🏗️ Y&Y App – AI-Powered Industrial ERP

An industrial-grade, microservices-based SaaS that combines live ERP inventory management with a state-of-the-art AI Domain-Expert Agent capable of diagnosing machinery issues in real-time.

Building monolithic apps is a thing of the past. Y&Y App showcases a robust, strictly decoupled Microservices Architecture, separating enterprise business logic (inventory) from complex AI workflows (Retrieval-Augmented Generation), all unified under a blazing-fast React frontend.

✨ The Elevator Pitch

Don't just track your equipment—troubleshoot it. Y&Y App uses a Retrieval-Augmented Generation (RAG) pipeline powered by Google's Gemini APIs and PostgreSQL pgvector. When a user reports a strange noise from a pump, the AI doesn't guess; it performs a vector similarity search to retrieve the exact manufacturer maintenance manual and synthesizes a safe, factual resolution.

🚀 Key Features

• Microservices Architecture: Independent scaling for UI, ERP, and AI logic.
• Enterprise-Grade ERP API: Built with .NET 8 Minimal APIs for extreme performance and type-safe data handling.
• AI RAG Pipeline: A Python FastAPI service utilizing gemini-embedding-001 and the ultra-low latency gemma-4-26b-a4b-it Mixture-of-Experts (MoE) model.
• Unified Vector Database: Google Cloud SQL (PostgreSQL) handles standard relational data and 768-dimensional mathematical vector embeddings in the same place via the pgvector extension.
• Serverless Deployment: APIs deployed on Google Cloud Run (scales to zero) and frontend hosted on Vercel.

🏗️ Architecture & Data Flow

graph TD
    User([User / Browser]) -->|HTTP Requests| UI[React + Vite Frontend - Vercel]
    
    UI -->|Fetch Inventory| ERP[.NET 8 Web API - CloudRun1]
    UI -->|Troubleshoot Query| AI[Python FastAPI - CloudRun2]
    
    ERP -->|SQL Queries| DB[(PostgreSQL + pgvector - Cloud SQL)]
    
    AI -->|1. Create Vector| LLM1[Gemini Embeddings 001]
    AI -->|2. Cosine Similarity Search <= >| DB
    AI -->|3. RAG Prompt Context| LLM2[Gemma-4-26B-A4B-IT]
    
    classDef frontend fill:#61dafb,stroke:#333,stroke-width:2px,color:black;
    classDef backend fill:#512bd4,stroke:#333,stroke-width:2px,color:white;
    classDef python fill:#3776ab,stroke:#333,stroke-width:2px,color:white;
    classDef db fill:#336791,stroke:#333,stroke-width:2px,color:white;
    classDef ai fill:#ea4335,stroke:#333,stroke-width:2px,color:white;

    class UI frontend;
    class ERP backend;
    class AI python;
    class DB db;
    class LLM1,LLM2 ai;

Loading

🛠️ Tech Stack

Tier	Technology	Why we chose it
Frontend	React + Vite	Fast compilation, modern hooks-based UI, easily deployed to Vercel.
Logic (ERP)	.NET 8 (C#)	Industry standard for secure, highly structured, and fast enterprise business data.
Logic (AI)	Python + FastAPI	The undisputed ecosystem for AI/ML integration and data processing.
Database	PostgreSQL + pgvector	Eliminates the need for a separate Vector DB by handling both standard SQL and Cosine Distance (<=>) vector math natively.
LLMs	Gemini API	Gemma 4-26b MoE offers incredible intelligence with low inference latency, ideal for real-time agents.
Cloud	Google Cloud Platform	Seamless integration via Cloud Run and Cloud SQL over secure Unix domain sockets.

💻 Local Development Setup

Want to run this monorepo locally? Follow these steps:

1. Database Configuration

Ensure you have a PostgreSQL instance running with the pgvector extension enabled.

1. Run the initial SQL script located in the tutorial to create the Products and manual_knowledge tables.
2. Insert the seed ERP data.

2. Run the ERP API (.NET)

cd yny.Api
# Update connection string in appsettings.json with your DB credentials
dotnet restore
dotnet run

API will be live at http://localhost:5000

3. Run the AI Agent (Python)

cd yny.AI
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

# Create a .env file with DB_URL and GEMINI_API_KEY
# Run the seed script to inject vector knowledge
python seed.py

# Start the API
python main.py

API will be live at http://localhost:8080

4. Run the Dashboard (React)

cd yny-ui
npm install
# Ensure .env connects to your local .NET and Python ports
npm run dev

App will be live at http://localhost:5173

🌐 Cloud Deployment

This application is fully containerized with Docker and designed for serverless architectures:

• Backend Services: Deployed via Google Cloud Run, utilizing secure Unix Domain Sockets for database connections without exposing TCP ports. Auto-deploys via GitHub Actions CI/CD.
• Frontend UI: Auto-deployed globally via Vercel.

🤝 Contributing

Contributions, issues, and feature requests are welcome! Feel free to check the issues page.

📝 License

This project is MIT licensed.