Basically, my question is simple: why does the definition of a locally presentable category require all colimits exist?
The motivation for this is that I was learning about algebraic posets, and had a certain idea of a Grothendieck topology in mind - namely, a covering family of $X$ must contain every compact subobject of $X$. I wanted to know what it would take for representable functors to be sheaves in this topology: after unwrapping definitions, this topology is subcanonical if and only if the poset is algebraic.
I was then studying a somewhat natural generalization of this, where the poset is replaced with a category, and covering sieves need to contain all maps from compact objects. (i.e., a subfunctor $S$ of $h_X$ is a covering sieve on $X$ iff $S(F)=Hom(F,X)$ for each compact $F$). You can then show (I'm less confident about this, but I think it holds!) that this topology is subcanonical if and only if every object is a filtered colimit of compact objects. (This can be thought of as a generalization of the previous paragraph, with filtered colimits replacing suprema).
So I had another statement about a certain kind of category, which I had translated as saying this topology was subcanonical. Indeed, these categories satisfy all the requirements of being locally finitely presentable - except that they don't have all their colimits (though they do have all their filtered colimits).
Now, I'm a little unclear on the definition of locally presentable categories - namely, some sources (such as ncatlab) only require that every object being a (suitably small) colimit of compact objects, while others (see for example here ) require that every object be a filtered colimit. In the latter definition, it would seem to make sense to only require (suitably small) filtered colimits to exist.
So returning to my original question: what is the intuition/motivation behind a locally presentable category, and how does the existence of colimits match that motivation?