[PR #2902] [MERGED] Add ComfyUI-NVML-Monitor #2866

New issue

Closed

opened 2026-10-28 21:13:16 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-28 21:13:16 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/ComfyUI-Manager/pull/2902
Author: @robomello
Created: 5/23/2026
Status: ✅ Merged
Merged: 5/31/2026
Merged by: @ltdrdata

Base: main ← Head: add-nvml-monitor

📝 Commits (1)

a0c4c78 Add ComfyUI-NVML-Monitor

📊 Changes

1 file changed (+10 additions, -0 deletions)

View changed files

📝 custom-node-list.json (+10 -0)

📄 Description

Adds ComfyUI-NVML-Monitor to the registry.

What it does: NVIDIA-first hardware telemetry node. Adds a draggable floating chip in ComfyUI showing CPU%, RAM, VRAM, and GPU utilization, plus a click-to-expand popup with detailed GPU stats (utilization, temperature, power draw vs limit, clocks, fan, processes) and a System tab (per-core CPU mini-bars, RAM, driver).

Why another monitor: existing options like ComfyUI-ADLX-Monitor (AMD/Windows) and ComfyUI-XPUSYS-Monitor (cross-vendor) target different use cases. This one is built specifically for NVIDIA + Linux, often running inside Docker. It uses pynvml/nvidia-ml-py directly and includes explicit accounting for VRAM held by external containers/host processes that NVML inside a container cannot attribute to a visible PID — surfaced as a clear "External (other containers / host)" row in the popup's process table.

Implementation:

Backend: Python (monitor.py) with 500 ms server-side cache, thread-safe
Frontend: vanilla JS, no build step, single web/nvml_monitor.js
HTTP API: GET /nvml_monitor/stats returning a JSON snapshot
License: MIT
Initial release: v0.1.0

Tested on RTX PRO 6000 Blackwell (driver 595.58.03) on Linux 6.17 with ComfyUI in Docker. Multi-GPU code path exists but currently single-GPU tested.

Repo: https://github.com/robomello/ComfyUI-NVML-Monitor

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/Comfy-Org/ComfyUI-Manager/pull/2902 **Author:** [@robomello](https://github.com/robomello) **Created:** 5/23/2026 **Status:** ✅ Merged **Merged:** 5/31/2026 **Merged by:** [@ltdrdata](https://github.com/ltdrdata) **Base:** `main` ← **Head:** `add-nvml-monitor` --- ### 📝 Commits (1) - [`a0c4c78`](https://github.com/Comfy-Org/ComfyUI-Manager/commit/a0c4c78e9ec2067ec444049392422404369f2425) Add ComfyUI-NVML-Monitor ### 📊 Changes **1 file changed** (+10 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `custom-node-list.json` (+10 -0) </details> ### 📄 Description Adds [ComfyUI-NVML-Monitor](https://github.com/robomello/ComfyUI-NVML-Monitor) to the registry. **What it does:** NVIDIA-first hardware telemetry node. Adds a draggable floating chip in ComfyUI showing CPU%, RAM, VRAM, and GPU utilization, plus a click-to-expand popup with detailed GPU stats (utilization, temperature, power draw vs limit, clocks, fan, processes) and a System tab (per-core CPU mini-bars, RAM, driver). **Why another monitor:** existing options like ComfyUI-ADLX-Monitor (AMD/Windows) and ComfyUI-XPUSYS-Monitor (cross-vendor) target different use cases. This one is built specifically for NVIDIA + Linux, often running inside Docker. It uses `pynvml`/`nvidia-ml-py` directly and includes explicit accounting for VRAM held by external containers/host processes that NVML inside a container cannot attribute to a visible PID — surfaced as a clear "External (other containers / host)" row in the popup's process table. **Implementation:** - Backend: Python (`monitor.py`) with 500 ms server-side cache, thread-safe - Frontend: vanilla JS, no build step, single `web/nvml_monitor.js` - HTTP API: `GET /nvml_monitor/stats` returning a JSON snapshot - License: MIT - Initial release: [v0.1.0](https://github.com/robomello/ComfyUI-NVML-Monitor/releases/tag/v0.1.0) Tested on RTX PRO 6000 Blackwell (driver 595.58.03) on Linux 6.17 with ComfyUI in Docker. Multi-GPU code path exists but currently single-GPU tested. Repo: https://github.com/robomello/ComfyUI-NVML-Monitor --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-28 21:13:16 +00:00