The build-in pathfinding panda has (pandai) is kind of grid based - at least the navmesh generating tools want a grid-mesh to make a navmesh. I would not suggest using pandai any way, it’s more of a showcase/proof of concept, last time I tried to use it it was prone to crashing or hanging and the agent movement speed depended on the framerate.
Some time ago I’ve made a implementation for Jump Point Search pathfinding, it’s a cool optimization for square, grid based maps. It uses a list of list (2d list) for the map so it should be easy to load maps by chunks, or you could replace that with a dict if your maps don’t fit a square. You may also want to run the pathinding on a separate thread (or better yet a process) - it’s just Python and can be a bit slow.