The Interviewer (PrepWise)

AI-powered mock interview platform built with Next.js, Firebase, Gemini, and Vapi voice agents.

Overview

PrepWise helps users practice interviews by generating role-specific interview questions, conducting real-time voice interview sessions, and producing structured feedback with scoring.

Core flow:

1. User signs up/signs in with Firebase Auth
2. User generates an interview set from role/level/tech stack inputs
3. Voice interview runs through Vapi
4. Transcript is analyzed with Gemini
5. Feedback is saved and displayed with scores + recommendations

Features

• Email/password authentication with secure server session cookies
• Interview question generation via Gemini (/api/vapi/generate)
• Real-time voice interview sessions using Vapi Web SDK
• Transcript-based feedback scoring and analysis
• Dashboard with user interviews and latest interviews
• Tech stack logo rendering and role-based interview cards

Tech Stack

• Next.js 15 (App Router)
• React 19 + TypeScript
• Tailwind CSS
• Firebase Auth + Firestore
• Firebase Admin SDK (server actions)
• Vapi Voice AI (@vapi-ai/web)
• AI SDK + Google Gemini (@ai-sdk/google)
• Zod + React Hook Form

Project Structure

The-Interviewer/
	app/
		(auth)/
		(root)/
		api/vapi/generate/route.ts
	components/
		Agent.tsx
		AuthForm.tsx
		InterviewCard.tsx
		ui/
	firebase/
		client.ts
		admin.ts
	lib/
		actions/
			auth.action.ts
			general.action.ts
		vapi.sdk.ts
		utils.ts
	constants/
		index.ts
	types/
	public/
	package.json

Environment Variables

Create a .env.local file:

# Firebase Client SDK
NEXT_PUBLIC_FIREBASE_API_KEY=
NEXT_PUBLIC_FIREBASE_AUTH_DOMAIN=
NEXT_PUBLIC_FIREBASE_PROJECT_ID=
NEXT_PUBLIC_FIREBASE_STORAGE_BUCKET=
NEXT_PUBLIC_FIREBASE_MESSAGING_SENDER_ID=
NEXT_PUBLIC_FIREBASE_APP_ID=
NEXT_PUBLIC_FIREBASE_MEASUREMENT_ID=

# Firebase Admin SDK
FIREBASE_PROJECT_ID=
FIREBASE_CLIENT_EMAIL=
FIREBASE_PRIVATE_KEY=

# AI + Voice
GOOGLE_GENERATIVE_AI_API_KEY=
NEXT_PUBLIC_VAPI_WEB_TOKEN=
NEXT_PUBLIC_VAPI_WORKFLOW_ID=

Notes:

• FIREBASE_PRIVATE_KEY must preserve newlines. If storing as one line, use \n and code will convert it.
• Keep all secret values out of version control.

Installation

npm install

Run Locally

npm run dev

Open http://localhost:3000.

Available Scripts

• npm run dev - start development server
• npm run build - production build
• npm run start - run production server
• npm run lint - lint project

Key Routes

• /sign-in - sign in page
• /sign-up - registration page
• / - dashboard with interview lists
• /interview - interview generation voice flow
• /interview/[id] - interview session page
• /interview/[id]/feedback - interview feedback page
• /api/vapi/generate - API route to generate interview questions and persist interview record

Data Collections (Firestore)

• users
  • name, email
• interviews
  • role, type, level, techstack, questions, userId, finalized, coverImage, createdAt
• feedback
  • interviewId, userId, totalScore, categoryScores, strengths, areasForImprovement, finalAssessment, createdAt

Feedback Evaluation Categories

The feedback model scores:

1. Communication Skills
2. Technical Knowledge
3. Problem Solving
4. Cultural Fit
5. Confidence and Clarity

Authentication Flow

• Client auth uses Firebase client SDK
• Server session cookie is created via Firebase Admin (setSessionCookie)
• Protected routes are enforced in layout files:
  • root routes require authentication
  • auth routes redirect authenticated users to dashboard

Deployment

• Vercel config exists in vercel.json
• Ensure all required env vars are configured in deployment settings

Author

Suprit Raj