MythForge

MythForge is a local chat server built on top of FastAPI and llama_cpp. It exposes a simple web UI and a JSON API for interactive story generation and experimentation built with FASTAPI with llama models.

Features

• Web interface served from the ui/ directory
• Character-based conversations with persistent memory
• Automatic goals created and updated throughout chat sessions
• System prompts and goal oriented prompts
• Clean text-based UI (HTML/JS/CSS) designed for PC and IPad usage
• Configurable model launch parameters via model_settings.json

Screenshots

Directory overview

• mythforge/ – Python backend implementation
• ui/ – Static files for the browser UI
• models/ – place your .gguf model files here
• chats/ – per-chat history storage
• global_prompts/ – system prompt files
• server_logs/ – JSON event logs

Requirements

MythForge requires Python 3.10 or later. Install dependencies using pip:

pip install fastapi uvicorn llama-cpp-python pydantic

Additional packages may be needed depending on your configuration.

Running

1. Copy or symlink your llama model into the models/ directory.

2. Launch the server with "RunMythForge.bat" python -m uvicorn mythforge.main:app --host 0.0.0.0 --port 8000

3. Connect with http:YourLocalIP:8000/MythForgeUI.html in your browser to access the UI.

API Endpoints

MythForge exposes a JSON API for programmatic access. Examples:

• POST /chat/send – Send a message to the assistant
• GET /settings/ – Retrieve current model settings
• PUT /settings/ – Update generation parameters
• GET /prompts/ – List all global prompts

Customization

Model parameters can be changed by editing model_settings.json. Global prompts can be added under global_prompts/ and will be available in the UI. Logs are stored under server_logs/.