How Qanoon answers
- Your question is matched against every section of the statutes in the corpus, using dense vectors and keyword search together, then re-ranked by a cross-encoder.
- The top provisions are shown to you before the answer is written, so you can read the law while the summary is still being generated.
- A small model writes the answer using only those provisions. Any citation it invents is deleted before you ever see it.
What it will not do: state law from memory. If retrieval finds nothing relevant, it says so instead of guessing.
Each question is answered independently. Follow-ups are treated as fresh questions, so include the context you need in each one.