ShelfLife

01 / Context

The problem to solve.

Household ingredients are easy to forget after purchase, which makes expiration dates, duplicate buying, and meal planning difficult to manage.

02 / Approach

From need to interface.

ShelfLife keeps a local pantry inventory, surfaces expiry information, and reduces manual entry through barcode and receipt scanning with optional recipe assistance.

What I built

• Native Compose interface
• Local pantry and expiry tracking
• Barcode and receipt scanning
• AI-assisted recipe ideas

03 / Implementation

How it fits together.

A Kotlin and Jetpack Compose Android client uses Room for local inventory persistence. ML Kit supports barcode and receipt scanning, Firebase provides cloud services and security rules, and a Cloudflare Worker proxies AI requests so provider credentials are not embedded in the app.

Decisions that shaped the build

• Used Room as the local source for essential pantry data
• Kept scan results reviewable before saving
• Routed AI requests through a Worker instead of calling the provider directly from Android

Challenges

• Keeping core pantry data useful without a network connection
• Turning imperfect barcode and receipt input into editable inventory records
• Calling AI features without shipping private service credentials in the APK

04 / Delivery & lessons

Built, tested, and still learning.

Testing

Kotlin unit tests cover pantry-item saving and DAO isolation, Firebase emulator tests validate Firestore rules, and Node tests cover Worker validation. An Android instrumentation test scaffold is present. Run `gradlew test`, `npm test` in `firebase-tests`, and `npm test` in `worker`; broader Compose UI, scanner-device, and end-to-end tests remain to be added.

Deployment

No public store release or downloadable build is linked.

What I took forward

• Local-first storage keeps everyday utilities resilient
• Scanner results need correction paths because recognition is imperfect
• Mobile secrets belong behind a server-side proxy