Portio

AI-powered calorie and nutrition tracker for iOS and Android.

Log food by typing or taking a photo. The app uses an LLM with tool-calling to look up accurate nutritional data — searching OpenFoodFacts for branded products and Google Search for everything else.

Features

• Log food via text or camera photo
• AI analysis (OpenRouter / Gemini) with parallel tool-calling
• Branded product lookup via OpenFoodFacts
• Nutrient search via Google Search (Serper)
• TDEE calculation using Mifflin-St Jeor formula
• AI-generated personalized calorie and macro goals
• Nutrition statistics and history
• Streaks — consecutive days with logged entries
• Nutrient imbalance and overshoot warnings
• Home screen widget
• iOS: HealthKit integration, Liquid Glass UI, deep links (calcal://camera, calcal://add)
• Android: Glance App Widget; Health Connect is disabled until the Android dependency/toolchain is upgraded

Setup

API keys are entered directly in the app on first launch (Settings screen). No config files needed.

Required keys:

• OpenRouter API key
• Serper API key
• Model name (e.g. google/gemini-2.0-flash-001)

Tech Stack

iOS

• Swift + SwiftUI
• SwiftData (local storage)
• WidgetKit + App Groups
• HealthKit

Android

• Kotlin + Jetpack Compose
• Room + DataStore
• Hilt (DI), Retrofit + OkHttp
• CameraX, Glance
• Glance App Widgets

Architecture

Both platforms follow MVVM + Repository pattern. The AI pipeline is identical: insert a placeholder item immediately for optimistic UI, fire the LLM request, update the record on completion.

flowchart TD
    UI["UI Layer"]
    UI --> Tracker["tracker — main screen"]
    UI --> Statistics["statistics — charts and history"]
    UI --> Streak["streak — streak history"]
    UI --> Settings["settings — goals and API keys"]
    UI --> Onboarding["onboarding — first launch profile setup"]
    UI --> Camera["camera — food photo capture"]

    Domain["Domain Layer"]
    Domain --> Model["model — FoodItem, NutritionResponse"]
    Domain --> Util["util — CalorieCalculator, StreakManager, NutrientWarningManager"]

    Data["Data Layer"]
    Data --> Local["local — Room / SwiftData"]
    Data --> Remote["remote — NutritionService, OpenFoodFactsService, SerperService"]
    Data --> Repo["repository — FoodRepository"]

    UI --> Domain
    Domain --> Data

Loading

LLM Loop

Each food query goes through a two-pass agentic loop via OpenRouter.

flowchart TD
    A([User query + optional image]) --> B[Insert placeholder item\nwith isProcessing = true]
    B --> C[Pass 1 — Tool gathering\nLLM with tools enabled]

    C --> D{finish_reason?}
    D -- tool_calls --> E[Execute tools in parallel]

    E --> F{Tool type}
    F -- openfoodfacts_search --> G[OpenFoodFacts API]
    F -- google_search --> H[Serper Google Search]

    G --> I[Merge tool results]
    H --> I

    I --> J[Pass 2 — Final analysis\nNew prompt with gathered data\ntools disabled]
    J --> K[LLM returns JSON\nfoods array]

    D -- stop --> K

    K --> L[Parse FoodArrayResponse]
    L --> M[Update placeholder\nwith nutrition data]
    M --> N[Update widget]
    N --> O([Done])

    C -- error / max 4 iterations --> P([Mark placeholder failed\nshow error])

Loading

Requirements

• iOS 26+ (liquid glass UI)
• Android 8.0+ (API 26+) (Material 3 UI)