⚠️ Preview Release

Razer AIKit is in preview. Features may change or have limitations before stable release. We appreciate your feedback!

Open-source AI development environment for engineers and researchers. Delivers cloud-grade performance and scalability directly on your desktop with out-of-the-box setup.

• AIKit CLI - Command-line interface for model lifecycle management
• vLLM Engine - Production-grade LLM inference with memory optimization
• LlamaFactory - Parameter-efficient fine-tuning framework
• Ray - Distributed computing for seamless multi-GPU scaling

nvidia-smi

nvidia-smi

# Create huggingface cache directory if it doesn't exist
mkdir -p $HOME/.cache/huggingface

# Pull and run Razer AIKIT
docker run -it \
  --restart=unless-stopped \
  --gpus all \
  --ipc host \
  --network host \
  --mount type=bind,source=$HOME/.cache/huggingface,target=/home/Razer/.cache/huggingface \
  --env HUGGING_FACE_HUB_TOKEN=<YOUR_TOKEN> \
  razerofficial/aikit:latest

Once inside the container, choose your preferred approach:

Option 1: Start with interactive guides

# Start Jupyter Lab for interactive guides (guides will be available at the outputted link)
jupyter lab --ip="0.0.0.0"

💡 Tip: Explore the notebooks/ folder for step-by-step guides and examples!

OR

Option 2: Run a model directly

# Run a lightweight coding model
rzr-aikit model run deepseek-ai/deepseek-coder-1.3b-instruct

For production deployments and monitoring, enable the full stack with Docker Compose:

git clone https://github.com/razerofficial/aikit.git && cd aikit
mkdir -p $HOME/.cache/huggingface
export HUGGING_FACE_HUB_TOKEN=<YOUR_TOKEN>
docker compose -f docker_compose/docker-compose.yaml up -d --pull always

Available Services:

• 📓 Jupyter Lab - Interactive notebooks with examples (http://localhost:8888)
• 📊 Grafana - GPU metrics, model performance, Ray cluster stats (http://localhost:3000)
• 💬 Open WebUI - Chat interface for model testing (http://localhost:1919)
• 🎯 Prometheus - Metrics collection (http://localhost:9090)

Razer AIKit is optimized for NVIDIA accelerated computing platforms with support for both x86-64 and ARM64 architectures.

Workstations & Development

• Razer Blade - High-performance gaming laptops with GeForce RTX 50 series, RTX 40 series, RTX 30 series, and RTX 20 series
• NVIDIA RTX Professional Workstations - RTX PRO 6000 Blackwell, RTX 6000 Ada, RTX A series

Data Center & Enterprise

• NVIDIA DGX Systems (x86-64 and ARM64)
  • NVIDIA GB10 (DGX Spark)
  • NVIDIA GB300, GB200
  • NVIDIA GH200, GH100
• Data Center GPUs: B200, B300, H100, H200, A100, L4, L40, L40S

GPU Requirements

• Minimum: NVIDIA GPU with Compute Capability 7.0 (Volta) or higher
• Supported Architectures: Blackwell, Hopper, Ada Lovelace, Ampere, Turing, Volta
• Detailed Information: GPU Compatibility Guide

We welcome contributions from the community! Special thanks to all our contributors who make this project possible.

Licensed under Apache License 2.0

Additional Resources

• Security Policy - Security reporting guidelines
• Open Source Notices - Third-party acknowledgments
• Contributing Guidelines - Detailed contribution instructions