On-device scan path
Prescription images and OCR text are designed to stay inside the Android app, with no patient-data backend.
Android engineering preview
A privacy-first Android project for local Korean dispensing-bag OCR, mandatory user review, and official public medicine-data lookup.
Prescription images and OCR text are designed to stay inside the Android app, with no patient-data backend.
Recognized medicine lines must be shown to the user before any official medicine details are treated as confirmed.
Official public datasets are built into generic SQLite artifacts with checksums, signatures, and source provenance.
Medical copy must come from official text, defined templates, or pharmacist-reviewed fixed copy.
Recognized medicine names, OCR text, and prescription images are not server search inputs.
Missing strength, dosage form, or conflicting evidence leads to review, ambiguity, or unresolved states.
The project excludes ads, behavioral analytics, accounts, and health-data monetization.
Current status: Goals 00–08 complete · v0.1.0 test build
CameraX scan with quality gate and bundled offline Korean OCR, deterministic fail-closed matcher, mandatory review, official-source display, local DUR wording, and encrypted private storage.
Reproducible, Ed25519-signed public drug database with independent verification and a rollback/revocation design.
An engineering preview, not cleared for production: pharmacist, regulatory, real-holdout, and pentest gates remain open and block production automatically.