Skip to content

gracebml/NutriCare-Agents

BranchesTags
 
 

Repository files navigation

NutriCare Agents: Intelligent Vietnamese Nutrition Platform

License: MIT Project Status: Active Made with Gemini Firebase: Hosting

Team: 404 Brain Not Found

Live Demo (Deployed on Firebase):

✨ https://studio--vietnamese-diet-planner.us-central1.hosted.app/ ✨

image

Table of Contents

Overview

NutriCare Agents is an advanced nutrition advisory system developed for the GDGOC Hackathon Vietnam 2025. The platform leverages cutting-edge AI technology including multi-modal Gemini APIs, Graph Neural Networks (GNN), and an orchestrated multi-agent LLM architecture to deliver personalized nutrition guidance tailored specifically for the Vietnamese population.

Our platform addresses crucial nutrition-related health challenges in Vietnam through a scientifically-grounded, culturally-appropriate approach that considers individual health conditions, taste preferences, regional cuisine variations, and budgetary constraints.

The Challenge

Vietnam faces significant nutrition-related health challenges:

  • Growing Non-Communicable Diseases: Diet-related conditions account for 74% of deaths in Vietnam, with rising rates of cardiovascular disease, type 2 diabetes, and certain cancers.

  • Digestive Health Issues: Poor dietary patterns affect approximately 10% of the population, leading to various gastrointestinal disorders.

  • Limited Personalization: Generic nutritional advice fails to address:

    • Individual health conditions
    • Cultural food preferences
    • Regional cuisine variations
    • Socioeconomic factors

  • Accessibility Barriers:

    • Economic: 78% of the population cannot afford professional nutritional counseling
    • Temporal: 65% lack sufficient time for comprehensive meal planning
    • Physical: Existing solutions often exclude elderly users and those with disabilities
    • Financial: 52% find nutritionally-optimal diets cost-prohibitive

These factors contribute to reduced quality of life, increased healthcare expenditures, and significant health inequities across different population segments.

Our Solution: NutriCare Agents

NutriCare Agents delivers a comprehensive solution through an innovative multi-modal, multi-agent AI platform:

Core Features

  1. Personalized Recommendation Engine: Utilizes Graph Neural Networks trained on extensive Vietnamese nutritional data (2000+ dishes, 500+ ingredients) and user health profiles to generate highly personalized meal recommendations.

  2. Multi-Agent AI System: Employs a sophisticated orchestration of specialized AI agents using Gemini and LangGraph:

    • Search Agent: Grounds recommendations in verified nutritional science
    • RAG Agent: Retrieves data from internal knowledge bases
    • Reasoning Agent: Applies logical inference to nutritional decisions
    • Content Writer: Crafts clear, culturally-appropriate explanations
    • Schedule Agent: Generates optimized daily/weekly meal plans
    • UI/UX Agent: Ensures accessible information presentation

  3. Multi-Modal Interaction: Supports:

    • Text input/output
    • Voice commands via Google Speech-to-Text
    • Image processing for food recognition (future)
    • Video generation for meal preparation guidance (upcoming)

  4. Location-Aware Recommendations: Integrates Google Maps JavaScript API and Places API to locate nearby restaurants and food vendors serving recommended dishes.

  5. Budget Optimization: Dynamically adjusts recommendations based on user-defined budget constraints, offering economical substitutions and seasonal adaptations.

  6. Explainable AI: Provides transparent rationales for all nutritional recommendations, citing authoritative sources including WHO guidelines and Vietnam National Institute of Nutrition data.

Responsible AI Implementation

NutriCare Agents exemplifies responsible AI development:

  • Equitable Access: Designed for users across socioeconomic strata and ability levels
  • Transparency: Visual representation of decision-making processes
  • Scientific Validity: All recommendations grounded in peer-reviewed nutritional science
  • User Privacy: Local data processing with user-controlled data management
  • Cultural Sensitivity: Deep integration of Vietnamese culinary traditions and regional variations
  • Continuous Improvement: Feedback mechanisms for ongoing refinement

Technology Architecture

Untitled-2025-03-26-2204

Core Technologies

Frontend

  • Framework: Next.js, React
  • Styling: Tailwind CSS
  • Authentication: Firebase Authentication
  • Maps Integration: Google Maps JavaScript API, Google Places API
  • Languages: TypeScript, HTML/CSS

Backend & Infrastructure

  • Serverless Functions: Firebase Cloud Functions
  • Database: Firebase Realtime Database
  • Storage: Google Cloud Storage
  • Analytics: Google BigQuery, Google Analytics
  • Development Environment: Firebase Studio (prototype development)
  • Deployment: Firebase Hosting

AI & Machine Learning

  • Foundation Models: Gemini 2.0 Flash (fine-tuned for Vietnamese nutrition)
  • AI Orchestration: LangChain, GenKit, Google GenAI SDK
  • Voice Interfaces: Google Text-to-Speech API, Google Speech-to-Text API
  • Multimodal Processing: Gemini API for image understanding and generation
  • Search & Grounding: Gemini API for web search grounding
  • Recommendation System: FastAPI backend hosted on Vertex AI
  • Data Processing: Data mining and model training via Google Colab

System Architecture Diagram

graph TD
    subgraph "User Interface Layer"
        UI[Next.js Web Application]
        Voice[Voice Interface]
        Map[Google Maps Integration]
        Auth[Google Authentication]
    end

    subgraph "API Gateway Layer"
        APIGateway[Firebase Cloud Functions]
    end

    subgraph "AI Orchestration Layer"
        GenKit[GenKit / GenAI SDK]
        LangChain[LangChain Orchestration]
        AgentSystem[Multi-Agent System]
    end

    subgraph "AI Agents"
        SearchAgent[Search Agent]
        RAGAgent[RAG Agent]
        ReasoningAgent[Reasoning Agent]
        ContentAgent[Content Writer]
        ScheduleAgent[Schedule Agent]
        UXAgent[UI/UX Agent]
    end

    subgraph "AI Services"
        Gemini[Gemini API]
        STT[Speech-to-Text]
        TTS[Text-to-Speech]
        ImgGen[Image Generation]
        VideoGen[Video Generation - Future]
    end

    subgraph "Recommendation System"
        VertexAPI[Vertex AI Endpoint]
        FastAPI[FastAPI Service]
        GNN[Graph Neural Network]
    end

    subgraph "External Services"
        GMaps[Google Maps API]
        Places[Google Places API]
        WebSearch[Web Search Grounding]
    end

    subgraph "Data Layer"
        Firebase[Firebase Realtime DB]
        CloudStorage[Google Cloud Storage]
        BigQuery[BigQuery Analytics]
    end

    UI --> APIGateway
    Voice --> STT --> APIGateway
    Auth --> UI
    
    APIGateway --> GenKit
    APIGateway --> LangChain
    APIGateway --> VertexAPI
    APIGateway --> GMaps
    APIGateway --> Places
    
    GenKit --> AgentSystem
    LangChain --> AgentSystem
    
    AgentSystem --> SearchAgent
    AgentSystem --> RAGAgent
    AgentSystem --> ReasoningAgent
    AgentSystem --> ContentAgent
    AgentSystem --> ScheduleAgent
    AgentSystem --> UXAgent
    
    SearchAgent --> WebSearch
    RAGAgent --> Firebase
    
    AgentSystem --> Gemini
    AgentSystem --> ImgGen
    AgentSystem --> VideoGen
    AgentSystem --> TTS
    
    VertexAPI --> FastAPI
    FastAPI --> GNN
    
    APIGateway --> Firebase
    APIGateway --> CloudStorage
    APIGateway --> BigQuery
    
    Map --> UI

    %% Styling
    classDef frontend fill:#f9f,stroke:#333,stroke-width:2px;
    classDef backend fill:#ccf,stroke:#333,stroke-width:2px;
    classDef ai fill:#cfc,stroke:#333,stroke-width:2px;
    classDef data fill:#ffc,stroke:#333,stroke-width:2px;
    classDef external fill:#fcf,stroke:#333,stroke-width:2px;

    class UI,Voice,Auth,Map frontend;
    class APIGateway,Firebase,CloudStorage,BigQuery backend;
    class GenKit,LangChain,AgentSystem,Gemini,STT,TTS,ImgGen,VideoGen ai;
    class SearchAgent,RAGAgent,ReasoningAgent,ContentAgent,ScheduleAgent,UXAgent ai;
    class VertexAPI,FastAPI,GNN ai;
    class GMaps,Places,WebSearch external;
Loading

Implementation Guide

Requirements

  • Node.js: v18.x or later
  • Google Cloud Account: With access to Gemini API and other Google Cloud services
  • Firebase Account: For authentication, database, and hosting services
  • API Keys:
    • Gemini API
    • Google Maps JavaScript API
    • Google Places API
    • Google Speech-to-Text/Text-to-Speech APIs

Setup Process

  1. Clone the repository:

    git clone https://github.com/technoob05/NutriCare_Agents
cd NutriCare_Agents

  2. Install dependencies:

    npm install

Environment Configuration

Create a .env.local file with the following configurations:

# Required Configurations
GOOGLE_GENAI_API_KEY=your_gemini_api_key
GEMINI_API_KEY=your_gemini_api_key
GOOGLE_MAPS_API_KEY=your_maps_api_key

# Firebase Configuration
NEXT_PUBLIC_FIREBASE_API_KEY=your_firebase_api_key
NEXT_PUBLIC_FIREBASE_AUTH_DOMAIN=your_auth_domain
NEXT_PUBLIC_FIREBASE_PROJECT_ID=your_project_id
NEXT_PUBLIC_FIREBASE_STORAGE_BUCKET=your_storage_bucket
NEXT_PUBLIC_FIREBASE_MESSAGING_SENDER_ID=your_messaging_sender_id
NEXT_PUBLIC_FIREBASE_APP_ID=your_app_id
NEXT_PUBLIC_FIREBASE_MEASUREMENT_ID=your_measurement_id

# Google Cloud Services
GOOGLE_APPLICATION_CREDENTIALS=path/to/service-account-key.json
VERTEX_AI_ENDPOINT=your_vertex_ai_endpoint

Deployment

Development Environment:

npm run dev

Production Build:

npm run build
npm run start

Firebase Deployment:

firebase login
firebase deploy

User Experience

The NutriCare Agents platform provides an intuitive, multi-modal user journey:

  1. Onboarding: Users create accounts via Google Authentication or email
  2. Health Profile: Users provide health information and dietary preferences
  3. Interaction Methods:
    • Text-based queries
    • Voice commands
    • Image uploads (future feature)
  4. Personalized Recommendations:
    • Daily/weekly meal plans
    • Specific dish recommendations
    • Nutritional insights
  5. Location Services: Find nearby restaurants serving recommended dishes
  6. Explanations: Receive detailed rationales for nutritional recommendations
  7. Budget Adaptation: Adjust plans based on financial constraints
  8. Data Management: Export or delete personal data as needed

Roadmap & Future Development

NutriCare Agents has an ambitious development roadmap:

Near-Term (6 Months):

  • Video Generation: AI-generated recipe videos
  • Image Recognition: Food photo analysis for nutritional content
  • Community Features: Recipe sharing and peer support

Medium-Term (12-18 Months):

  • Wearable Integration: Connection with Google Fit and other health trackers
  • Extended Health Conditions: Support for 50+ medical conditions
  • Regional Specialization: North, Central, and South Vietnamese cuisine variants

Long-Term Vision:

  • Clinical Partnerships: Integration with healthcare providers
  • Research Platform: Anonymized data for nutrition research
  • International Adaptation: Frameworks for other cultural contexts

Impact Goals:

  • Reduce nutrition-related healthcare costs by 18% within 3 years
  • Improve quality of life metrics for users with chronic conditions by 30%
  • Decrease food waste by 25% through optimized meal planning
  • Support sustainable food systems through local ingredient recommendations

Development Team

404 Brain Not Found Team:

  • Nguyễn Lâm Phú Quý: Project Lead / Product Manager / Data Scientist
  • Huỳnh Trung Kiệt: AI Engineer (Recommendation Systems)
  • Đào Sỹ Duy Minh: AI Engineer (Multi-Agent Systems)
  • Bàng Mỹ Linh: AI Engineer (NLP Specialist)
  • Phan Bá Thanh: Data Engineer / Infrastructure Specialist

Contributions

We welcome contributions from the community. Please review our contribution guidelines before submitting pull requests.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact Information

About

No description, website, or topics provided.

Resources

Readme

License

MIT license
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages

  • TypeScript 99.4%
  • Other 0.6%