Skip to content

this-Demir/hotel-microservices-platform

BranchesTags

Repository files navigation

StayEase - Microservices Hotel Booking System

.NET Next.js TypeScript Tailwind CSS Docker GitHub Actions Azure Vercel AWS Lambda AWS Cognito Supabase MongoDB Redis RabbitMQ OpenAI Resend

Overview

StayEase is a Hotels.com-style hotel booking platform developed as a final project for SE 4458 — Software Architecture & Design of Modern Large Scale Systems.

The system is built on a microservices architecture with eight independently deployed services. All traffic enters through an Ocelot API Gateway which handles JWT validation via AWS Cognito before routing to downstream services. Hotel search, booking, comments, AI-assisted booking, and real-time notifications are each handled by a dedicated service. Relational data lives in Supabase (PostgreSQL), comments in MongoDB Atlas, and cache in Upstash Redis. An asynchronous booking notification pipeline runs over CloudAMQP RabbitMQ. A nightly AWS Lambda job monitors room capacity and alerts hotel administrators when occupancy exceeds 80% for the following month.

Contents

Live URLs

Note: The live deployment has been taken down. The services are no longer running.

Architecture Overview

The system follows a microservices architecture. All client requests are routed through a central API Gateway (Ocelot) where JWT tokens are validated. Downstream services communicate synchronously via REST and asynchronously via RabbitMQ.

Full diagram: docs/diagrams/microservices.md

graph TD
    AdminClient[Admin Client]
    Client[Client]
    IAM[IAM<br>AWS Cognito]
    Gateway[API Gateway<br>Ocelot]
    HotelSvc[Hotel Service]
    CommentsSvc[Comments Service]
    AiAgent[AI Agent Service]
    NotifSvc[Notification Service]
    Lambda[Cron Jobs<br>AWS Lambda]

    Redis[(Hotel Cache<br>Redis)]
    Queue[(Queue<br>RabbitMQ)]
    PgDB[(PostgreSQL<br>Supabase)]
    MongoDB[(NoSQL DB<br>MongoDB Atlas)]
    Resend[Resend<br>Email]
    OpenAI[OpenAI<br>GPT-4o-mini]
    EventBridge[EventBridge<br>nightly]

    AdminClient -->|manage hotels and rooms| Gateway
    Client -->|search hotel / book hotel| Gateway
    Client -->|authenticate| IAM
    AiAgent -->|search and book via tool calls| Gateway

    Gateway -->|JWT validated here| HotelSvc
    Gateway --> CommentsSvc
    Gateway --> AiAgent

    HotelSvc <-->|cache| Redis
    HotelSvc --> PgDB
    HotelSvc -->|new reservation| Queue

    CommentsSvc --> MongoDB

    Queue -->|consume| NotifSvc
    NotifSvc --> Resend
    NotifSvc --> PgDB

    AiAgent --> OpenAI

    EventBridge -->|nightly trigger| Lambda
    Lambda --> PgDB
    Lambda --> Resend
Loading

Deployment

Note: The deployment has been taken down. The information below reflects the architecture as it was built and deployed.

Full CI/CD and deployment diagram: docs/diagrams/deploy_overall.md

Service Platform
client Vercel
admin-client Vercel
api-gateway Azure Container Apps (external ingress)
hotel-service Azure Container Apps (internal)
comments-service Azure Container Apps (internal)
notification-service Azure Container Apps (internal)
ai-agent-service Azure Container Apps (internal)
cron-jobs AWS Lambda + EventBridge

CI/CD is handled by GitHub Actions with path-filtered workflows — a change to one service only redeploys that service. The test job (xUnit) gates the build step on every push to master.

flowchart LR
    Dev[Developer<br>git push to master]
    GH[GitHub Actions<br>path-filtered workflows]
    Tests[dotnet test<br>xUnit]
    GHCR[GHCR<br>Container Registry]
    ACA[Azure Container Apps<br>Germany West Central]
    Lambda[AWS Lambda<br>us-east-1]
    Vercel[Vercel<br>auto-deploy]

    Dev -->|push| GH
    GH -->|test job| Tests
    Tests -->|pass — gates build| GH
    GH -->|docker build + push| GHCR
    GH -->|az containerapp update| ACA
    GH -->|dotnet lambda deploy-function| Lambda
    GH -->|triggers| Vercel
    GHCR -->|pull on deploy| ACA
Loading

Services

Full service relationship diagram: docs/diagrams/microservices.md

Service Responsibility
api-gateway Ocelot reverse proxy — JWT validation, rate limiting, single entry point for all traffic
hotel-service Hotel and room admin CRUD, search with Redis cache and 15% member discount, booking engine (SELECT FOR UPDATE), in-app notifications
comments-service Hotel ratings and comments stored exclusively in MongoDB Atlas
notification-service RabbitMQ consumer — sends booking confirmation emails via Resend and writes in-app notification rows
ai-agent-service GPT-4o-mini orchestration with tool calling to hotel-service Search and Book APIs
cron-jobs AWS Lambda triggered nightly by EventBridge — checks room capacity below 20% for next month and alerts hotel admins
client User-facing Next.js app — search, book, AI chat, comments, notifications, My Bookings
admin-client Admin Next.js panel — hotel and room management, availability windows, image gallery, reservations, notifications

Data Models

Full schema with all fields and assumptions: docs/4_Data_Models.md

PostgreSQL (Supabase)

erDiagram
    Hotels {
        UUID Id PK
        string Name
        string LocationPoint
        text Description
        string AdminEmail
        string AdminSub "nullable"
        string ImageUrl "nullable"
        double Latitude "nullable"
        double Longitude "nullable"
    }
    Rooms {
        UUID Id PK
        UUID HotelId FK
        string RoomType
        decimal BasePrice
    }
    RoomAvailability {
        UUID Id PK
        UUID RoomId FK
        date StartDate
        date EndDate
        boolean IsVacant
        int TotalCapacity
        int ReservedCount
    }
    Reservations {
        UUID Id PK
        UUID RoomId FK
        string UserId
        date CheckInDate
        date CheckOutDate
        int GuestCount
        decimal PricePaid
    }
    Notifications {
        UUID Id PK
        string UserId
        string Title
        text Body
        boolean IsRead
        timestamp CreatedAt
    }
    HotelImages {
        UUID Id PK
        UUID HotelId FK
        string Title
        string ImageUrl
        timestamp CreatedAt
    }

    Hotels ||--o{ Rooms : "has"
    Rooms ||--o{ RoomAvailability : "has"
    Rooms ||--o{ Reservations : "has"
    Hotels ||--o{ HotelImages : "has"
Loading

MongoDB Atlas (comments-service only)

erDiagram
    HotelComments {
        ObjectId _id PK
        UUID hotelId "ref to Supabase Hotels"
        string userId "Cognito sub"
        date travelDate
        float overallRating
        object categoryRatings
        string commentText
        string adminReply "nullable"
        timestamp createdAt
    }
Loading

categoryRatings is an embedded object with fields: cleanliness, staff, facilities, ecoFriendly.

Assumptions

Full assumptions: docs/5_Assumptions.md

  • LocationPoint is a plain text field used for keyword search (LIKE '%Istanbul%'). It is not parsed as coordinates. Latitude and Longitude are separate nullable columns used only for map pins.
  • The 15% member discount is applied server-side whenever a valid Cognito JWT is present on the search request.
  • SELECT FOR UPDATE on the RoomAvailability row is the sole concurrency guard for booking. There is no payment gateway — bookings are confirmed immediately on commit.
  • PricePaid is captured at booking time and is independent of future BasePrice changes.
  • The Cognito sub claim is used as the canonical user identifier across all services.
  • JWT validation happens exclusively at the Ocelot API Gateway. Downstream services trust forwarded headers.
  • Any authenticated user can post a comment. There is no verified-guest restriction.
  • The nightly Lambda job checks capacity for the next calendar month. Alert threshold is less than 20% remaining capacity.
  • The AI agent is stateless — no chat history is persisted between sessions.
  • There is no booking cancellation flow.

Issues Encountered

Full issue log: docs/6_Issues_Encountered.md

  • AWS Lambda does not support .NET 9 — upgraded cron-jobs to net10.0.
  • .NET JWT middleware remaps the sub claim by default — fixed with MapInboundClaims = false on AddJwtBearer.
  • Frontend was sending the access token instead of the ID token — the access token does not carry the email claim, causing downstream rejections.
  • Queue name mismatch between publisher (booking-events) and consumer (booking.events) silently dropped all booking notifications.
  • RabbitMQ consumer crashed on email failure and requeued messages indefinitely — isolated the email call in a try-catch so the message is always acknowledged.
  • Supabase direct connection requires IPv6 — switched to the session pooler URL for IPv4 compatibility.
  • Upstash Redis removed its free tier mid-project — migrated to Redis Cloud with a connection string format change.
  • Initial deployment target was Google Cloud Run — switched to Azure Container Apps for better internal service networking.

Local Development

Prerequisites: Docker, .NET 9 SDK, Node.js 22

# Start local infrastructure (Postgres, MongoDB, Redis, RabbitMQ)
docker-compose up

# Run a .NET service
cd src/hotel-service
dotnet run

# Run the user client
cd src/client
npm install && npm run dev

# Run the admin client
cd src/admin-client
npm install && npm run dev -- --port 3001

# Run tests
cd src/hotel-service-tests
dotnet test

Each service reads configuration from appsettings.Development.json (gitignored). Copy .env.example to .env and fill in credentials before running docker-compose up.

Repository Structure

/src
  /api-gateway                  .NET 9 Ocelot reverse proxy
  /hotel-service                .NET 9 Web API — Admin, Search, Booking
  /hotel-service-tests          xUnit tests
  /comments-service             .NET 9 Web API — MongoDB only
  /comments-service-tests       xUnit tests
  /notification-service         .NET 9 Web API — RabbitMQ consumer + Resend email
  /notification-service-tests   xUnit tests
  /ai-agent-service             .NET 9 Web API — OpenAI GPT-4o-mini orchestration
  /ai-agent-service-tests       xUnit tests
  /cron-jobs                    AWS Lambda (.NET 10) — nightly capacity checker
  /client                       Next.js — user app
  /admin-client                 Next.js — admin panel
/.github/workflows              Per-service CI/CD with path filtering
/docs                           Architecture, data models, assumptions, issues

License

MIT

About

Cloud-native hotel booking platform built with .NET 9 microservices, Ocelot API Gateway, Next.js, RabbitMQ, and OpenAI GPT-4o-mini

Resources

Readme

License

MIT license

Contributing

Contributing
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Contributors

Languages