Ollama Model Selector xO🎯

If you're wiring up local-LLM nodes in ComfyUI, you'll eventually be typing a model name like dolphin-llama3 into a text field by hand somewhere, and eventually you'll typo it. OllamaModelSelect is the fix: a dropdown of whatever Ollama models are actually installed on your machine, so you pick instead of type.

That's it, that's the whole node - but the schema gives away something genuinely useful about how it works, which is more than you'd normally get from a one-line utility. The default option in the dropdown, when nothing's configured, is the literal string "Ollama not found - Is it installed?". That's not a placeholder someone forgot to remove - it's the node's live error state. It means this node actually queries your local Ollama server for its installed-model list when the graph loads, and it fails loud rather than silent if it can't reach one. That's a small thing, but it's more honest than a lot of custom nodes manage.

How it works

At graph-load time it asks your local Ollama installation what models you have, and populates its dropdown from the answer. If it can't reach Ollama at all, it falls back to that one error string as the only choice.

The inputs and outputs that matter

• model_name - an enum/dropdown, default dolphin-llama3. This is the only input, and it's just a selector.
• MODEL_NAME (STRING) - the single output, and worth pointing out explicitly: it's a plain string, not some special model handle. It exists purely to feed OllamaTextGen's model_name field, which on that node is a free-text STRING input rather than a synced dropdown - this node is what saves you from typing that string yourself and getting it wrong.

How to install it

Standard pack install: ComfyUI Manager, search ComfyUI Neural Nodes, install and restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/xobiomesh/ComfyUI_xObiomesh

then restart. This node needs something the pack itself doesn't install: an Ollama server actually running on your machine. Grab it separately from ollama.com, then ollama pull dolphin-llama3 (or whichever model you want) before this node has anything to show you - Ollama manages its own model downloads entirely outside ComfyUI.

Worth knowing going in: this pack has essentially no public footprint. A search for it turns up nothing, whereas the more established alternative for this exact job - stavsap/comfyui-ollama - shows up in well over a hundred community threads. That doesn't make this pack broken, just obscure; you're on the schema and your own testing more than a community knowledge base.

Common issues & troubleshooting

Dropdown only shows "Ollama not found - Is it installed?" - Ollama isn't running, or isn't reachable on its default port (11434). Start the Ollama app or service first, then reload the ComfyUI page - this node reads the model list once at load time, not continuously, so starting Ollama after ComfyUI is already open won't update the dropdown until you refresh.

Dropdown is missing a model you know you pulled, or still shows one you deleted. Same cause: it's a snapshot from page load. Run ollama list in a terminal for ground truth, then reload ComfyUI to sync the dropdown.

Selection here works fine but OllamaTextGen still fails. This node only hands off a string - it doesn't verify the downstream node can actually reach Ollama itself. If OllamaTextGen's host_url points somewhere different (a remote machine, a Docker container), a clean selection here tells you nothing about whether generation will succeed.