A single LLM with retrieval is one model answering one prompt, with citation as an instruction it can skip. Burna AI's engine is twelve specialized agents in a cascading constraint pipeline: each agent's output bounds the valid output space of every agent after it, so the output space narrows at every step. The engine cannot produce a grade outside the defined CTCAE set or contradict its own upstream findings, and every grade must cite the CTCAE criterion and the line in the record that supports it: a grade with an incomplete citation set is flagged before it is signed. Attribution uses WHO-UMC and Kramer with per-drug scores and the reasoning shown, and a clinician signs every grade. The constraint is architectural, not a prompt instruction. Two patents filed on the cascading constraint pipeline.
Interactive demo
Try Burna right now, no sign up required.