type - choose appropriate settings
- Add characters (optional) - create protagonists, sidekicks, villains
- Customize settings - set tone, word limit, and special instructions
- Enable interactive mode for choose-your-adventure stories
- Click "Create Story" and watch the magic happen!
- Generate an interactive story with choice points enabled
- Read the initial story segment (200-300 words)
- Select from 3 choices (A, B, or C) to continue the adventure
- Each choice generates a new 200-300 word continuation
- Story branches infinitely based on your decisions
- Track your adventure with built-in story statistics
Visit /test-adventure to experience a pre-built choose-your-adventure story featuring Maya's Crystal Cave Adventure.
- Frontend: Next.js 14 with TypeScript
- UI Framework: Material-UI 5
- AI Engine: Ollama with custom StoryForge fine-tuned model
- Database: SQLite with better-sqlite3
- Styling: Material-UI theme system
- State Management: React hooks and context
StoryForge/
├── src/
│ ├── app/ # Next.js App Router
│ │ ├── api/ # API endpoints
│ │ │ ├── stories/ # Story generation APIs
│ │ │ └── test-db/ # Database testing
│ │ ├── create/ # Story creation page
│ │ ├── test-adventure/ # Interactive demo
│ │ └── globals.css # Global styles
│ ├── components/ # React components
│ │ ├── story/ # Story-related components
│ │ └── layout/ # Layout components
│ ├── lib/ # Utilities and services
│ │ ├── database/ # Database management
│ │ ├── ollama/ # AI client integration
│ │ └── middleware/ # API middleware
│ └── types/ # TypeScript definitions
├── memory-bank/ # Project documentation
├── scripts/ # Database and utility scripts
├── training/ # AI model training pipeline
│ ├── scripts/ # Training and preprocessing scripts
│ ├── models/ # Fine-tuned models and checkpoints
│ ├── training-datasets/ # Children's story datasets
│ └── logs/ # Training logs and metrics
└── data/ # SQLite database files
POST /api/stories/generate- Generate initial stories with choicesPOST /api/stories/continue- Continue stories based on user choicesGET /api/stories/generate- Health check for story generation serviceGET /api/test-db- Database connectivity test
Create a .env file in the project root:
# Database
DATABASE_URL="./data/storyforge.db"
# Ollama Configuration
OLLAMA_BASE_URL="http://localhost:11434"
OLLAMA_MODEL="storyforge-qwen-fine-tuned:latest"
# Optional: API Keys for external services
# ANTHROPIC_API_KEY=your_key_here
# OPENAI_API_KEY=your_key_hereThe application uses a custom fine-tuned model optimized for children's story generation:
- Model Name:
storyforge-qwen-fine-tuned:latest - Base Model: Qwen2.5-0.5B-Instruct (494M parameters)
- Specialization: Fine-tuned for children's stories on 3,685 curated stories
- Training Details:
- Training Steps: 924 steps across 4 epochs
- LoRA Configuration: r=16, alpha=32
- Training Loss: 5.69 → Optimized for story generation
- Training Date: August 8, 2025
- Model Size: 994 MB in Ollama
- Integration: Ready-to-use via Ollama with included Modelfile
StoryForge includes a complete training pipeline for fine-tuning language models specifically for children's story generation.
- 3,685 high-quality children's stories from classic fairy tale collections
- Sources: Andersen, Beatrix Potter, Brothers Grimm, Chinese Fairy Tales, and more
- Preprocessing: Stories formatted with age-appropriate prompts and safety guidelines
- Validation Split: 90% training, 10% validation
# 1. Preprocess the dataset
python training/scripts/preprocess_and_save.py
# 2. Fine-tune the model
python training/scripts/fine_tune_model.py
# 3. Manage and deploy models
python training/scripts/model_manager.py- Base Model: Qwen/Qwen2.5-0.5B-Instruct (494M parameters)
- Fine-tuning Method: LoRA (Low-Rank Adaptation)
- Training Steps: 924 steps across 4 epochs
- Batch Size: 8 with gradient accumulation
- Learning Rate: 3e-4 with warmup
- Hardware: Optimized for NVIDIA A100 GPU
- Training Loss: Reduced to 5.69 (optimized for story generation)
- Specialization: Enhanced understanding of children's story structure
- Safety: Improved age-appropriate content generation
- Speed: Ultra-fast inference on consumer hardware
The fine-tuned model is already integrated and ready to use:
- Model Name:
storyforge-qwen-fine-tuned:latest - Modelfile: Included at
training/models/Modelfile.storyforge - Size: 994 MB in Ollama
- Status: ✅ Successfully tested and deployed
npm run db:test# Test story generation API with custom model
curl -X POST http://localhost:3000/api/stories/generate \
-H "Content-Type: application/json" \
-d '{"prompt":"A brave explorer finds a treasure","ageGroup":"9-12","storyType":"adventure"}'Visit /test-adventure for a complete interactive demo.
- Pre-generation filtering of inappropriate prompts
- Post-generation analysis of story content
- Age-appropriate vocabulary enforcement
- Balanced safety approach that allows creative storytelling
- COPPA compliant data handling
- No personal information collection from children
- Parental oversight capabilities
- Safe content generation with multiple validation layers
Located in src/app/api/stories/generate/route.ts, templates can be customized for:
- Different age groups (5-8, 9-12, 13-16)
- Various story types (Adventure, Mystery, Fantasy, etc.)
- Tone adjustments (Exciting, Funny, Educational, etc.)
Material-UI theme configuration in src/app/ClientLayout.tsx:
- Color schemes
- Typography settings
- Component styling
- Responsive breakpoints
- Connection pooling for AI requests
- Rate limiting to prevent abuse
- Content caching for improved response times
- Efficient database queries with SQLite
- Responsive design for all devices
- Request tracking and logging
- Error handling with graceful fallbacks
- Performance metrics collection
- Health check endpoints
We welcome contributions! Please see our contributing guidelines:
- Fork the repository
- Create a feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
- Follow TypeScript best practices
- Use Material-UI components consistently
- Maintain child safety standards
- Add tests for new features
- Update documentation
This project is licensed under the MIT License - see the LICENSE file for details.
- Ollama for providing excellent local AI capabilities
- Material-UI for the beautiful component library
- Next.js team for the amazing React framework
- DeepSeek for the efficient language model
- Open source community for inspiration and tools
- Issues: GitHub Issues
- Discussions: GitHub Discussions
- Documentation: Check the
memory-bank/directory for detailed docs
Made with ❤️ for young storytellers everywhere
StoryForge - Where imagination meets technology! ✨