ComfyUI-SystemOne
ComfyUI nodes that branch workflows on typed Choice, Yes/No and Score judgments from Laya, Jev or Cloudflare Clef.
Overview
ComfyUI System One lets a ComfyUI graph ask natural-language questions such as which style a prompt wants or whether content is too graphic, and route on the typed answer. It provides Choice, Yes/No, Score and Map nodes, plus Clef-based image check and pick-best-image nodes for automatic QA of generated images.
The nodes speak one question-and-answer shape to three interchangeable System One backends: Laya running locally inside ComfyUI, Jev as TypeSafe's hosted API, and Cloudflare's Clef or Clef-flash on Workers AI, which can also see images. A single Backend node switches a whole graph between them. Choice nodes fall back to a configured default when confidence is below a threshold, so an unsure model does not pick a branch on a coin flip. The README states Laya answers take roughly 0.03 to 0.3 seconds on an Apple Silicon Mac after a one-time load.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which style does this prompt want? | — |
| Yes / no | Is this too graphic for kids? | yes · no |
| Score | Where does this prompt fall on the given levels?(inferred) | — |
| Yes / no | Does this generated image fail the yes/no check? | yes · no |
How it works
- Choose a backend (laya, jev, clef-flash, clef) on the Backend node
- Pass the prompt or JSON as state with instructions and options
- Read choice/probability/score plus confidence
- Route with Map, or use the fallback when confidence is below min_confidence
Strengths
- Local free backend (Laya) or hosted backends behind one interface
- Confidence threshold with explicit fallback branch
- Image QA filtering with Clef
- Available from the Comfy Registry
Caveats
- Hosted backends need TypeSafe or Cloudflare credentials
- Laya weights download from Hugging Face on first use
Getting started
comfy node install comfyui-systemone (or search System One in ComfyUI Manager), then export TYPESAFE_API_KEY or Cloudflare credentials for hosted backends.At a glance
- Type
- Service
- Status
- beta
- Language
- Python
- Runtime
- ComfyUI custom nodes
- Has code
- yes
Use cases and integrations
How curated lists describe it
ComfyUI nodes that branch workflows on natural-language judgments from System One models (local Laya or hosted Jev).
hellogumbo/awesome-jev
Frequently asked questions
Which models can it use?
Laya locally, Jev via TypeSafe's API, and Cloudflare Clef or Clef-flash via Workers AI.
Does it cost anything?
Laya is free and local; Jev and Clef are billed per input token by their providers.
What happens when the model is unsure?
The Choice node returns the configured fallback when confidence is below min_confidence.
Related projects
Claude Code plugin where a local decision model (Laya, or Jev) flags profane or abusive prompts and an instruct model rewrites them.
ModelprototypeVideo demo classifying 50 mock YouTube comments with Jev (intent, value, video idea) and comparing speed and cost with GPT-4o-mini.
ModelprototypePythonLocal browser agent that picks each step's operation and element with a Jev-shaped model such as a fine-tuned Laya, no cloud needed.
MCP serverbetaPythonComputer-use platform (Cua Spaces, Driver, Lume) with CUA-S1 small decision models and a preview jev-use example for bounded actions.
ProjectbetaPythonFork of the jev-chat Windows app adding a key-free local LAYA judgment backend and a DPI-aware screen capture fix.
ProjectbetaPythonChinese multi-seat exploration engine for game and rule-system design; Jev only gates whether to start deep exploration or ask.
ProjectbetaJavaScript