ComfyUI Node
Z-Image API Config
Configure LLM API connection. Supports OpenRouter, local servers, and direct HuggingFace model loading.
Z-Image API Config
- config
◄provideropenrouter►
◄modelqwen/qwen3-235b-a22b:free►
◄api_key►
◄local_endpointhttp://localhost:11434/v1►
◄llm_path►
◄auto_download_fallbackfalse►
◄quantization4bit►
◄deviceauto►
CategoryZ-Image
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| provider | COMBO | openrouter | Select API provider: openrouter (cloud), local (API server), or direct (HuggingFace model loading) |
| model | STRING | qwen/qwen3-235b-a22b:free | Model identifier. OpenRouter: provider/model-name | Local: model name from server | Direct: HuggingFace repo ID |
| api_keyopt | STRING | API key for OpenRouter. Get one at https://openrouter.ai/keys | |
| local_endpointopt | STRING | http://localhost:11434/v1 | Local LLM server endpoint (Ollama: 11434, LM Studio: 1234, vLLM: 8000) |
| llm_pathopt | STRING | Custom path to local LLM models. If set, model name is treated as folder name. If empty, downloads from HuggingFace. | |
| auto_download_fallbackopt | BOOLEAN | false | If enabled: try manual path first, fall back to auto-download if not found. If disabled: manual path is strict (fails if not found). |
| quantizationopt | COMBO | 4bit | Model precision. 4-bit saves VRAM, 8-bit is balanced, FP16/None gives best quality |
| deviceopt | COMBO | auto | Device for direct model loading |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| config | ZIMAGE_CONFIG | — |