🚀 Interview Copilot

A role-based AI-powered interview practice platform that simulates real technical interviews with structured feedback.

---

✨ Overview

Interview Copilot helps developers practice interviews in a focused and structured way.

Instead of random AI chat, the system enforces:

• One question at a time
• Immediate structured feedback
• Controlled interview flow

Each response is evaluated with:

• Score
• Strengths
• Weaknesses
• Ideal Answer

---

🎯 Features

• 🎭 Role-based interviews (Frontend, Backend, DSA)
• 🧠 AI-generated questions & evaluation
• 📊 Real-time scoring and progress tracking
• ⏱️ Session timer
• 📱 Fully responsive UI
• 🔁 Controlled interview flow (fixed number of questions)
• 🧹 Cleaned and structured AI output (no messy formatting)

---

🏗️ Tech Stack

Frontend

• Next.js 15 (App Router)
• Tailwind CSS

Backend

• Next.js API Routes

AI

• Google Gemini API (@google/genai)

---

🧠 System Design

The application is built with a clear separation of concerns:

1. Frontend

• Handles UI rendering, chat state, and user interaction
• Parses and displays structured feedback

2. API Layer

• Controls interview flow (question → answer → feedback)
• Limits total number of questions
• Formats and validates AI responses

3. AI Layer

• Generates interview questions
• Evaluates answers based on structured prompts

⚠️ AI does NOT control application logic. The backend ensures deterministic behavior.

---

🛠️ Installation

1. Clone the repository

git clone https://github.com/krishna7054/chat-bot.git
cd chat-bot

2. Install dependencies

npm install

3. Setup environment variables

Create a .env.local file:

GEMINI_API_KEY=your_api_key_here

4. Run the development server

npm run dev

🤖 AI Handling Strategy (Important)

To avoid unreliable AI outputs ("AI slop"), the system follows strict principles:

✅ Structured Prompting

AI is forced to return responses in a fixed format:

• Score
• Strengths
• Weaknesses
• Ideal Answer
• Next Question

✅ Controlled Parsing

Responses are parsed and split into:

• Feedback
• Next Question

Only required data is extracted.

✅ Validation Layer

• Score extracted using regex
• Missing or malformed responses are handled safely

✅ Cleaning Layer

• Removes markdown artifacts (*, #, etc.)
• Ensures clean UI rendering

✅ Code-Controlled Logic

• Interview flow is NOT controlled by AI
• Backend enforces:
  • Question count
  • Interview completion
  • Response sequencing

💡 AI is treated as a tool, not a source of truth.

📊 Interview Flow

1. User selects a role
2. AI generates Question 1
3. User submits answer
4. AI evaluates response
5. Feedback is displayed
6. Next question is asked
7. Process repeats until completion