Project_README.md

StoryCraftAI

Learn how to automate story generation using deep learning and deploy a full-featured web application with CI/CD.

Project Description

StoryCraftAI is an automatic story generation system built entirely from scratch using deep learning. This project covers the entire pipeline—from scraping and preprocessing classic short stories to training various neural network architectures and generating creative narratives from a user-provided seed text.

Key Features

• Data Scraping & Preprocessing:
  • Automatically scrapes classic short stories from classicshorts.com.
  • Cleans text by removing unwanted characters and tokenizes sentences into words.
• Multiple Model Architectures:
  • GRU & LSTM: Optimized for sequence prediction.
  • Bidirectional Models: (Bidirectional-LSTM and Bidirectional-GRU) Capture richer context.
  • Hybrid Approaches: Combine strengths of different models for better performance.
• Training & Evaluation:
  • Generates training sequences (50 words as input and 1 as output).
  • Tracks performance with accuracy and loss metrics.
  • Saves training history for visualization.
• Interactive Web Application:
  • Flask-based web interface for real-time story generation.
• CI/CD & Deployment:
  • Deployed on Hugging Face Spaces with continuous integration via GitHub Actions.

Getting Started

Prerequisites

• Python 3.8 or later.
• Virtual environment (recommended).

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/StoryCraftAI.git
   cd StoryCraftAI
   

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate   # On Windows: venv\Scripts\activate
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Download NLTK data:

   python -c "import nltk; nltk.download('punkt')"
   

5. Run the application locally:

   python app.py
   

Training the Models

Pipeline Overview

• Data Scraping: Fetches stories from classicshorts.com.
• Preprocessing: Tokenizes text and creates sequences of 51 tokens (50 input, 1 target).
• Model Building: Supports architectures like gru, lstm, bi_di_gru, and bi_di_lstm.
• Training: Uses categorical crossentropy loss. Training logs and performance graphs are generated.

Execution

Run the training scripts using Jupyter notebooks in the Notebooks/ folder or directly from the command line. Save the trained model, tokenizer, and training history in the Models/ directory.

Deployment

Using Docker

1. Build and run the Docker container locally:

   docker build -t storycraftai .
   docker run -p 7860:7860 storycraftai
   

CI/CD on Hugging Face Spaces

• Configured for CI/CD using GitHub Actions. Every push to the main branch triggers an automated rebuild and redeployment.

Contributing

Contributions are welcome! Please:

1. Fork the repository.
2. Create a new branch.
3. Submit a pull request detailing your modifications.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

For questions or feedback:

• Email: imamitsingh958@gmail.com
• GitHub: @amitsingh-24
• Linkedin: @Linkedin

Enjoy creating and exploring creative stories with StoryCraftAI!