Automate ComfyUI Workflow with LLM Agent

Need a Windows ComfyUI workflow multiple actions automation with a local LLM agent.
I.e. agent needs to:
- operate loading input images from destination folder,
- analyzing position,
- select a most similar positioned reference image,
- load selected reference image,
- start the job,
- wait for ending,
- move the saved image from destination folder to the predefined location,
- etc.
So the job is to prepare the pipeline for comfyUI automation and test different local LLM Agent models for the best correct execution of the task.
Ask technical questions in chat.
p.s.
Don't bid if you cant test LLM agents with comfyUI on you local machine (GPU) or remote machine (GPU), i.e. don't waste your time with ChatGPT copy-pasting, YOUR task is to test all variants and give a working one as a result, else it will be 100% failure, dispute & BAD rating.
p.p.s. funds can be adjusted a bit if needed & 5* rating guaranteed if you will have a WORKING result! No ChatGPT theories will work!
NB: MAX bid is 45 - don't place higher it will be auto-rejected! (can be adjusted after detailed tech discussion)