☁️Easy API LLM Loader

The pack's regular LLM API loader wants a base_url and api_key typed straight into the node - fine once, annoying every time you drop a new copy of the node into a workflow or share the graph with someone else who then has to paste in their own credentials. This node is the shortcut around that: one dropdown, model_name, and you're done. Everything the dropdown needs comes from config.ini instead of the node itself.

How it actually works

The README lays out the setup this points at directly: open config.ini in the pack's project folder, and for each model you want available, fill in its base_url and api_key (or, for a relay/proxy API, whatever endpoint you're routing through) using config.ini.example as the template. Once a model's entry exists there, it shows up as a choice in model_name on this node - you're picking a preset instead of re-typing credentials on every node instance. That's the entire value proposition: configure once centrally, reuse everywhere in the graph.

Ollama users get an extra shortcut baked into that same config: point base_url at http://127.0.0.1:11434/v1/, put ollama as the API key placeholder, and your model name as model_name - no real key needed because there's nothing to authenticate against on localhost.

The inputs and outputs that matter

There's exactly one input, model_name, and its tooltip is refreshingly direct about what it is: "The model name." It's an enum populated dynamically from your config.ini, which is why it ships with zero built-in choices - if you haven't configured anything yet, the dropdown will be empty, and that's expected, not broken. The single output, model (tooltip: "The loaded model"), is a custom type that plugs into this pack's downstream LLM-calling nodes the same way any of its other model loader outputs do.

Installing it

Get the pack first, either way:

• ComfyUI Manager: search "comfyui_LLM_party", install, restart ComfyUI.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/heshengtao/comfyui_LLM_party, then pip install -r requirements.txt from inside the pack's own folder using ComfyUI's Python (portable builds: python_embeded\python.exe -m pip install -r requirements.txt), then restart.

That gets you the node. It doesn't get you a working model_name dropdown - that needs the config.ini step above done first.

Common issues & troubleshooting

model_name dropdown is empty. This is the single most likely thing to trip people up on their first use - the schema itself ships with zero default choices (choices: 0), because this node has nothing hardcoded to offer. You have to add at least one model to config.ini and restart ComfyUI before anything appears here.

Wrong model gets picked, or credentials look stale. Because everything routes through config.ini centrally, if you edit an entry there after ComfyUI is already running, it won't pick up the change until you restart - a common trap when you're iterating on API keys or trying a different base_url for a relay service.

Just need to test one model quickly, don't want to touch config.ini. Use the pack's full (non-Easy) LLM API loader instead and paste base_url/api_key directly into that node - it's more typing per node, but it skips the config file entirely, which is genuinely faster for a one-off test than editing config.ini for something you'll only use once.