AI Voice Interview Agent (Thrisha Karkera)

An interactive voice-based AI interview assistant that listens to interviewer questions and responds with intelligent answers in both text and speech.

This project combines speech recognition, predefined logic, and LLM fallback to simulate a structured technical interview.

---

🚀 Live Demo

👉 Deployed on Hugging Face: https://huggingface.co/spaces/Thrisha2005/voicebot

---

✨ Features

• 🎤 Voice Input — Ask questions using your microphone
• 🧠 Smart Question Matching — Detects intent using keyword-based matching
• 📚 Predefined Answers — High-quality, structured responses for common interview questions
• 🤖 LLM Fallback (Groq + LLaMA) — Handles unexpected questions dynamically
• 🔊 Voice Output (TTS) — Converts answers into speech using gTTS
• 💬 Text Output — Displays conversation clearly
• 🧾 Conversation Memory — Maintains chat history for context

---

🛠️ Tech Stack

• Python

• Gradio — UI interface

• Groq API

  • Whisper (whisper-large-v3) for speech-to-text
  • LLaMA (llama-3.3-70b-versatile) for fallback responses

• gTTS — Text-to-speech

• pydub — Audio processing

---

🧩 How It Works

1. 🎤 User records a question
2. 🔄 Audio is converted to .wav format
3. 🧠 Whisper (via Groq) transcribes speech → text
4. 🔍 System checks for a predefined answer using smart keyword matching
5. 🤖 If no match → fallback to LLaMA model
6. 🔊 Response is converted to speech (gTTS)
7. 💬 Output is shown as both text + audio

---

🧠 Smart Answer System

The assistant prioritizes scripted, high-quality answers for common interview questions like:

• "Tell me about yourself"
• "What are your strengths?"
• "Why should we hire you?"
• "What are your weaknesses?"
• "Where do you see yourself in 5 years?"

If a question doesn't match → it uses LLM fallback to generate a response.

---

📂 Project Structure

├── app.py              # Main application (Gradio UI + logic)
├── requirements.txt   # Dependencies
└── README.md          # Documentation

---

▶️ Run Locally

1. Clone the repository

git clone https://github.com/Thrisha200578/voicebot.git cd voicebot

2. Install dependencies

pip install -r requirements.txt

3. Set API Key

export GROQ_API_KEY=your_api_key_here

4. Run the app

python app.py

---

🔑 Environment Variables

Variable	Description
GROQ_API_KEY	API key for Groq

---

⚠️ Limitations

• Works best with clear voice input
• Scripted answers rely on keyword matching (not full NLP intent detection)
• Internet required for Groq API calls
• gTTS may introduce slight latency

---

🔮 Future Improvements

• ✅ Replace keyword matching with semantic search / embeddings
• ✅ Add multilingual support
• ✅ Improve voice quality with advanced TTS models
• ✅ Add real-time streaming responses
• ✅ Expand interview question coverage

---

📸 UI Preview

(Add screenshots from your Hugging Face app here)

---

🤝 Contributing

Contributions are welcome! Feel free to fork the repo and improve the system.

---

📜 License

MIT License

---

🙌 Acknowledgements

• Groq for ultra-fast inference
• Open-source speech & audio libraries
• Gradio for rapid UI development

---

👤 Author

Thrisha Karkera

---