AI Voice-Controlled Desktop Assistant

A modern, cross-platform, AI-powered voice assistant for your desktop. Instantly open or close apps, files, folders, and websites using natural language voice commands. Combines local fuzzy search, LLM intent extraction, and real-time web search for a truly human, productive experience.

---

🚀 Key Features

• Voice-Activated: Control your desktop with your voice using advanced speech recognition.
• Open & Close Anything: Instantly open or close apps, files, folders, and Chrome tabs with natural language ("open terminal", "close downloads folder", "open kunaal updated resume file").
• AI-Powered Intent Extraction: Uses LLM (Mistral) for robust, context-aware command understanding.
• Fuzzy File Search: Finds files even with typos, word order changes, or partial names ("open ai internal 2 pdf file" → AI Internal - 2.pdf).
• Smart App & System App Handling: Recognizes and opens/closes system apps (Terminal, Finder, Settings, Activity Monitor, etc.) by name.
• Bulletproof Website Handling: Opens the correct website in Chrome using Google Custom Search API ("open cbit in chrome" → https://www.cbit.ac.in).
• Deep, Fast Indexing: Recursively indexes all user-facing folders (Downloads, Desktop, Documents, /Applications, etc.) for instant lookups.
• Natural Language: No rigid syntax—just speak as you would to a human assistant.
• Cross-Platform: macOS (full), Windows/Linux (open/close app logic can be extended).

---

🧠 Example Commands

• open terminal
• close terminal
• open downloads folder
• close downloads folder
• open kunaal updated resume file
• open ai internal 2 pdf file
• open gmail in chrome
• close gmail in chrome
• open notion
• open chatgpt
• open codechef
• open python file
• search for best pizza near me

---

🏆 Why This Project Stands Out

• True AI Desktop Agent: Combines LLM, fuzzy search, and real-time web search for human-like understanding and action.
• Enterprise-Ready: Handles ambiguous, typo-prone, and context-rich queries with ease.
• Productivity Booster: No more hunting for files, apps, or websites—just say what you want.
• Modern Stack: Python, SpeechRecognition, PyAudio, OpenAI/Mistral LLM, Google Custom Search API, AppleScript (macOS).
• Extensible: Easily add new commands, system integrations, or support for more platforms.

---

⚡️ Quick Start

1. Clone the Repository

   git clone https://github.com/Yuvakunaal/AI-Voice-Desktop-Assistant.git
   cd AI-Voice-Desktop-Assistant

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables:

   • Create a .env file with your API keys:

     OPENROUTER_API_KEY=your_openrouter_api_key_here
     GOOGLE_API_KEY=your_google_custom_search_api_key
     GOOGLE_CSE_ID=your_custom_search_engine_id

   🔑 How to Get OpenRouter API Key

   1. Go to https://openrouter.ai/.
   2. Log in with your account.
   3. Go to Settings.
   4. Navigate to Keys.
   5. Click on Create Key, give it a name.
   6. Copy the generated key.
   7. Paste it as the value of the OPENROUTER_API_KEY variable in your project (.env).

4. Set up Google Custom Search API:

   • Instructions here
   • Add your API key and CSE ID in commands/website_opener.py.

5. Run the assistant:

   python3 main.py

6. Speak your command!

---

🛠️ Tech Stack

• Python 3.8+
• SpeechRecognition, PyAudio
• Mistral LLM (via OpenRouter)
• Google Custom Search API
• AppleScript (macOS automation)
• Fuzzy matching, deep file indexing

🤖 AI-Powered Features

This project leverages advanced AI capabilities including:

• Natural Language Understanding: Mistral LLM processes voice commands with human-like comprehension.
• Intent Extraction: AI determines whether you want to open/close apps, files, folders, or websites.
• Smart Matching: Fuzzy search algorithms find the right files even with imperfect queries.
• Context Awareness: Understands relationships between words and resolves ambiguities.

🎯 How It Works

1. Voice Input: Speak your command naturally.
2. Speech Recognition: Converts speech to text using Google's speech recognition.
3. AI Intent Parsing: Mistral LLM analyzes your command and extracts intent + target.
4. Smart Search: Finds the best matching app, file, folder, or website.
5. Execution: Opens/closes the requested item using system automation.
6. Web Integration: Uses Google Custom Search to find correct websites for ambiguous queries.

🌟 Advanced Features

• Real-time Web Resolution: Finds correct websites even for obscure queries.
• Antonym Handling: Intelligently avoids conflicting file matches (e.g., "internal" vs "external").
• LLM Tie-Breaking: Uses AI to choose between multiple potential matches.
• System App Integration: Special handling for macOS system applications.

👨‍💻 Developer

Kunaal – GenAI, AI, Python Enthusiast

🙏 Acknowledgments

• OpenRouter for providing access to Mistral LLM.
• Google for speech recognition and custom search APIs.
• The open-source community for the libraries and tools used.

⭐ If you like this project, please support by starring the repository!