IF LLM List Models 📚

The main IF_LLM node has a dirty little secret: its llm_model field is a plain text box, and it validates your answer against what the provider actually reports. Type llama3.2 when the server calls it llama3.2:latest and you'll get bounced with an "Invalid LLM model selected" error mid-workflow. That's exactly the problem this node exists to kill. IF_LLM_ListModels asks any of the pack's 13 providers "what do you have?" and prints the answer, numbered, in the order the provider hands it over.

It's a tiny node, but it's the kind of thing you'll reach for every time you set up a new machine or point the pack at a new backend. It saves you from the guess-and-check loop.

How it works

Under the hood it's a thin wrapper around the same model-listing code the main node uses for validation. You pick a provider - ollama, llamacpp, kobold, lmstudio, textgen, groq, gemini, openai, anthropic, mistral, transformers, xai, or deepseek - and it hits that provider's list endpoint. For local servers that means an HTTP call to base_ip:port (the default localhost:11434 is Ollama's port, which tells you what the author expects you to run). For API providers it needs a key, pulled from the environment or from the external_api_key field you can type in.

Two things it does beyond the output:

• It writes the list to disk at ComfyUI/model_lists/<provider>_models.txt, so you get a file record, not just a widget you have to wire up.
• The refresh boolean forces a fresh query. That matters: if you ollama pull a new model mid-session and the list looks stale, flip refresh and re-run.

The inputs that matter

Honestly, three:

• llm_provider - which backend to interrogate.
• base_ip / port - only touch these for local servers on another machine or a non-default port.
• external_api_key - fill this for hosted APIs if your key isn't in the environment or .env.

The output is a single model_list STRING that looks like === Available Models for OLLAMA ===\n1. llama3.2\n2. .... Wire it into a text display node - this pack ships IF_LLM_DisplayText - or just read the file it writes. Copy the exact names into the main node's llm_model field and the validation error disappears.

Installing it

It ships with the whole pack, so install once:

cd ComfyUI/custom_nodes
git clone https://github.com/if-ai/ComfyUI-IF_LLM

then restart ComfyUI - or just search "IF_LLM" in ComfyUI Manager and hit install. One real gotcha before you do: the README warns this pack conflicts with IF_AI_tools, so remove that if you have it, or the two fight over the same node names. The requirements file pulls in a lot - torch, transformers, triton, decord, moviepy - so the first install takes a while.

Troubleshooting

• You see No models available or provider requires valid API key/connection. That string is literally in the code, and it means one of three things: Ollama (or whatever server) isn't running, the base_ip/port are wrong, or the API key is missing. Check the last one first - the pack reads keys from env vars with exact names like OPENAI_API_KEY or from a .env file.
• The list is empty but Ollama is clearly running. Check the port. ollama serve listens on 11434 by default, but a second instance or a Docker container will happily squat on something else.
• The main node still rejects a name the list showed. Some servers report the model with a :tag suffix (llama3.2:latest), and that's the string you have to paste in.