Does the knowledge base handle Hebrew and Arabic documents?

Yes — right-to-left documents are tokenized and chunked correctly, so retrieval works as well as it does for English.