(Deno) Local LLM Loader：将本地大语言模型接入ComfyUI工作流的节点
英文摘要
The (Deno) Local LLM Loader is a new ComfyUI node that connects local LLMs (Ollama, LM Studio, llama.cpp, vLLM, or any OpenAI-compatible local server) directly into image generation workflows. It enables tasks like turning short ideas into detailed prompts, generating prompts from images via vision-capable models, and chaining multiple LLM steps (e.g., draft → review → final cleanup). The accompanying (Deno) Local LLM Reviewer node can pass or block image outputs based on review text, allowing iterative refinement without leaving ComfyUI. Designed as a local-first tool, it requires no remote API keys and keeps models loaded for efficient prompt chains. The nodes are available through ComfyUI Manager under "Deno Custom Nodes" and on GitHub.
中文摘要
(Deno) Local LLM Loader 是一款新的 ComfyUI 节点，可将本地大语言模型（Ollama、LM Studio、llama.cpp、vLLM 或任何 OpenAI 兼容的本地服务）直接接入图像生成工作流。它支持将简短想法扩展为详细提示词、借助视觉模型从图像生成或审查提示词，以及串联多个 LLM 步骤（如草稿→审阅→最终整理）。配套的 (Deno) Local LLM Reviewer 节点能根据审阅文本通过或阻断图像输出，从而在 ComfyUI 内实现迭代优化。该节点以本地优先为设计理念，无需远程 API 密钥，并在提示链运行时保持模型加载。节点可通过 ComfyUI Manager 搜索“Deno Custom Nodes”安装，代码托管在 GitHub。
关键要点
Connects local LLMs (Ollama, LM Studio, llama.cpp, vLLM, etc.) directly inside ComfyUI workflows, no remote APIs needed.
直接在 ComfyUI 工作流中接入本地大语言模型（Ollama、LM Studio、llama.cpp、vLLM 等），无需远程 API。
Facilitates prompt generation: short idea expansion, image-to-prompt with vision models, and multi-step LLM chains.
简化提示词生成：支持简短想法扩展、利用视觉模型从图像生成提示词，以及多步骤 LLM 链。
Includes (Deno) Local LLM Reviewer node to conditionally pass or retry image outputs based on LLM review text.
包含 (Deno) Local LLM Reviewer 节点，可根据 LLM 审阅文本有条件地通过或重新生成图像输出。
Available via ComfyUI Manager ("Deno Custom Nodes") and open-source on GitHub.
可通过 ComfyUI Manager 搜索“Deno Custom Nodes”安装，并在 GitHub 上开源。