🚀 AI Powered Interview Assistant

A Vite + React + Redux web app that simulates a timed technical interview.
It parses a candidate’s resume, generates tailored questions with Google Gemini, times and records answers, auto-scores each response, and produces a final AI summary & recommendation.

Stack: React, Redux Toolkit, Ant Design, Vite, (optional Tailwind classes for gradients), Google Gemini API.

---

✨ Features

• Resume-driven questions (6 total: 2 Easy, 2 Medium, 2 Hard)
• Per-question timers (auto-submit on timeout)
• Auto scoring using Gemini (0–15 with sub-rubrics)
• Final AI summary & recommendation
• Session persistence (Redux + Persist)
• Cross-tab sync (keep UI state consistent across tabs)
• Robust Gemini error handling (targeted retries for 429/503 + safe fallbacks)
• Dark UI theme with subtle neon accents (fully customizable in App.css)
• Deploy ready (Vercel / Netlify guides included)

---

🔐 Environment Variables

This project uses Vite. All variables must be prefixed with VITE_.

Setup

1. Copy the example file:

cp .env.example .env

2. Open .env and add your Gemini API key:

VITE_GEMINI_API_KEY=your_google_gemini_api_key_here

⚠️ Frontend-only apps expose API keys to users. For production, consider proxying Gemini calls through a small server (or Vercel/Netlify serverless function) to keep the key private and enforce your own rate limits.

---

🧑‍💻 Local Development

# 1) Clone the repository
git clone https://github.com/Sarg3n7/AI-Powered-Interview-Assitant.git

# 2) Install deps
npm install

# 3) Copy .env.example → .env and add your key
cp .env.example .env

# 4) Run dev server
npm run dev

# 5) Build / Preview
npm run build
npm run preview

• Vite serves at http://localhost:5173 by default.

---

🧠 How AI Calls Work

All Gemini interactions live in src/services/geminiService.js:

• generateQuestions(resumeText, candidateName)
  Returns exactly 6 questions (with difficulty & time limit). If parsing fails, it falls back to deterministic, sensible defaults.

• scoreAnswer(question, answer)
  Requests a JSON payload with a total score (0–15) and sub-scores: correctness (0–5), completeness (0–3), clarity (0–2), problemSolving (0–3), codeQuality (0–2). Safe parsing with a fallback heuristic if needed.

• createSummary(candidateData, sessionData)
  Builds a final professional summary + recommendation + strengths + improvements. Again, safe parsing + fallback.

Targeted Retries

safeGenerate(prompt, context) wraps Gemini calls and:

• Retries on 429 (rate limit) and 503 (service unavailable)
• Parses server-provided "retryDelay" or "Please retry in Xs" when available
• Uses capped exponential backoff with jitter otherwise
• Logs clean, helpful console messages

This is why you might still see 429/503 in the console occasionally—it’s normal and handled automatically. UI remains smooth.

---

🧭 UI Notes

• The top bar, tabs, cards, and chat follow a dark theme with neon accents.
• Scrollbars and overflow behavior in InterviewChat.jsx are tuned so the question block and message list each have their own scroll containers (ChatGPT-like), preventing layout jumps when alerts appear.

You can tweak colors in App.css.

---

🧪 Scripts

// package.json (relevant bits)
{
  "scripts": {
    "dev": "vite",
    "build": "tsc -b || true && vite build",   // tsc is optional; ignore if not using TS
    "preview": "vite preview"
  }
}

---

☁️ Deploy

Option A — Vercel (recommended for Vite)

1. Push your repo to GitHub.
2. Import to Vercel → Project Settings:
   • Framework Preset: Vite
   • Build Command: npm run build
   • Output Directory: dist
3. Add Environment Variable:
   • VITE_GEMINI_API_KEY → your key (apply to Production & Preview)
4. Deploy.

Option B — Netlify

Site Settings → Build & deploy → Build settings

• Base directory: . (or your subfolder)
• Build command: npm run build
• Publish directory: dist
• Environment variables: VITE_GEMINI_API_KEY
• Click Deploy site.

---

🛟 Troubleshooting

I see 429/503 in console logs.
That’s normal under load. The app retries with backoff. If it persists:

• Ensure your Gemini quota is sufficient.
• Reduce parallel requests (one question at a time already helps).
• Consider a backend proxy to centralize rate-limiting.

“Missing Gemini API key” error.
Set VITE_GEMINI_API_KEY in your .env (local) or in Vercel/Netlify project settings (cloud). Rebuild after setting.

White screen after deploy

• Confirm Output Directory is dist.
• Don’t set CRA-style homepage in package.json.
• Check the browser console for missing env var.

Selectors warnings (React-Redux)
If you see “selector returned a different result” in dev tools, memoize selectors using createSelector or return stable references.

---

📄 License

MIT © Sarg3n7