ComfyUI-ollama_killer

ComfyUI-ollama_killer
★ 3

进程管理VRAM释放跨平台文本透传
自动检测并优雅终止 Ollama 进程，释放 VRAM 并返回详细状态，同时将文本输入原样透传，提升工作流性能与内存管理。
💡 在运行 Ollama 模型后自动关闭进程释放显存
🍴 2 Forks💻 Python🔄 2025-12-10
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/9d76119b2ef2
📄 README

OllamaKiller Node for ComfyUI

OllamaKiller is a utility node for ComfyUI designed to manage VRAM usage more efficiently by automatically terminating the Ollama process (ollama.exe on Windows, ollama on macOS/Linux). This is particularly useful for users with limited VRAM, allowing them to clear up memory after running models and improve workflow performance.

Features

  • Cross-Platform Support: Works on Windows, Unconfirmed with macOS, and Linux so test yourself.
  • Automatic Process Detection: Finds and terminates Ollama processes
  • Graceful Termination: Attempts clean shutdown before force killing
  • Detailed Status: Provides clear feedback about process termination
  • Seamless Integration: Passes through text input while managing processes in the background

    • Purpose

    Running Ollama in ComfyUI workflows can consume significant VRAM. The OllamaKiller node automatically terminates the Ollama process after each use, freeing up valuable memory and enhancing performance in subsequent generations.

    How It Works

  • When triggered, the node searches for any running Ollama processes
  • It first attempts to gracefully terminate the process
  • If needed, it can force kill stubborn processes
  • The original text input is passed through to the output
  • Process status is returned separately for monitoring

    • Installation

    Method 1: Git Clone

    # In your ComfyUI custom_nodes directory
git clone https://github.com/theshubzworld/ComfyUI-ollama_killer.git

    Method 2: ComfyUI Manager

  • Open ComfyUI
  • Go to Manager tab
  • Search for “OllamaKiller”
  • Click Install

    • Restart ComfyUI after installation.

    Usage

  • Add OllamaKiller to Your Workflow:
  • Place the node after any Ollama model node
  • Connect the text output to the OllamaKiller input
  • The node will automatically handle process termination

  • Inputs:
  • text: Connect any text input (required)
  • trigger: Toggle to manually trigger process termination (optional)
  • force_kill: Enable to force kill stubborn processes (optional)

  • Outputs:
  • status: Information about process termination
  • output_text: The original input text (passed through)

    • Node Information

  • Node Class: OllamaKiller
  • Category: utils
  • Input Types:
  • Text (String): Any text input
  • Trigger (Boolean): Manual trigger (optional)
  • Force Kill (Boolean): Force process termination (optional)
  • Output Types:
  • Status (String): Process termination status
  • Output Text (String): Original input text

    • Example Workflow

  • Add an Ollama model node to your workflow
  • Connect its output to OllamaKiller’s input
  • Connect OllamaKiller’s output to the next node in your workflow
  • Run the workflow – Ollama processes will be automatically managed

    • Troubleshooting

  • If processes aren’t being terminated, try enabling force_kill
  • Ensure you have the necessary permissions to manage processes
  • Check the status output for detailed error messages

    • License

    This project is open-source and available under the MIT License. Contributions are welcome!

    Credits

    Maintained by Shubz World