Large language models (LLMs) are great at composition and reasoning, but their factual recall is limited to what was baked into their weights at pre-training time.
RAG remedies this by:
- Retrieving relevant pieces of external knowledge on-the-fly with a fast vector search;
- Supplying those snippets to a generator model, which uses them to craft the final answer.
Download from https://github.com/Henningzen/srag
clojure -X:run-xclojure -T:build test
clojure -X:testclojure -T:build ci
java -jar target/no.jansenh/srag-0.1.0-SNAPSHOT.jarWe keep a changelog.
Copyright © 2025 Henning Jansen Distributed under the GNU General Public License v3.0