The difficulty in accurately identifying drug information from packaging, especially for the elderly or when the packaging is damaged, poses significant health risks due to potential misuse. This research addresses this challenge by developing a comprehensive, mobile-based solution. We introduce an Android application that implements a multi-layered approach: (1) on-device Optical Character Recognition (OCR) using Google's ML Kit for fast text extraction, (2) the Jaro-Winkler algorithm to correct and enhance the accuracy of imperfect OCR results, (3) a curated database of 300 common Indonesian drugs, and (4) an innovative chatbot powered by a Large Language Model (LLM) to provide interactive and reliable drug information. The system was evaluated through a series of comprehensive tests.

The results demonstrate high functional stability and a robust Top-5 identification accuracy of 96% across 150 test cases, maintaining 98% accuracy in low-light conditions and 90% on damaged packaging. Furthermore, the application achieved a "Good" usability score of 78.67 from the System Usability Scale (SUS) and received high-to-very-high validity ratings for its content from pharmaceutical experts. This study validates the effectiveness of combining OCR with advanced text matching and a controlled LLM to create a reliable and user-friendly public health tool.