Instead of embedding chunks, extracting Subject-Verb-Object triples gives exact memory control. Would love feedback on this approach.
· Sign up or log in to comment
