ResuWrite

Overview

ResuWrite is an AI-powered novel generation system that transforms a person's resume into a creative story. Using a series of specialized LLM chains, the system analyzes a resume, creates a character profile, and then generates a complete novel with chapters, plots, and events tailored to that character.

Architecture

graph TD
    subgraph Input
        resume[Resume PDF]
        subject[Subject]
        genre[Genre]
        author[Author]
    end

    subgraph Chains
        MC[MainCharacterChain]
        TC[TitleChain]
        PC[PlotChain]
        CC[ChaptersChain]
        WC[WriterChain]
    end

    subgraph Output
        profile[Character Profile]
        title[Novel Title]
        plot[Full Plot]
        chapters[Chapters List]
        book[Complete Novel]
        doc[Word Document]
    end

    resume --> MC
    MC --> profile
    
    profile --> TC
    subject --> TC
    genre --> TC
    author --> TC
    TC --> title
    
    profile --> PC
    subject --> PC
    genre --> PC
    author --> PC
    title --> PC
    PC --> plot
    
    profile --> CC
    subject --> CC
    genre --> CC
    author --> CC
    title --> CC
    plot --> CC
    CC --> chapters
    
    profile --> WC
    genre --> WC
    author --> WC
    title --> WC
    plot --> WC
    chapters --> WC
    WC --> book
    
    book --> doc
    chapters --> doc
    title --> doc

Loading

Components

1. LLM Provider

ResuWrite uses Groq as the LLM provider, leveraging powerful language models like llama-3.3-70b-versatile or mixtral-8x7b-32768 through the LangChain integration.

2. Chains

The system is composed of five specialized LLM chains:

MainCharacterChain

• Purpose: Analyzes a resume and creates a character profile
• Input: PDF resume
• Output: Textual description of the person's profile
• Process: Loads and splits PDF, combines content, and runs through LLM

TitleChain

• Purpose: Generates an appropriate title for the novel
• Input: Subject, genre, author, and character profile
• Output: Novel title
• Process: Considers all inputs to create a genre and author-appropriate title

PlotChain

• Purpose: Creates a complete plot for the novel
• Input: Subject, genre, author, profile, and title
• Output: Detailed plot description
• Process: Uses a helper chain to generate storytelling attributes, then creates a full plot

ChaptersChain

• Purpose: Generates a structured chapter list
• Input: Subject, genre, author, profile, title, and plot
• Output: Dictionary of chapters and their descriptions
• Process: Creates a prologue, numbered chapters, and an epilogue

WriterChain

• Purpose: Writes the actual content for each chapter event
• Input: Genre, author, title, profile, plot, previous events, chapter summary, and current event
• Output: Paragraphs of story content
• Process: Maintains continuity while generating new content for each event

3. Document Processing

• PDF Loading: Uses PyPDFLoader to extract text from resumes
• Word Document Creation: Uses python-docx to format and save the final novel

Installation

1. Clone the repository
2. Install dependencies:

   pip install python-dotenv langchain-groq langchain pypdf2 python-docx

3. Create a .env file with your Groq API key:

   GROQ_API_KEY=your_api_key_here
   

Usage

1. Place a resume PDF in the ./docs folder
2. Edit the main() function in app.py to set your desired:
   • Subject
   • Author name
   • Genre
   • Resume file name
3. Run the application:

   python app.py

4. The generated novel will be saved as a Word document in the ./docs folder

File Structure

.
├── app.py             # Main application code
├── .env               # Environment variables (API keys)
├── docs/              # Input and output documents
│   ├── Divija_Prof.pdf  # Example resume
│   └── Generated_Novel.docx  # Output file
├── README.md          # This file
└── requirements.txt   # Dependencies

Workflow

1. Resume Analysis: The system begins by loading and analyzing a resume PDF
2. Character Creation: A profile is generated based on the resume content
3. Story Framework: Title, plot, and chapter structure are created
4. Content Generation: Each chapter is populated with events and paragraphs
5. Document Creation: The final novel is formatted and saved as a Word document

Customization

You can customize the system by:

• Modifying the prompt templates in each chain
• Adjusting the LLM parameters (temperature, model)
• Adding additional chains for more complex story elements
• Expanding the event generation for more detailed chapters

Limitations

• The quality of the generated novel depends on the resume content and LLM capabilities
• Long novels may require multiple API calls and can be time-consuming
• The current implementation uses a simplified approach to event generation

Future Enhancements

• Implement more sophisticated event planning
• Add character relationship mapping
• Include theme and motif analysis
• Provide style customization options
• Create a web interface for easier use

Dependencies

• langchain-groq
• langchain
• python-dotenv
• PyPDFLoader
• python-docx

License

This project is licensed under the MIT License.