Forge Mind

A JavaFX desktop application for local LLM chats, with integrated Google Drive synchronization for chat history and a powerful web search feature.

---

About the Project

Forge Mind is a desktop application designed to provide a secure and private AI chat experience. Unlike web-based solutions, this app runs large language models (LLMs) locally on your machine, ensuring your data never leaves your computer. With a sleek user interface built with JavaFX, it seamlessly integrates with Google Drive to back up your conversations, offering both privacy and convenience.

Code Name: Project Plasma

Key Features

• Local LLM Integration: Runs popular LLM models like Phi-2 2.7B GGUF or Mistral-7B GGUF directly on your machine using the llama.cpp C++ library via JNI.
• One-Time Local Downloads: Automatically downloads the required LLM models the first time the app is run.
• Google Drive Synchronization: Securely authenticates with Google Drive to store and retrieve your chat history, making your data accessible and backed up.
• Flexible Chat Modes: Offers three distinct modes to suit your needs:
  • Normal Chat: Saved and synced to Google Drive.
  • Temporary Chat: Private conversations that are not saved.
  • Web Search Chat: Integrates with a search API to provide up-to-date, factual information.
• Modern UI: A clean and intuitive user interface built with the JavaFX framework.

Technologies Used

• Front-End: JavaFX
• Back-End: Java
• AI Engine: llama.cpp (via JNI)
• Cloud Storage: Google Drive API for Java

Getting Started

To get a local copy up and running, follow these simple steps:

1. Clone the repository:

   git clone https://github.com/tux-byte-club/ForgeMind.git

2. Open the project in your favorite Java IDE (e.g., IntelliJ IDEA, Eclipse).
3. Configure JNI to work with your local llama.cpp build.
4. Run the application. The LLM models will be downloaded automatically on the first run.

Roadmap

Here's the plan for getting Forge Mind up and running. This checklist will be updated as we make progress!

• Phase 1: Core Functionality

  • Set up basic JavaFX UI with a chat window.
  • Implement Java Native Interface (JNI) to call a llama.cpp function.
  • Successfully load and run a simple LLM model.
  • Implement basic chat functionality (send input, receive output).

• Phase 2: Data & Storage

  • Implement Google Drive API integration.
  • Handle OAuth 2.0 authentication for the user's Google account.
  • Create methods to save and load chat history from Drive.
  • Create a local cache for chat history.

• Phase 3: Features

  • Implement the "Normal Chat" mode with Google Drive saving.
  • Implement the "Temporary Chat" mode that doesn't save history.
  • Implement the "Web Search" functionality and feed results to the LLM.
  • Add a visual download progress bar for the first-time model download.

• Phase 4: Refinement & UX

  • Improve the chat UI with timestamps and user/AI labels.
  • Add error handling for network issues, API errors, etc.
  • Optimize model loading and performance.
  • Prepare for release!

---

System Requirements

To ensure a smooth and efficient experience, your system should meet the following minimum requirements. Performance will scale directly with your hardware, especially your RAM and CPU.

• Operating System: Windows 10/11, macOS, or a modern Linux distribution.
• RAM:
  • Minimum: 8 GB is required for the application and the smaller LLM model (e.g., Phi-2 2.7B).
  • Recommended: 16 GB or more is highly recommended for running larger models (e.g., Mistral-7B) without performance issues.
• Processor (CPU): A modern multi-core processor is essential, as llama.cpp is heavily optimized for CPU usage.
• Storage: At least 10-15 GB of free disk space is needed to download and store the application and the LLM model files.
• Java Runtime: Java 17 or a more recent version of the JRE/JDK is required to run the application.

---

License

This project is released under the MIT License.

---