📊 Service Status Check

When a remote-service node gives you an empty dropdown, the first question is always: is the server actually running, and does it see my models? The Service Status Check node exists to answer that for the Nexa SDK path without you running curl commands. It's the pack's diagnostic node - less a workflow piece, more a "why is nothing working" tool that surfaces both sides of the equation: what the remote server reports and what your local model folder contains.

How it works

Three inputs, all boring and that's fine:

• base_url - the Nexa SDK server address, default http://127.0.0.1:8080. Nexa's local server speaks OpenAI-compatible on that port, same as Ollama and LM Studio on theirs.
• models_dir - the local model directory it scans. Note the default is a /tmp/ComfyUI/models/LLM path, which is a quirk worth knowing: your real ComfyUI models live under ComfyUI/models/LLM/GGUF, so if the local_models output looks empty or wrong, set this to your actual directory.
• refresh - a boolean you flip to force a rescan and re-ping.

The node returns three strings: status (whether the server answered), remote_models (what the server reports), and local_models (what it found on disk). That's the whole point - seeing the two lists side by side tells you instantly whether the problem is on the server side (empty remote_models) or your side (empty local_models).

When you'd reach for it

Nexa SDK is the least common of the three backends this pack supports, so this node is squarely aimed at people running Nexa's local inference (it's a real thing - a local AI SDK that runs GGUF models and serves them HTTP-style on 8080, made by nexa.ai). The practical uses: sanity-checking that the server is up before you wire a config node, and confirming a freshly downloaded model actually landed where the pack looks. It's an output node, so you use it as a standalone diagnostic - run it, read the three strings, fix whatever's broken.

Install

Same pack, same command:

cd ComfyUI/custom_nodes
git clone https://github.com/walke2019/ComfyUI-GGUF-VLM.git
cd ComfyUI-GGUF-VLM
pip install -r requirements.txt

Restart ComfyUI and the node's there. The honest take: if you're not using Nexa, you don't need this node at all - the equivalent diagnosis for Ollama or LM Studio is just clicking Refresh Models on the config nodes. But if you're on Nexa, it's the difference between guessing at which half of the pipeline is broken and knowing in one run.