I can't give you the exact commands, but what you need to do is roughly
this: enable links from the sensor's pixel array through all the entities on
the way to ccdc output video node, and configure the format on the pads at
the ends of those links. The media bus code is the same GRBG10 everywhere
(AFAIR that's what you get from the sensor), size as well. Then you can
capture from the ccdc output video node.

If you want yuv, then add preview, and resizer as well if you need scaling.

media-ctl -p and media-ctl --print-dot and dotfile (in graphviz package) are
helpful. To configure the device, you'll need media-ctl options -l and -V.