c-base LLM Project

Welcome to the c-base LLM (Large Language Model) project! This initiative aims to integrate a locally-hosted, open-source AI system into the c-base environment, enhancing its functionality as a partially reactivated spaceship computer (c-beam). Below are the key requirements and objectives of the project.

Key Objectives

• Implement a self-hosted LLM tailored to the c-base history.
• Enable multi-user and multi-instance interactions for different rooms (e.g., Workshop vs. Nerd Area).
• Develop a sandbox environment for experimentation and development.

Technical Requirements

AI Model

• Feasibility: An LLM that can be fine-tuned with minimal effort to integrate c-base history.
• Efficiency: Smaller models preferred, optimized for local hardware.
• Compatibility: Support for LLM-compatible agent systems and modular design.

Voice Integration

• Voice-to-Text: Evaluate and integrate suitable models, such as Whisper, for converting voice commands to text.
• Input Methods: Support for sending voice messages via platforms like Matrix.
• Privacy-Conscious: Integrate push-to-talk microphones for user privacy somewhere at c-base location.

Multi-User and Multi-Instance Support

• Ensure simultaneous responses for different rooms.
• Focus on sufficient VRAM capacity for handling multiple interactions.

Hardware Requirements

• Budget Options: Feasible with used hardware (~€1000) using LLaMa or LoRA models.
• High-End Options: Explore NVIDIA Project Digits (~$3000) for cutting-edge performance.
• Future-Proofing: Consider waiting for the second generation of such hardware for better pricing and stability.

Audio and Communication System

• Revamp the c-base ELA system for room-specific announcements.
• Equip c-beam with spatial audio features.

Community Involvement

• Engage key members (e.g., "lazee" and "sasquatsch") with expertise in self-hosted AI technologies.
• Secure volunteers with coding skills for long-term support and development.

Financial Considerations

• Medium- to high-budget projects are possible, funded through c-base reserves.
• Ensure the majority of the c-base crew supports the project.

Next Steps

1. Collect and refine additional requirements.
2. Evaluate potential models and hardware solutions.
3. Develop a prototype in a sandbox environment.
4. Gradually roll out the solution across c-base.

Future

1. Connection to c-beam controlling of c-base lights etc. via c-beam app

Tech.Doc: https://code.c-base.org/kind