oobaboogaOpenAI

oobabooga's text-generation-webui was, for a while, the way to run local LLMs with a nice interface, and NodeGPT is a pack with roots reaching back to exactly that scene - its earliest build talked straight to oobabooga's textgen API. By the AutoGen era it had graduated to the modern approach: this node connects to text-generation-webui through its OpenAI-compatible endpoint, so it can sit in the same "LLM provider" slot as the ChatGPT, LM_Studio, and Ollama nodes.

The name is doing real work: it tells you there's a server running somewhere that speaks OpenAI protocol, and this node is just a client for it. Which is why one of its defaults is delightful - api_key defaults to the string any string here is fine. That's the author flagging, right in the widget, that local servers don't verify keys. It's honest, and it's a useful reminder every time you look at the node.

The fields

• model - defaults to gpt-3.5-turbo, which is a fake name in the OpenAI-compatible sense: text-generation-webui serves whatever model you loaded, and the name is mostly cosmetic. Change it if your setup cares.
• api_key - defaults to any string here is fine. Local, so it genuinely doesn't matter.
• api_type - openai.
• api_base - defaults to http://127.0.0.1:5001/v1. This is the one that matters: it's where text-generation-webui's OpenAI-compatible API listens. You must have oobabooga running with that API enabled, or every call dies here.
• api_version - 2023-05-15, an OpenAI API version stamp that's basically inert against a local server; leave it.

Output: LLM, feeding any agent.

Getting it running

In text-generation-webui, enable the OpenAI-compatible API extension, then:

cd ComfyUI/custom_nodes
git clone https://github.com/xXAdonesXx/NodeGPT

Restart ComfyUI so requirements auto-install, wire oobaboogaOpenAI → Assistant → Chat, and your local model does the thinking. Expect the pack's known rough edges - chats that don't stop, agents that echo - plus the usual local-model slowness. And one more honest note: as of this writing the repo returns 404 and the author's account is gone, so the clone is historical. The pattern, though - "run a server, point a node at it, treat it like OpenAI" - is exactly how the ecosystem still does local LLMs today, which is why this node is worth understanding even now that text-generation-webui has ceded the spotlight to Ollama and LM Studio.