Interview AI

Table of Contents

1. Introduction
2. Features
3. Project Architecture
4. Installation
5. Environment Variables
6. Usage
7. Example
8. Tech Stack

---

Introduction

Interview AI is a tool designed to aid interviewers by dynamically generating relevant follow-up questions based on the candidate's responses. This project ensures that the interviewer never runs out of insightful questions, keeping the interview flowing smoothly and enhancing its depth.

Features

• Real-Time Follow-Up Questions: Using AI-powered models, this app listens to interview responses, converts them from speech-to-text, and suggests contextually relevant follow-up questions in real-time.
• Large Language Model Integration: Hosts an LLM for text processing, accessible via a Docker container.
• Streamlit Interface: Interviewers can view generated questions and other insights in an intuitive Streamlit dashboard.
• Docker Deployment: Containerizes the model and deploys it using Docker.

Project Architecture

• Speech-to-Text Conversion: Captures the candidate’s spoken answers and converts them into text format.
• Vector Database: Stores candidate responses and retrieves similar past responses to generate follow-up questions.
• LLM Model: Processes responses and suggests relevant questions based on context.
• Streamlit Interface: Displays the generated questions to the interviewer in real-time.
• S3 Storage: Used to store audio as text data as part of the interview logs.

Installation

1. Clone the repository:

   git clone https://github.com/UBH-Fall-2024/ub-hacking-create-your-repo-here-transformers.git
   

2. Set up the Docker environment:

   docker build -t speech-to-text
   

3. Start the Docker container:

   docker run -p 8000:8000 speech-to-text
   

4. Install dependencies for local Python development (optional):

   pip install -r requirements.txt
   

Environment Variables

• MODEL_PATH: Path to the Vosk model for speech-to-text recognition.
• LOG_PATH: Directory path for logging the fails index.
• Update these in a .env file or within Docker environment variables as needed.

Usage

• Start the Service: Run the Docker container as described in the Installation section.
• Make Requests: Access the service API at http://localhost:8000 for speech-to-text transcription.

Example

• Example transcription request and response: curl http://localhost:8000/ask

Tech Stack

• Speech-to-Text: Vosk
• Vector Database: FAISS
• Large Language Model: Llama-3.2-3B-Instruct
• UI Framework: Streamlit
• Cloud Storage: Amazon S3
• Containerization: Docker