drawthingsai/draw-things-community

The community repository for the Draw Things app.

What it solves

This project provides the public-facing core logic for the Draw Things app, allowing users to run image generation models, samplers, and trainers outside of the main sandboxed application. It specifically enables the ability to self-host an inference server (via gRPC) so that the Draw Things app can offload heavy computation to a separate Mac or Linux machine.

How it works

The repository contains re-implementations of image generation models and data models. It offers two primary ways to interact with this code:

  1. CLI Tools: A draw-things-cli for local image generation and LoRA training on macOS.
  2. gRPC Server: A gRPCServerCLI that can be hosted on macOS or CUDA-capable Linux (via Docker) to act as a remote backend for the Draw Things app's "Server-Offload" feature.

Who it’s for

  • Advanced Users: Those who want to generate images or train LoRAs via a command-line interface on macOS.
  • Self-hosters: Users with powerful Linux servers (RTX 20xx through H100) who want to provide remote GPU power to their Draw Things app.
  • Developers: Those interested in the underlying implementation of the app's models and samplers.

Highlights

  • Cross-Platform Inference: Supports macOS and CUDA-accelerated Linux.
  • Server-Offload: Allows the Draw Things app to connect to a self-hosted gRPC server for remote generation.
  • LoRA Training: Includes a CLI for local LoRA training on macOS.
  • Flexible Deployment: Provides pre-built binaries and Docker images for quick setup of the gRPC server.

Related

  • Project
    SeldonIO/MLServer

    An open-source inference server that provides a standardized REST and gRPC interface for serving machine learning models across various frameworks.

  • Project
    jolibrain/deepdetect

    A deep learning runtime and REST server that provides a unified API for training and inference across images, text, and tabular data.

  • Project
    predibase/lorax

    A multi-LoRA inference server that allows users to serve thousands of fine-tuned LLMs on a single GPU to reduce serving costs.

  • Dispatch
    llama.cpp and llama.app: Local LLM Inference and Ecosystem

    llama.cpp and its official home llama.app provide high-performance, hardware-agnostic local LLM inference, now featuring a simplified 'llama serve' command and integration with local coding agents like Pi.

  • Project
    autonomous-ai/autonomous-computer

    Open‑source hardware designs for building personal AI computers with multiple high‑end GPUs, enabling on‑premise inference, data privacy and zero‑per‑token costs.