Initialize a Draw Things GPU server with GPUScript, including script sync, Docker/CUDA/NVIDIA runtime setup, 7T data disk mounting, mergerfs, and end-to-end GPU verification.
Set up and verify a new Draw Things CPU proxy and Envoy server using the scripts in Scripts/ServerManagement/CPUScript. Use when Codex needs to bootstrap a root-access Linux CPU server over SSH/Tailscale, upload model-list and envoy-config.yaml, install Docker/Tailscale images, install or copy compute.drawthings.ai TLS certificates, launch proxy_service and envoy_grpc_web_proxy, and debug port 443, Envoy 8443, TLS, or gRPC Echo connectivity.
Use when writing or refactoring UIKit UI in this repo, especially Workflow/ViewController wiring, Controller/View components, SnapKit layout, Workspace/Dflat-driven UI state, Objective-C adapter responders, or app threading boundaries.
Add a new image or video generative model to the Draw Things app / CLI with a compile-first, end-to-end workflow across SwiftDiffusion, tokenizer plumbing, text encoder, fixed encoder, UNet / DiT runtime, VAE, converter and quantizer tooling, and CLI validation.
Validate Draw Things LoRA training end to end with draw-things-cli, including tiny-dataset training, loss and scaler checks, checkpoint sanity, and base-versus-LoRA generation comparison.
Add or tighten Draw Things LoRA trainer support for generative models available in the Draw Things app / CLI, covering LoRA builders, trainer dispatch, tokenizer and fixed-encoder wiring, checkpoint export, numerical debugging, and validation.