SD 1.5 and SXDL and influence of clipVision model

[Creative-comfyUI]

I recently started working with Ipadapter, a very interesting tool.

It seems that we can use a SDXL checkpoint model with the SD1.5 for clip vision and SD1.5 IPadapter model, which I thought it was not possible, but not SD1.5 checkpoint with SDXL clip vision and IPadapter model (strange results). Using totaly SDXL or SD sometimes gives what is written in the prompt but not exactly what is expected from Ipadapter and sometimes the opposite.

What is the relationship between Ipadapter model, Clip Vision model and Checkpoint model?

How does the clip vision model affect the result?

Where can we find a clip vision model for comfyUI that works because the one I have bigG, pytorch, clip-vision-g gives errors.