This project has been created as part of the 42 curriculum by iouhssei, mel-rhay, aessadik,achahbal, aait-bou

Rihla — Student Travel Platform for Morocco

Description:

Rihla is a community-driven travel platform built for students exploring Morocco. It combines location-based discovery, social interaction, and AI-powered support to make travel easier, safer, and more social.

MVP Features

Area	Description
Authentication	Email register/login, logout, Google + 42 OAuth, cookie JWT + CSRF, change password (local accounts), link providers in Settings
Users & profiles	Profiles (username, avatar, bio), edit profile, change-password flow
File uploads	Avatars via profiles service
Location & places	Google Places discovery, city & saved places, planner integration
Activity-based discovery	Places categorized by activity type
Ratings & reviews	Community reviews + aggregates (review-places)
Real-time	friends-service: REST + WebSockets (chat / notifications; polish ongoing)
AI	Trip planner (Gemini) + places suggestions (places-service)

Technical stack

Frontend technologies and frameworks

• React with TypeScript for a typed, component-based UI.
• Vite as the build tool and dev server for fast iteration and modern ESM output.
• React Router for client-side routing.
• Tailwind CSS (v4) for utility-first styling, with Radix UI primitives and helpers such as class-variance-authority, clsx, and tailwind-merge for consistent, accessible components.
• GSAP for motion where richer animation is needed; Lucide React for icons.

Backend technologies and frameworks

• Node.js across services, with a microservices layout: separate processes for auth, profiles, friends, AI places, reviews, favorites, planner, etc.
• NestJS (with Passport, JWT, validation, Prisma) for auth-service and profiles-service, where a structured module architecture and strong typing pay off for security and user data.
• Express for lighter HTTP services (e.g. friends-service, planner-service, ai-places-service, review-places, fav-places) with Prisma where persistence is required.
• Argon2 for password hashing; cookie-based sessions and CSRF patterns aligned with the SPA.
• WebSockets (ws in friends-service) for realtime chat and notifications, behind the HTTPS gateway.
• Google Generative AI (@google/generative-ai) in the planner path; Gemini used for trip planning and related AI features.
• ioredis and express-rate-limit (with Redis backing) for caching and throttling on high-churn routes (e.g. AI and places).

Database system and why it was chosen

• PostgreSQL is the primary relational store (containerized in the stack, with Prisma migrations per service).
• It was chosen for ACID transactions, a clear relational model (users, profiles, reviews, favorites, planner data, etc.), strong tooling, and straightforward scaling patterns for a multi-service backend. Prisma gives schema-first modeling, migrations, and type-safe access from TypeScript/JavaScript services.

Other significant technologies and libraries

• Redis for caching and rate limiting, reducing load on Postgres and external APIs.
• HashiCorp Vault (with Docker secrets files) for centralized secret management and bootstrap of credentials.
• Docker and Docker Compose for reproducible environments and service orchestration.
• Nginx with ModSecurity as a TLS-terminating reverse proxy and API gateway to the internal network.
• Swagger UI for API exploration; Netdata for monitoring; Logbull and a log forwarder for log aggregation (optional depending on configuration).

Justification for major technical choices

• Microservices isolate failure and deployment boundaries (auth vs. profiles vs. realtime vs. AI), match the ft_transcendence-style scope, and let each area use the smallest stack that fits (Nest for “core domain” services, Express for focused HTTP/WS workers).
• Prisma + PostgreSQL keeps data access consistent and migratable across several repos/services without ad hoc SQL sprawl.
• Vite + React prioritizes developer experience and a responsive SPA while staying easy to serve behind a single HTTPS entrypoint.
• Redis is a pragmatic addition for performance and abuse protection on AI- and search-adjacent endpoints.
• Vault + gateway support security-by-design: secrets are not baked into images, and traffic can be inspected and routed uniformly.

Features list

Public and marketing

• Landing / hero (/) with discovery entry and navigation.
• Privacy (/privacy).
• Healthcheck (/healthcheck) for service URL checks and diagnostics.

Authentication and account security

• Registration (/register) and login (/login).
• Logout and session teardown via auth API.
• Google OAuth and 42 Intra OAuth; OAuth success handling (/oauth-success).
• HTTP-only cookie session model with JWT and CSRF token flow for mutating requests.
• Change password for local accounts (/profile/change-password).
• Settings (/settings): linked OAuth providers and related account configuration.

Profile and preferences

• Profile view (/profile) and edit profile (/profile/edit): username, display name, avatar, bio, interests (JSON on profile).
• Avatar upload via profiles service / uploads path behind the gateway.
• Interests onboarding or editing (/interests).

Places, discovery, and content

• Home discovery experience (/home): search and exploration UI.
• City page (/city): city-scoped discovery and place detail context.
• Google Places–backed discovery and autocomplete (via ai-places-service and gateway routes).
• AI place suggestions (Gemini + caching) for cities.
• Saved places (/saved): list and manage favorites (fav-places), including category metadata per place.
• Reviews and ratings for places: create/read community reviews (review-places), aggregates exposed to the UI where implemented.

Trip planning (AI)

• Planner (/planner): generate and work with AI trip plans (Gemini via planner-service), persisted as TripPlan records (city, days, preferences, JSON plan, dates); trip plan modal surfaces per-day activities (name, time, category, ratings, favorites).

Social and realtime

• Friends (/friends): friend requests and friendships (friends-service REST).
• Web chat (/webchat): WebSocket-based messaging (ws), conversations and messages stored in Postgres.
• Notifications (/notifications) with realtime notification delivery / toasts (NotificationRealtimeProvider, glass toast UI).
• Nav badges for unread or actionable items where wired.

Experience and resilience

• Protected routes for authenticated sections.
• Theme switching (light/dark) via ThemeProvider.
• Global toasts for feedback (GlassToastProvider / stack).
• 404 handling (NotFoundPage).

Instructions

Prerequisites

• Docker Engine with the Compose V2 plugin (docker compose …, as used by the Makefile).
• Git, to clone the repository.
• Optional (local development without the full stack): Node.js 18+ (npm or pnpm). Some services (for example ai-places-service) document Node 20+ for running outside Docker; the full stack builds Node versions inside images.
• OAuth / API access as needed: Google OAuth, 42 Intra OAuth, Google Gemini, Google Places (see secret files below).
• Sufficient disk and RAM for many containers (database, Vault, microservices, reverse proxy, monitoring).

Configuration

1. Repository root .env

docker-compose.yml loads ./.env at the repo root for multiple services (including frontend). Create it (not committed) before starting the stack. At minimum, define ports and shared settings your deployment expects, for example:

# Ports
PORT_FRONT=5173
PORT_POSTGRES=5432
PORT_VAULT=8200

# CORS / redirects
FRONTEND_URL=http://localhost:5173

# JWT (shared across backend services)
JWT_ACCESS_SECRET=change_me
JWT_ACCESS_EXPIRES_IN=15m

# Postgres — Prisma / services (use a URL that matches your setup; Docker services usually reach Postgres on the compose network)
DATABASE_URL=postgresql://postgres:password@database:5432/rihla

# External APIs (also mirrored for Vault where applicable)
GEMINI_API_KEY=your_gemini_api_key
GOOGLE_PLACES_API_KEY=your_google_places_api_key

# Redis
REDIS_URL=redis://redis:6379

2. Vault secret files

The vault service mounts Docker secrets: from backend/devops/secrets/. Create these files (typically one line each, gitignored) before the first full start:

• google_client_id
• google_client_secret
• callback_url
• frontend_url
• gemini_api_key
• google_places
• fortytwo_client_id
• fortytwo_client_secret
• fortytwo_callback

Run the project

1. Clone the repository and cd into the project root.
2. Add ./.env and populate backend/devops/secrets/ as above.
3. Ensure Docker is running.
4. From the project root, build and start the full stack:

   make all

   This runs docker compose -f ./docker-compose.yml build … then up -d (see Makefile targets below). The first run can take several minutes.
5. Wait until healthchecks settle (especially Vault, database, then app services).
6. Open the app at https://localhost (port 443). The gateway uses a self-signed certificate; your browser will warn you—continue for local development.
7. Logs: make logs or docker compose -f ./docker-compose.yml logs -f [service].
8. Stop: make stop. Tear down and remove volumes/images: make clean or make fclean (destructive; see Makefile).

Hostnames behind the gateway (HTTPS on port 443): the ModSecurity/nginx config uses names such as auth.localhost, netdata.localhost, and logbull.localhost for some tools; the default server handles localhost for the main UI and API paths.

Makefile targets

From the repo root, with DOCKER=docker and COMPOSE=./docker-compose.yml:

Target	What it does
make all	Full docker compose build (--no-cache, --parallel, --force-rm, --pull) then up -d --remove-orphans.
make front	Build and start only the frontend service.
make logs	docker compose … logs -f (all services), or logs -f $1 if shell $1 is set.
make start / make stop	docker compose … start / stop.
make clean	Stop stack, down --remove-orphans --rmi all -v, then prune listed containers, images, volumes, networks.
make fclean	clean plus docker system prune -a --volumes -f.
make re	fclean then all.

Note: The Makefile defines make backend, but docker-compose.yml has no backend service, so that target will fail until the compose file and Makefile are aligned. Use make all for the full stack.

Team Information

Ilyass Ouhsseine (iouhssei)

Roles: Product Owner, Project Manager, Frontend Developer
Description:
Defined and led the product vision of Rihla, a travel discovery platform built on a microservices architecture (React / Node.js). Managed the full product lifecycle — from ideation and feature prioritization to delivery — while contributing to frontend development. Focused on AI-powered trip planning, personalized recommendations, and seamless user experience.

---

Amine Essadiki (aessadik)

Roles: Tech Lead, DevOps Engineer
Description:
Led the infrastructure and DevOps strategy, ensuring scalability, reliability, and security across services. Designed and maintained CI/CD pipelines, containerized services (Docker), and managed deployment workflows. Oversaw system architecture decisions and performance optimization.

---

Mohammed El Rhayour (mel-rhay)

Roles: Tech Lead, Backend Developer
Description:
Architected and developed core backend services using a microservices approach. Designed APIs, handled business logic, and ensured system consistency and performance. Collaborated closely on authentication, data models, and service communication.

---

Anas Chahlabani (achahbal)

Roles: Mobile Developer
Description:
Developed the mobile application experience for Rihla, focusing on performance, usability, and consistency with the web platform. Integrated APIs and contributed to delivering a smooth cross-platform user experience.

---

Ali Ait Bouih (aait-bou)

Roles: Backend Developer
Description:
Contributed to backend service development, including API implementation, database interactions, and feature integration. Worked on ensuring reliability and maintainability of services.

Project Management

Team Organization & Task Distribution

The team of five was organized around complementary roles, with each member owning specific domains while collaborating across boundaries:

Member	Role	Responsibilities
Ilyass Ouhsseine (iouhssei)	Product Owner, Project Manager, Frontend Dev	Product vision, feature prioritization, sprint planning, AI features, frontend development
Amine Essadiki (aessadik)	Tech Lead, DevOps Engineer	Infrastructure, CI/CD, Docker orchestration, Nginx gateway, Vault, security, deployment
Mohammed El Rhayour (mel-rhay)	Tech Lead, Backend Developer	Core backend architecture, API design, auth flows, data models, service communication
Anas Chahlabani (achahbal)	Mobile Developer	Mobile application, API integration, cross-platform UX consistency
Ali Ait Bouih (aait-bou)	Backend Developer	Backend service development, API implementation, database interactions, feature integration

Agile Workflow

The team followed an Agile methodology with iterative development cycles:

• Sprint planning at the start of each iteration to align on deliverables
• Daily follow-ups to track progress and surface blockers early
• Sprint retrospectives to improve processes over time
• Features were broken down into actionable tasks with clear ownership
• Continuous delivery was prioritized, with each service independently deployable

Project Management Tools

• GitHub Issues & Pull Requests — task tracking, code review, and change management. Feature branches followed the convention feature/your-feature-name; commits followed Conventional Commits (feat:, fix:, docs:, etc.)
• GitHub Projects / Kanban board — visual progress tracking across backlog, in-progress, and done columns

Communication Channels

• Discord — primary async communication channel: dedicated channels per service domain, design decisions, and daily check-ins
• In-person / video meetings — sprint planning, retrospectives, architecture discussions, and decisions requiring synchronous alignment

---

Tech Stack

Frontend

Tool	Version	Purpose
React	19.2	Modern UI framework with hooks
Vite	7.2	Lightning-fast build tool and dev server
TypeScript	5.9	Type-safe JavaScript
Tailwind CSS	4.1	Utility-first styling framework
Radix UI	1.4	Unstyled, accessible component primitives
React Router	7.13	Client-side routing with protected routes
Lucide React	0.563	Icon library
GSAP	3.14	Animation library
class-variance-authority + clsx + tailwind-merge	—	Consistent, conditional class composition
ESLint	9	Code quality & consistency

Justification: React 19 + Vite offers the best developer experience for a fast-moving SPA — near-instant HMR, ESM output, and a component model that scales well across many pages. Tailwind + Radix keeps styling consistent and accessible without a heavy component library. GSAP handles richer animations that CSS transitions can't easily express.

Backend

Tool	Version	Purpose
NestJS	10	Auth + Profiles microservices (structured modules, DI, guards)
Express	4/5	Lightweight HTTP services (places, planner, reviews, favorites, friends)
TypeScript	5.x	Type safety across all services
Prisma	6.x	Schema-first ORM with migrations
Passport.js	0.7	OAuth 2.0 & JWT middleware
Argon2	—	Industry-standard password hashing
ws	—	WebSocket server for real-time chat and notifications
ioredis	—	Redis client for caching and rate limiting
@google/generative-ai	—	Gemini API integration for trip planning

Justification: NestJS is used for the two highest-security services (auth and profiles) where its module system, guards, and dependency injection provide the strongest safety guarantees. Express is used for the remaining, more focused services where a lighter footprint fits better. Separating services this way means a crash or vulnerability in one domain (e.g. AI places) cannot directly bring down authentication.

Database System

PostgreSQL 15+ is the primary relational store, accessed through Prisma in every service that requires persistence.

Why PostgreSQL:

• ACID transactions ensure data integrity across reviews, favorites, and user records
• Strong relational model fits the domain well — users, profiles, friendships, reviews, and trips all have clear relationships
• Excellent TypeScript tooling via Prisma — schema-first modeling, auto-generated migrations, and fully type-safe queries
• Proven scalability patterns for when the platform grows
• Each service manages its own Prisma schema and migrations, keeping domain boundaries clean

Redis 7+ complements PostgreSQL as a caching and rate-limiting layer, reducing load on the database and external APIs on high-churn routes.

Other Significant Technologies

Technology	Purpose
Docker + Docker Compose	Reproducible environments and full-stack orchestration
Nginx + ModSecurity	TLS-terminating reverse proxy, WAF, and API gateway
HashiCorp Vault	Centralized secrets management — credentials never baked into images
Elasticsearch + Kibana	Centralized log aggregation and visualization
Netdata	Real-time system and container metrics

Justification for Major Technical Choices

Microservices architecture: Each service has independent deployment boundaries and failure domains. A bug in the AI planner cannot take down authentication. It also lets each service use the smallest stack that fits its needs.

Cookie-based JWT + CSRF: More secure than storing tokens in localStorage for a web SPA. The double-submit CSRF pattern protects mutating requests, and HTTP-only cookies prevent XSS exfiltration.

Nginx + ModSecurity as gateway: Traffic inspection, WAF rules, and TLS termination happen in one place, so internal services can communicate over plain HTTP without worrying about encryption and filtering individually.

Vault for secrets: API keys and OAuth credentials are injected at runtime rather than committed to the repo or baked into images, aligning with security-by-design principles.

---

Database Schema

Persistence is PostgreSQL, accessed through Prisma in each service. The stack uses one database instance; each service ships its own schema.prisma and migrations for the tables it owns. userId values align with User.id from the auth/profiles domain (string CUIDs), so reviews, favorites, chat, and trips all refer to the same identity without cross-schema foreign keys in every service.

Visual Overview (ER Diagram)

erDiagram
  User ||--o{ Account : "1:N"
  User ||--o| Profile : "0..1"
  Conversation ||--o{ ConversationParticipant : "N:M bridge"
  Conversation ||--o{ Message : "1:N"

  User {
    string id PK
    string email UK
    string hashPassword
    boolean isEmailVerified
    datetime createdAt
    datetime updatedAt
  }

  Account {
    string id PK
    string userId FK
    string provider
    string providerAccountId
    string accessToken
    string refreshToken
    int expiresAt
    datetime createdAt
  }

  Profile {
    string id PK
    string userId FK_UK
    string username UK
    string displayName
    string avatar
    string bio
    string status
    json interests
    datetime createdAt
    datetime updatedAt
  }

  FriendRequest {
    string id PK
    string fromUserId
    string toUserId
    datetime createdAt
    datetime updatedAt
  }

  Friendship {
    string id PK
    string userLowId
    string userHighId
    datetime createdAt
  }

  Conversation {
    string id PK
    datetime createdAt
  }

  ConversationParticipant {
    string conversationId FK
    string userId
  }

  Message {
    string id PK
    string conversationId FK
    string senderId
    string content
    datetime createdAt
  }

  Notification {
    string id PK
    string userId
    string type
    string title
    string body
    json data
    boolean read
    datetime readAt
    boolean archived
    datetime archivedAt
    datetime createdAt
  }

  Review {
    string id PK
    string placeName
    string city
    string userId
    int rating
    string comment
    datetime createdAt
    datetime updatedAt
  }

  SavedPlace {
    string id PK
    string userId
    string placeName
    string city
    string category
    string address
    string image
    string placeId
    float rating
    string status
    datetime savedAt
  }

  TripPlan {
    string id PK
    string userId
    string city
    int days
    json plan
    date tripStartDate
    date tripEndDate
    datetime createdAt
    datetime updatedAt
  }

Loading

Tables and Relationships (by Service)

auth-service / profiles-service: User is the root identity. Account rows are many-to-one with User (unique on [provider, providerAccountId]). Profile is an optional one-to-one with User. Deleting a user cascades to both accounts and profile.

friends-service: FriendRequest stores directed requests (unique pair). Friendship stores an undirected pair created after acceptance. Conversation has many Message rows and many users via ConversationParticipant (composite PK: conversationId + userId). Notification is per userId, indexed for unread and archive queries.

review-places: Review — one review per user per place name + city (@@unique([userId, placeName, city])).

fav-places: SavedPlace — one saved row per user per place name + city, with optional placeId, rating, image, and status.

planner-service: TripPlan — per userId and city, with days, preferences (Postgres text array), structured plan JSON, and optional trip date bounds.

Key Fields and Data Types

Category	Fields	Type
Identifiers	User.id, Account.id, Profile.id, FriendRequest.id, Friendship.id, Notification.id	String — cuid()
Identifiers	Conversation.id, Message.id, Review.id, SavedPlace.id, TripPlan.id	String — uuid()
Text	emails, usernames, provider keys, message content, review comments, place names, addresses	String
Booleans	User.isEmailVerified, Notification.read, Notification.archived	Boolean
Numbers	Account.expiresAt (epoch), Review.rating (1–5), TripPlan.days, SavedPlace.rating	Int / Float
JSON	Profile.interests, Notification.data, TripPlan.plan (serialized AI itinerary)	Json
Arrays	TripPlan.preferences (native Postgres text array)	String[]
Timestamps	createdAt, updatedAt across all models	DateTime
Date-only	TripPlan.tripStartDate, TripPlan.tripEndDate	Date (@db.Date)

ft_transcendence Plan (Total Score: 23/14)

---

iouhssei: [Minor] Custom-made design system: Reusable components, color palette, typography. [Major] Recommendation system: Machine learning based on user behavior.[Minor] Use a frontend framework: React, Vue, Angular, Svelte, etc. mel-rhay: [Minor] Notification system: For creation, update, and deletion actions. [Major] Implement real-time features: WebSockets, real-time updates, efficient broadcasting.[Minor] Use a backend framework: Express, Fastify, NestJS, Django, etc.[Major] Allow users to interact with other users: Chat, Profile, Friends system. aait-bou: [Minor] Use an ORM: Database Object-Relational Mapping. [Minor] Remote authentication: OAuth 2.0 (Google, GitHub, 42, etc.). achahbal: [Minor] Server-Side Rendering (SSR): Improved performance and SEO. [Minor] Support for additional browsers: Min 2 additional browsers (Firefox, Safari, etc.).[Minor] Health check & status page. [Major] Custom Major Module: Substantial technical complexity. Requires justification. aessadik: [Major] WAF/ModSecurity + HashiCorp Vault: Strict WAF config and secure secrets management. [Major] Backend as microservices: Loosely-coupled services.[Major] Prometheus & Grafana: Monitoring and alerting.

refactor this in markkdown and improve wrinting 3:12 PM Identified request to restructure and enhance project documentation formatting

The user wants me to reformat and improve the writing of their ft_transcendence plan in markdown. Done ft_transcendence — Project Plan

Total Score: 23/14 iouhssei Type Module Minor Custom Design System — Reusable components, color palette, and typography. Minor Frontend Framework — React, Vue, Angular, Svelte, or equivalent. Major Recommendation System — Machine learning model based on user behavior. mel-rhay Type Module Minor Notification System — Alerts for creation, update, and deletion actions. Minor Backend Framework — Express, Fastify, NestJS, Django, or equivalent. Major Real-Time Features — WebSocket integration with efficient event broadcasting. Major User Interactions — Chat, user profiles, and friends system. aait-bou Type Module Minor ORM Integration — Database access via Object-Relational Mapping. Minor Remote Authentication — OAuth 2.0 support (Google, GitHub, 42, etc.). achahbal Type Module Minor Server-Side Rendering (SSR) — Improved performance and SEO. Minor Cross-Browser Support — Compatibility with at least 2 additional browsers (Firefox, Safari, etc.). Minor Health Check & Status Page — Service availability monitoring endpoint. Major Custom Major Module — Substantial technical complexity; requires written justification. aessadik Type Module Major WAF / ModSecurity + HashiCorp Vault — Strict WAF configuration with secure secrets management. Major Microservices Architecture — Backend decomposed into loosely-coupled, independently deployable services. Major Prometheus & Grafana — Full monitoring stack with metrics collection and alerting.

in markdown 3:12 PM Deduced user's markdown export preference

Here's your .md file, ready to download or share with the team. Ft transcendence plan Document · MD Want to be notified when Claude responds?

Claude is AI and can make mistakes. Please double-check responses. Ft transcendence plan · MD

ft_transcendence — Project Plan

Total Score: 23/14

---

iouhssei

Type	Module
Minor	Custom Design System — Reusable components, color palette, and typography.
Minor	Frontend Framework — React, Vue, Angular, Svelte, or equivalent.
Major	Recommendation System — Machine learning model based on user behavior.
Minor	Cross-Browser Support — Compatibility with at least 2 additional browsers (Firefox, Safari, etc.).

---

mel-rhay

Type	Module
Minor	Notification System — Alerts for creation, update, and deletion actions.
Minor	Backend Framework — Express, Fastify, NestJS, Django, or equivalent.
Major	Real-Time Features — WebSocket integration with efficient event broadcasting.
Major	User Interactions — Chat, user profiles, and friends system.
Minor	Health Check & Status Page — Service availability monitoring endpoint.

---

aait-bou

Type	Module
Minor	ORM Integration — Database access via Object-Relational Mapping.
Minor	Remote Authentication — OAuth 2.0 support (Google, GitHub, 42, etc.).

---

achahbal

Type	Module
Minor	Server-Side Rendering (SSR) — Improved performance and SEO.
Major	Custom Major Module — Substantial technical complexity; requires written justification.

---

aessadik

Type	Module
Major	WAF / ModSecurity + HashiCorp Vault — Strict WAF configuration with secure secrets management.
Major	Microservices Architecture — Backend decomposed into loosely-coupled, independently deployable services.
Major	Prometheus & Grafana — Full monitoring stack with metrics collection and alerting.

Custom Module (Bonus)

RIHLA-MOBILE - The AI Travel Explorer

---

🎖️ Major Module of Choice: AI Travel Orchestration Engine

Grade: Major Module (2 Points)

∗ Why I chose this module

Traditional travel applications often suffer from a disconnect between data discovery and logistical planning. Furthermore, modern users overwhelmingly prefer to plan and navigate their journeys using mobile devices while on the go. I chose to implement a custom AI Travel Orchestration Engine within a native mobile environment because it meets the user where they are: in the palm of their hand. This project bridges the gap, creating a frictionless transition from raw geospatial discovery to a mathematically logical multi-day itinerary, optimized specifically for the mobile-first traveler.

∗ What technical challenges it addresses

1. The Hallucination Problem: Large Language Models (LLMs) often generate non-existent locations or outdated addresses. My engine addresses this by implementing a Custom Prompt Bridge. Instead of allowing the AI to guess, the engine programmatically injects real-time, verified data from the Google Places API into the LLM context, forcing the AI to work strictly with physical, verifiable landmarks and coordinates.
2. Geospatial Logic: Calculating a logical travel sequence for a human (avoiding diagonal zig-zags across a city) requires complex heuristics. The orchestration engine solves this by using AI to group activities based on neighborhood density and logical time-of-day constraints (e.g., matching museum hours with morning slots and restaurants with evening slots).
3. Structured Data Parsing: Converting natural language AI thought processes into a strictly-typed JSON schema that a mobile UI can render without crashing is a significant technical hurdle. I implemented a robust parsing layer that enforces a specific schema on every generation.

∗ How it adds value to your project

It transforms RIHLA from a simple search directory into a dynamic personal travel assistant. By automating the planning phase, it saves users hours of research and ensures that their trips are geographically optimized. It takes the "guesswork" out of tourism, providing value that standard map applications currently lack.

∗ Why it deserves Major module status (2 points)

This module is substantial because it involves a complex multi-stage pipeline:

• Stage 1 (Live Data Acquisition): Asynchronous fetching of real-world geospatial datasets.
• Stage 2 (Contextual Injection): Transforming raw JSON into a natural language context for the LLM.
• Stage 3 (AI Orchestration): Using Gemini-2.5-Flash to solve the combinatorial problem of itinerary design.
• Stage 4 (Dynamic UI Rendering): Mapping a complex AI-generated payload into a vertical timeline interface with deep-linked native GPS actions. By integrating disparate state-of-the-art technologies (React Native + Google Maps + Google Generative AI) into one seamless system, this module demonstrates high technical complexity and creativity far beyond a standard feature.

---

⚙️ Tech Stack

• Framework: React Native / Expo (TS)
• Navigation: Native Stack
• APIs: Google Places, Google Gemini 2.5 Flash
• Target: iOS Only

npm run ios