Build a production-ready full stack mobile-first app called fricker — the friend picker.

Core Concept: A social-lite layer for real friends: find people by @username, send mutual friend requests, then pick who to see next. Swipe or browse AI-suggested activities, send a plan, and your friend can accept or decline.

Public profiles are minimal (display name, avatar, @username) for discovery; richer hangout preferences stay private until you connect.

Platforms:

1. React Native app (primary platform)
2. Optional Next.js web dashboard
3. Shared backend API

Preferred Stack:

• React Native + Expo + TypeScript
• Next.js + TypeScript + Tailwind
• Express.js + TypeScript
• Firebase Auth
• Firestore
• Firebase Storage
• Zod validation
• Zustand or Context
• Clean scalable folder structure

Core AI Requirement: Create an abstraction layer so the app can connect to any AI provider later.

Supported providers should be easy to swap:

• LM Studio local models
• OpenAI
• Anthropic
• Ollama
• Custom REST endpoints
• Future providers

Use provider adapter pattern.

Example:

AiProvider interface:

• generateSuggestions(friendProfile): Promise<Suggestion[]>

Implement providers:

• LMStudioProvider
• OpenAIProvider
• AnthropicProvider
• MockProvider

Backend chooses provider from config/env.

Main Features:

Authentication:

• Sign up
• Login
• Logout
• Persistent session

Friend Records (private only): Users manually store their real-life friends.

Fields:

• name
• nickname
• photo
• phone number
• address / suburb
• hobbies
• favourite foods
• favourite activities
• personality notes
• budget level
• availability notes
• tags
• birthday (optional)

Swipe Flow:

Main screen shows friend cards.

Swipe Left:

• skip for now

Swipe Right:

• want to hang out

After swipe right: Generate 3 personalized suggestions using AI.

Example outputs:

• Coffee in Carlton then bookstore visit
• Basketball at local courts then burgers
• Beach drive and lunch
• Trivia night nearby
• Arcade then ramen

Suggestions Screen:

Each suggestion should include:

• title
• short reason why it matches friend
• estimated cost
• estimated duration
• optional nearby place ideas

Actions:

• Save suggestion
• Generate more
• Copy invite text
• Open SMS
• Mark completed

History: Track completed hangouts.

Fields:

• friend
• activity
• date
• notes
• rating

Firestore Structure:

users/{userId} users/{userId}/friends/{friendId} users/{userId}/hangoutHistory/{historyId} users/{userId}/savedSuggestions/{suggestionId} users/{userId}/settings/app

API Requirements:

REST API with Express.

Routes example:

POST /api/auth/signup POST /api/auth/login

GET /api/friends POST /api/friends PUT /api/friends/:id DELETE /api/friends/:id

POST /api/suggestions/:friendId POST /api/history GET /api/history

AI Architecture:

Create:

/services/ai/provider.interface.ts /services/ai/lmstudio.provider.ts /services/ai/openai.provider.ts /services/ai/anthropic.provider.ts /services/ai/mock.provider.ts

LM Studio provider should support local endpoint such as: http://127.0.0.1:1234/v1

Frontend calls backend only. Backend talks to AI provider.

UI Requirements:

• Modern polished UI
• Mobile-first
• Swipeable animated cards
• Smooth transitions
• Reusable components
• Dark mode
• Great empty/loading states

Deliverables:

1. Full project folder structure
2. Firestore schema
3. Backend route structure
4. AI provider adapter implementation
5. React Native screens
6. Swipe card implementation plan
7. Step-by-step build order

Start by generating the complete architecture and folder structure first.