AI Companion Web App (Frontend + Backend) – Using Existing Prompt & Persona System PROJECT OVERVIEW I am building an AI companion web app. The core AI prompt, persona rules, and behavior logic are already finalized (~90%). Your role is NOT to design AI behavior, but to implement the frontend and backend system that uses the provided prompt exactly as-is. Think of this as building the container, plumbing, and UI around an existing AI “brain”. SCOPE OF WORK (WHAT YOU WILL BUILD) FRONTEND (Web App) Clean, simple web UI Pages required: Login / Sign up page Chat page Persona selector (simple buttons, e.g. “Aisya”, “Nor”) Frontend requirements: Responsive Simple chat UI (like ChatGPT style) No complex animations needed Focus on usability AUTHENTICATION Google Login (OAuth) Email & password login (optional but preferred) Session handling User must stay logged in across refresh BACKEND Secure API endpoint for chat Backend calls OpenAI API (API key via ENV, NOT hardcoded) Backend assembles prompt before each AI call Prompt assembly order (IMPORTANT): Core system prompt (provided by me, do not edit) Persona prompt (based on user selection) Memory summary (from database) Latest user message PERSONA SYSTEM (IMPORTANT) Users can select a persona (e.g. “Aisya”, “Nor”) Persona selection should: Update a user field (e.g. persona_active) NOT trigger an AI call Persona only affects prompt when user sends a message Persona rules: Persona content will be provided by me Do NOT modify persona prompt text Just load the correct persona prompt dynamically MEMORY SYSTEM (SIMPLE) This is NOT advanced AI memory. Requirements: Database table for memory Each memory record contains: user_id type (preference / emotion) short text summary updated_at Rules: Store summary, not full chat logs Limit memory injection to max 3–5 items Memory logic can be basic (manual or rule-based) DATABASE Minimum required tables: Users Memories Optional: Conversation log (basic) TECH STACK (Preferred) You may propose alternatives, but preferred stack is: Frontend: Next.js (React) Backend: API routes / server functions Database: PostgreSQL / Supabase / Firebase Auth: Google OAuth Hosting: Vercel / equivalent If you recommend a different stack, explain WHY. IMPORTANT RULES (READ CAREFULLY) DO NOT: Edit, rewrite, or “improve” the AI prompt Design AI behavior or personality Auto-store full chat history as memory Hardcode API keys Make architectural decisions without discussion YOU MUST: Use prompt text exactly as provided Load prompts from files or constants Keep logic modular and readable Comment code where logic matters Deliver source code under my ownership DELIVERABLES Working web app (login → chat → persona switch) Source code (GitHub repository under my account) Environment variable setup instructions Simple README explaining: how to run locally how to deploy Deployed test version (optional but preferred) ACCEPTANCE CRITERIA Project is considered complete when: User can login with Google User can chat with AI AI responds using provided prompt Persona switching works correctly Memory is stored and injected properly No prompt editing by developer IDEAL DEVELOPER Comfortable with AI integrations Understands prompt-based systems Does not “freestyle” behavior logic Communicates clearly Will ask questions before making assumptions APPLICATION QUESTIONS (IMPORTANT) Please answer the following when applying: Have you built AI chat apps before? Share examples. Which stack do you propose and why? How would you implement persona switching without changing AI logic? Are you comfortable using an existing prompt without editing it? FINAL NOTE This project is about precision and discipline, not creativity in AI behavior. You are building the system, not the personality.