Omni-Sight Tracker

​## Inspiration I built Omni-Sight Tracker because I wanted to solve a common problem: losing track of physical project components, tools, and materials in a busy workspace. Many people start projects but struggle to stay organized as parts get moved or stored away. I wanted to create a "visual memory" for the real world that helps users catalog their inventory instantly. ​## What I Learned Building this project taught me the importance of data integrity and the difference between "demo" data and "real-world" utility. I learned how to manage a live PostgreSQL database to ensure that user logs are persistent. I also gained experience in prompt engineering with the Gemini 3 Flash API to ensure it provides concise, helpful labels for identifying physical assets. ​## How I Built It The app is built using a modern full-stack approach on Replit. ​Frontend: React with a tactical, high-contrast UI for field use. ​Backend: Node.js handling the logic for image processing and database communication. ​AI: Integration with Gemini 3 Flash for real-time object identification. ​Database: Drizzle ORM and PostgreSQL to store every scan securely. ​## Challenges I Faced One of the primary technical challenges was managing the application state between the manual 'Scan' mode and the automated 'Persistence' mode. I found that initializing the camera for a new scan would occasionally conflict with the background tracking grid, causing the persistence toggle to reset. I learned that handling multiple real-time streams—AI identification, GPS logging, and camera hardware—requires very precise state management to prevent one from overriding the other

Built With

• drizzle-orm
• gemini-3-flash-api
• node.js
• postgresql
• react
• replit
• tailwind-css
• typescript