Interview Copilot

Real-time interview assistant with speech recognition and AI-powered suggestions. Captures system audio, transcribes it locally with Whisper, and provides intelligent responses via AI.

Features

• Real-time audio capture - Captures system audio (microphone or loopback)
• Local speech-to-text - Whisper.cpp runs entirely on your machine (no cloud costs)
• Voice Activity Detection - Automatically detects speech segments
• AI-powered suggestions - Sends transcribed questions to an AI API for suggested responses
• Terminal UI - Clean terminal interface with transcript and suggestions panels
• Cross-platform - Works on Linux, macOS, and Windows

Installation

Prerequisites

• Rust (1.70+)
• Cargo
• C++ compiler (for building whisper.cpp)
• CMake

Linux Dependencies (Arch/EndeavourOS)

sudo pacman -S base-devel cmake alsa-lib

macOS

xcode-select --install

Windows

• Visual Studio Build Tools with C++ support
• CMake

Build

git clone https://github.com/vinycalves/interview-copilot.git
cd interview-copilot
cargo build --release

Quick Start

1. Download a Whisper model:

   cargo run --release -- download-model medium

   Choose from: tiny, base, small, medium, large-v3

2. List available audio devices:

   cargo run --release -- list-devices

   Select a [LOOPBACK] device to capture system audio, or a microphone for regular input.

3. Set your AI API key:

   export OPENAI_API_KEY="your-key-here"

4. Run:

   cargo run --release

Usage

TUI Controls

Key	Action
Space	Pause/Resume audio capture
R	Clear transcript
Q	Quit
D	Download a model (from within the app)

Configuration

Config file: ~/.config/interview-copilot/config.toml

[general]
language = "en"
model = "medium"

[audio]
device = "auto"
silence_threshold = 0.02
silence_duration_ms = 1500

[ai]
api_url = "https://api.openai.com/v1/chat/completions"
api_key = ""
model = "gpt-4"

Audio Setup

To capture system audio (what you hear from the interviewer):

Linux (PipeWire/PulseAudio):

• Loopback devices are listed as "[LOOPBACK]" in list-devices
• Select the monitor of your output device

macOS:

• Install BlackHole virtual audio driver
• Route system audio through it

Windows:

• WASAPI loopback devices are detected automatically

Architecture

Audio Input → VAD → Whisper.cpp → AI API → Terminal UI
  (cpal)           (local)                 (ratatui)

• cpal: Cross-platform audio capture library
• whisper-rs: Rust bindings for whisper.cpp (local transcription)
• reqwest: HTTP client for AI API calls
• ratatui: Terminal UI framework

License

MIT