PodBench
Personal ComfyUI workflow library and ephemeral RunPod GPU pod launcher that eliminates idle VRAM billing and manual custom-node provisioning.
Managing complex ComfyUI node graphs across ephemeral cloud GPUs usually means either paying for idle network volumes or spending 25 minutes re-pulling 40GB of safetensors checkpoints and custom nodes on every cold boot.
Deterministic Graph Dependency Resolution
PodBench parses exported ComfyUI API and UI JSON graphs to extract every required checkpoint, LoRA, ControlNet weight, and custom node git commit hash. Launching a workflow spins up a targeted RunPod instance (RTX 4090 / A100), mounts a content-addressed model cache, and tears the pod down automatically after an idle timeout.
More Projects
All projects →AI phone receptionist for Australian tradies that answers missed calls on-site, qualifies jobs, quotes call-out rates, and books calendar slots in real time.
AST-aware containerisation tool that inspects repository lockfiles and entrypoints to emit hardened, non-root multi-stage Dockerfiles and Compose stacks.
