Procrasti-Hater

"I see you scrolling your phone instead of working."

Procrasti-Hater is an AI-powered surveillance agent that forces you to be productive. It monitors your screen process and webcam activity in real-time, detecting distractions like sleeping, using your phone, or launching games. When you slack off, a personalized AI persona (e.g., Gigachad, Gordon Ramsey) verbally roasts you.

🎥 Demo

🔥 Key Features

• 👀 Real-time Vision Monitoring: Detects if you are sleeping, looking at your phone, absent, or looking away (gazing elsewhere) using MediaPipe.
• 🖥️ Process & Window Tracking: Instantly detects distraction apps (Games, Netflix, Social Media) based on active window titles.
• 🗣️ Verbal Abuse (TTS): Uses ElevenLabs to generate voice responses. The AI interrupts you immediately when a violation is detected.
• 🎭 Multiple Personas: Choose your supervisor - form the supportive "Anime Girl" to the ruthless "Drill Sergeant" or "Gordon Ramsey".
• 🎙️ Interactive Excuses: You can talk back to the agent via microphone. The AI judges if your excuse is valid or scolds you harder (LLM-based).
• 📊 Session Report: At the end of a work session, get a ruthless "Session Review" with a score and a detailed breakdown of your sins.

🛠️ Tech Stack

Client

• Framework: Python, PyQt6
• Vision: MediaPipe (Face Landmarkers, Pose Detection)
• Audio: SoundDevice (Push-to-Talk)
• Communication: LiveKit Client SDK (Data Channels for real-time packets)
• UI Design: Custom "Pip-Boy" style retro interface.

Agent

• Framework: LiveKit Agents
• Intelligence: Google Gemini (Reasoning & Persona)
• Voice: ElevenLabs TTS (Streaming Audio)
• Logic: Asyncio-based event loop handling detection packets and managing conversation state.

📂 Project Structure

ProcrastiHater/
│
├── 📂 agent/                 # [Backend] The AI logic running on the server
│   ├── main.py               # Entry point. Handles LiveKit events & orchestration.
│   ├── llm.py                # Interacts with OpenAI for generating scolding text.
│   ├── memory.py             # Manages session history & cooldowns.
│   └── prompts.py            # System prompts defining the Personas.
│
├── 📂 client/                # [Frontend] The desktop application
│   ├── main.py               # Entry point. Launches UI & Background Services.
│   ├── config.py             # Configuration loader (.env).
│   │
│   ├── 📂 services/          # Background worker threads
│   │   ├── vision.py         # Webcam analysis (Sleep/Phone/Absence detection).
│   │   ├── screen.py         # Active window monitoring.
│   │   ├── livekit_client.py # Network storage for sending packets.
│   │   └── stats.py          # Local session statistics tracking.
│   │
│   └── 📂 ui/                # PyQt6 Widgets
│       ├── main_window.py    # Dashboard (Personality selection, Stats).
│       ├── floating_widget.py# Always-on-top character overlay.
│       └── stats_view.py     # End-of-session report card.
│
└── 📂 shared/                # Shared types & protocols
    ├── protocol.py           # JSON Packet structure.
    └── constants.py          # Event definitions (WINDOW_CHANGE, SLEEPING, etc).

🚀 Getting Started

Prerequisites

1. Python 3.10+
2. LiveKit Cloud Project (Url & Secret)
3. OpenAI API Key
4. ElevenLabs API Key

Installation

1. Clone the repository

   git clone https://github.com/yourusername/procrastihater.git
   cd procrastihater

2. Install dependencies

   pip install -r requirements.txt

3. Configure Environment Create a .env file in the root directory:

   LIVEKIT_URL=wss://your-project.livekit.cloud
   LIVEKIT_API_KEY=your_api_key
   LIVEKIT_API_SECRET=your_api_secret
   OPENAI_API_KEY=sk-proj-...
   ELEVEN_API_KEY=...

Usage

1. Start the Agent

   python agent/main.py start

   The agent will connect to the room and wait for a user.

2. Start the Client

   python client/main.py

   The GUI will launch. Select a persona and click START.

🎮 Controls

• Alt+S: Toggle Microphone (Push-to-Talk)
• Alt+B: Toggle Debug Window (View webcam feed & vision landmarks)

📜 License

MIT License