Osee11

In Benin, reading a handwritten prescription can be difficult for patients and the people helping them. PharmaClear is a prototype that transcribes photographed prescription text, makes uncertain readings visible, and presents documented information from a cited drug label for a pharmacist to review. The Meta Responses API runs muse-wonder-1.0 to read the image and return a transcription with self-reported confidence. A separate deterministic table supplies the cited interaction information; the model does not generate it. Built during the hackathon: the image-reading flow, offline photo queue, C relay with disk queue, sourced reference table, English/French/Fon interface, and evaluation harness. The project history and build process are described in the README’s Hackathon scope and How this was built sections. The evaluation is small, model confidence is not calibrated, and a pharmacist has not reviewed the table. The prototype has not been clinically validated. Next steps include a broader evaluation and pharmacist review.