Ersatz Echos

Ersatz Echos is an AI-powered history generator that creates rich, structured timelines for fictional worlds. With customizable parameters and the ability to draw inspiration from user-provided documents, Ersatz Echos crafts unique histories spanning centuries or millennia. The tool outputs its generated histories in a structured JSON format, making it easy to integrate with other world-building applications. Whether you're a game developer, novelist, or simply a creative enthusiast, Ersatz Echos provides a foundation for building immersive, consistent fictional universes.

Features

• AI-driven history generation using advanced language models
• Customizable timeline parameters (start year, end year, number of events, etc..)
• Structured JSON output for easy integration with other tools
• Experimental features for enhanced world-building capabilities
  • Ability to extract information from user-provided PDF documents for added context

Getting Started

Prerequisites

• Access to an API (OpenAI, OpenRouter, etc.)
• Python 3.11+ installed

Installation

1. Clone the repository:

   git clone https://github.com/DaemonIB/ersatz-echos.git
   

2. Navigate to the project directory:

   cd ersatz-echos
   

3. Install the required dependencies:

   pip install -r requirements.txt
   

4. Copy the example configuration files:

   cp user_context.example.json user_context.json
   cp config.example.json config.json
   

5. Edit config.json and replace YOUR_API_KEY_HERE with your LLM API key.

6. (Optional) If you want to provide additional context for the LLM to base the timeline on:

   • Edit user_context.json and replace the sample data with your own data.

Usage

Run the Python script:

python main.py --output history.json

The generated history will be saved to the file specified with the --output flag, or history.json if not specified.

Configuration

• config.json: Contains settings for the LLM API and other parameters.
• user_context.json: Allows you to provide additional context for the LLM to base the timeline on. The structure is generic, and any top-level category can be used (e.g., Locations, Characters, Funny Hair Styles). Individual entities within each category need a name and description.

Notes

• The default model has been selected based on cost and ability to produce JSON. For OpenAI, openai/gpt-3.5-turbo-0125 is compatible, along with the models specified here.
• Other models may work, but their success rate may be lower.

Experimental Features

• Document Extraction: Enabled via the document_extraction flag in config.json.

License

This project is licensed under Apache-2.0.