Nodes/ComfyUI-QING/vLLM-T2I丨API
ComfyUI Node

vLLM-T2I丨API

A ComfyUI node in QING/API with 18 inputs and 4 outputs.

By sheengoa·Created 12 months ago·Updated 3 months ago· 16
vLLM-T2I丨API
    • image
    • raw_response
    • status
    • debug_info
    prompt
    negative_prompt
    platformvLLM
    modeldefault
    base_urlhttp://127.0.0.1:8091
    t2i_endpointchat_completions
    width1024
    height1024
    num_inference_steps50
    true_cfg_scale4.00
    top_p0.90
    repetition_penalty1.10
    seed0
    n1
    response_formatb64_json
    timeout_sec120
    api_key
    advanced_options
    CategoryQING/API

    Inputs (18)

    NameTypeDefaultDescription
    promptSTRING图像生成提示词
    negative_promptSTRING负面提示词;若后端支持会随请求提交
    platformCOMBOvLLM服务平台(用于动态拉取 /v1/models)
    modelSTRINGdefault模型 id
    base_urlSTRINGhttp://127.0.0.1:8091服务根地址
    t2i_endpointCOMBOchat_completionschat_completions 为 vLLM-Omni 文档主路径;images_generations 用于部分 LoRA 等场景
    widthSTRING1024宽度(字符串整数)
    heightSTRING1024高度(字符串整数)
    num_inference_stepsINT501–250
    true_cfg_scaleFLOAT4.000–30写入 extra_body.true_cfg_scale;更多参数请使用 advanced_options
    top_pFLOAT0.900–1
    repetition_penaltyFLOAT1.100.5–2
    seedINT00–2147483647
    nINT11–4仅 images_generations 生效;n>1 时输出 IMAGE 批次
    response_formatCOMBOb64_json仅 images_generations 生效
    timeout_secINT1205–600
    api_keyoptSTRING留空时使用环境变量或 QING 配置
    advanced_optionsoptSTRING高级选项节点输出的 JSON(可通过 custom_json 兜底扩展)

    Outputs (4)

    NameTypeDescription
    imageIMAGE
    raw_responseSTRING
    statusSTRING
    debug_infoSTRING