I recently came across a project that looks promising: [WordLlama](https://github.com/dleemiller/WordLlama?tab=readme-ov-file#s...). It appears to be well-suited for semantic chunking, though I haven’t had a chance to try it out yet.