Clinical note → verified facts → FHIR
Every fact, traced back to the sentence it came from.
Paste a note and ClinicalLens pulls out problems, medications, vitals, allergies, and the plan. Each fact is checked against the original text, and anything the model can't back up is set aside instead of shown as fact.
How grounding works
- 01Segment
The server splits the note into numbered sentences with exact character offsets. Plain code, no AI involved.
- 02Extract
The model returns schema-checked JSON. Every fact must cite its sentence numbers and a word-for-word quote.
- 03Verify
Code searches the note for each quote. Found: the fact is linked to its source. Not found: it's flagged and left out of the export.