🩺 VRGDG Llama CPP Doctor 🩺

If you've used GGUF models in ComfyUI, you know the phrase "llama-cpp-python" is less a library name and more a prayer. The bindings are infamously finicky to install - they compile against a specific CUDA setup, get confused by pre-existing llama-cpp packages, and break silently across Python environments. VRGDG_LlamaCppDoctor is the pack author's answer: a diagnostic node that runs a bunch of checks on your llama-cpp-python situation and hands you a report and an install hint instead of a blank error.

The pack ships llama-cpp-python in its requirements (and the node exists because enough people hit install hell that she automated the diagnosis). The doctor looks at your Python environment, checks for known-bad packages that collide with llama-cpp-python (it maintains a list of the usual suspects), and can read nvidia-smi for GPU/driver details. It then emits a structured report plus a suggested fix command.

The inputs that matter

There are only two, and both are checkboxes:

• include_install_help - default on. Adds the cleanup commands and documentation for fixing a broken llama-cpp-python to the report.
• include_nvidia_smi - default on. Tries to read nvidia-smi for driver and GPU info when it's on PATH, so the report can speak to your actual CUDA situation.

That's it. This is a diagnostic node - you run it, read the outputs, fix your install, done.

The outputs

• status - a short human verdict.
• report - the full diagnostic dump.
• support_bundle - collected environment details you can paste into an issue or share for help.
• install_hint - the suggested fix command(s) when include_install_help is on.
• python_exe - which Python interpreter ComfyUI is running, which is half the battle when people install into the wrong environment.

How to use it

Drop it into an empty graph, queue it, and read report in a text preview node (or wherever you surface strings). It's not a fix-it button - it's a diagnostic that tells you what's broken so the fix is obvious. If the report points at a conflicting package, uninstall the bad one and reinstall the correct bindings, typically something like:

pip uninstall -y llama-cpp-python llama-cpp llama_cpp
python -m pip install llama-cpp-python --force-reinstall --no-cache-dir

(That's the shape of the fix; the node's install_hint gives you the version of it that matches your environment.)

Install & notes

cd ComfyUI/custom_nodes
git clone https://github.com/vrgamegirl19/comfyui-vrgamedevgirl

Or search vrgamedev in ComfyUI Manager. It's part of the pack's LLM module; if the node doesn't load, it means the whole LLM.py module failed to import, which is itself often a llama-cpp-python problem - which, given what this node does, is a bit on the nose.

Honest review: it's a niche helper, but it's aimed at one of the most genuinely painful recurring installs in the ComfyUI ecosystem. When you've been staring at a llama-cpp-python compile error for an hour, a node that prints "here's exactly what's wrong and here's the command" is worth more than any amount of forum-diving. Run it before you open GitHub issues - the support bundle is exactly what maintainers ask for.