Ai Ml

AI & Machine Learning Configuration

This repository provides first-class support for AI/ML development via local LLMs and hardware acceleration.

Local LLMs with Ollama

  • Service Integration: Managed natively via modules/services/ai/ollama.nix.
  • Hardware Acceleration: Dynamically switches the Nix package (ollama-cuda, ollama-rocm, ollama-cpu) depending on the detected hardware profile.
  • Terminal Translation: A local, privacy-first translation tool is available via hey .translate. It uses the translategemma:4b model to provide high-quality translations directly in your shell or via an interactive Rofi menu (Mod + Alt + T).
  • AI Developer Tools:
    • hey .commit-gen: Generates commit messages from your git diff (use --funny for a sarcastic version).
    • hey .fortune: A witty AI-generated developer fortune in ASCII art.
    • hey .roast: A brutal AI roast of your current system performance and setup.
  • UI Integrations: nextjs-ollama-llm-ui (or open-webui) is configured to provide a web interface for the local language models, bound to 127.0.0.1:11434.

Data Science Environments

  • Unified Suite: This repository uses a 3-tiered architecture for Data Science. For detailed instructions on Python, R, JupyterHub, and Reproducible Research workflows, see docs/data-science.md.
  • JupyterHub: Multi-user interactive data science environments (services.dev.jupyterhub).
  • Python ML Stack: Seamless integration with PyTorch, Transformers, and libraries for data modeling.

Hardware Support

  • Profiles available for NVIDIA CUDA, AMD ROCm, and Intel to accelerate heavy parallel computations.