How adrianhajdin/ai_mock_interviews works
Project type and purpose
• Project type: Full-stack web application built with Next.js (React)
• Purpose: Prepwise – a job interview preparation platform.
– Frontend for UI/UX and pages.
– Backend logic via Next.js API routes.
– Uses AI voice agents (Vapi), Google Gemini for question generation and feedback.
– Firebase for authentication and data persistence.
File structure
• app/
– Next.js “app router” folder containing:
• (auth) routes for sign-in / sign-up
• (root) routes for dashboard and interview flows
• api/vapi/generate/route.ts – serverless endpoint for question generation
• components/ & components/ui/
– Reusable UI components (Agent, InterviewCard, FormField, shadcn UI wrappers)
• lib/
– actions/ (auth.action.ts, general.action.ts) – business logic for Firebase and AI calls
– utils.ts – styling helpers and tech-logo resolution
– vapi.sdk.ts – wrapper around Vapi AI service
• firebase/
– admin.ts & client.ts – Firebase Admin SDK for backend and client SDK for front end
• constants/, types/
– Shared types and configuration
• public/
• Configuration files
– next.config.ts, tsconfig.json, postcss.config.mjs
README for additional insights
• Tech stack: Next.js, Firebase (Auth, Firestore), Tailwind CSS, Vapi AI, Google Gemini, shadcn/ui, Zod
• Features:
– Authentication via Firebase
– Interview creation with AI-generated questions
– Real-time AI-driven voice interviews (Vapi)
– Feedback scoring via Google Gemini prompt in general.action.ts
Key reminders
• Dynamic routing: illustrate interview/[id] and feedback subpages
• Detail interactions in lib/ (auth.action.ts vs. general.action.ts) as separate flows
• Don’t overcrowd: group related pages under one “Next.js Frontend” cluster but call out critical API routes