This repository contains ComfyUI nodes that integrates with the OpenAI API: it allows you to use language models and vision language models within your workflow.
It is KISS by design and intended for those who only wants basic capabilities without having to import massive projects like LLM party: only the chat completions endpoint (with vision support) is implemented as it should be enough for 99% of use cases.
Thanks to its simplicity the project has a low footprint: it only has 1 external dependency (3 in total) !
openaithe official OpenAI API bindings for Pythonnumpyfor computation but it is already a dependency of ComfyUIPillowalso already used by ComfyUI for image processing
The default base_url parameter value targets the official OpenAI API endpoint by default but by changing it, you can also use this project with any OpenAI API compatible servers like Ollama, vLLM, TGI, etc...
Multiples images are supported as long as they are fed batched to the chat completion node.
If you want to customize the chat completion, you can chain options to modify the request. Most common options are available as predefined nodes but you can inject any key/value pair using the Extra body node.
Options nodes are available for:
seedtemperaturemax_tokenstop_pfrequency_penaltypresence_penaltydeveloper_roleextra_body(for any other key/value pair)
Search for OpenAI API in the Custom Nodes Manager and install it.
On the github interface, click the green <> Code button and then Download ZIP. Extract the root folder of the zip file into your ComfyUI/custom_nodes directory.
