This Nix flake provides reproducible, isolated environments for various AI projects, primarily targeting users of NixOS and Linux. It aims to simplify the setup and execution of complex AI tools like ComfyUI and InvokeAI, which often suffer from challenging dependency management.
How It Works
The flake leverages Nix's declarative package management to bundle AI applications and their dependencies. This ensures consistent execution across different environments and simplifies installation. It utilizes Cachix for distributing pre-built binaries, significantly reducing build times. The flake supports GPU acceleration by configuring environment variables for host drivers, including experimental support for Windows via WSL.
Quick Start & Requirements
- Install/Run:
nix run github:nixified-ai/flake/<commit-hash>#<project-name>
(e.g., nix run github:nixified-ai/flake/2aeb76f52f72c7a242f20e9bc47cfaa2ed65915d#comfyui
)
- Prerequisites: Nix package manager. For GPU usage, compatible NVIDIA or AMD drivers are required.
- Configuration: Users need to configure Nix to trust the
ai.cachix.org
binary cache.
- Docs: https://github.com/nixified-ai/flake
Highlighted Details
- Provides pre-configured Nix environments for ComfyUI and InvokeAI.
- Supports NVIDIA GPU acceleration on Linux and Windows (via WSL).
- Offers a NixOS module for easier integration.
- Includes a binary cache for faster setup.
Maintenance & Community
- The project mentions InvokeAI and text-generation-webui as "unmaintained" within the flake, suggesting a focus on newer or actively maintained projects.
- A Matrix channel is available for discussion:
#nixified.ai:matrix.org
.
Licensing & Compatibility
- The README does not explicitly state a license for the flake itself. The included projects (ComfyUI, InvokeAI, text-generation-webui) have their own licenses. Compatibility for commercial use depends on the licenses of the individual AI projects packaged.
Limitations & Caveats
- InvokeAI and text-generation-webui are marked as unmaintained, requiring the use of specific older commits.
- AMD ROCm support for InvokeAI is noted as broken due to Nixpkgs limitations.
- ComfyUI requires manual configuration of custom nodes and models.