Is sentence splitter still optimal for

the sentence splitter isn't splitting into single sentences, its splitting into chunks that respect sentence boundaries

Ok but how does it factor in things such as titles, subsections and paragraphs under subsections? <h1> vs <h2> etc

I also want to respect subsection boundaries!

I also need to know the maximum size of an m3 chunk in terms of ASCII characters

section boundaries are harder -- those should probably be split before applying a the sentence splitter, using your own algorithm

Even if I am using a bge-m3 embedding model?

What is maximum size of chunk I can use with bge-m3?

bge-m3 has an 8k context limit