AI Assistant - A Modular, voice-controlled system 🚀

-I don't click, I command

This is a voice-activated AI Assistant that listens to your commands, understands them with GPT, executes them using Playwright, and responds back to you - like a real assistant should. It's not Alexa, It's not Siri. It's yours. Built with open tools. Runs on locally. Works on your voice.

✨Features

~🔎 Wake word detection via Porcupine
~🗣 Speech to text using OpenAI Whisper
~❝ ❞ Command parsing via GPT (openAI API)
~🦾 Automation using Playwright(open browser, search, scrap, etc.)
~📖 Text to speech output via Edge TTS
~📏 Modular structure: easy plugin new commands, functions
~🌐 Local-Only design -no internet dependency except GPT(parsing)

🤖Tech Stack:-

🛠️ Component - Tool

1. Wake word - Porcupine(custom keyword model)
2. Speech recognition - OpenAI Whisper(local model "small/medium")
3. Natural Language Parsing - OpenAI GPT API
4. Automation - Playwrigth
5. Text-to-speech - Edge TTS
6. Runtime - Python 3.10+

📂Folder Structure:-

|-main.py --> entry point, connects
|-wakeword.py --> porcupine
|-listener.py --> Whisper + parser
|-automate.py --> playwright
|-requirements.txt --> all the required libraries to run the program

👨🏻‍💻How it works:

1. Wake word detect via porcupine custom model
2. Your voice is recorded and passed to whisper
3. Whisper transcribes the voice to text
4. GPT AI parses the text and determines the action
5. Playwright runs the automation task
6. Result is converted into Speech by Edge TTS
7. Assistant responds - like a friend who codes.

🛑KNOWN LIMITATIONS:-

1. Currently optimized for single microphone model
2. Mic detection/quality varies across systems
3. GPT API call requires internet connection(for command parsing)
4. No GUI yet - purely CLI and audio-driven -I focused on building intelligence, not appearance - for now

📈Future plans:-

1. Add local NLP fallback to replace GPT
2. Add GUI with Tkinter or React dashboard(post MERN)
3. Add multi mic/device compatibility
4. Plugin-style commands(e.g., emails, music)
5. Dockerize the assistant for deployment

🧗‍♂️Why this exists:-

--> I built this project not as a clone of something else, but as a milestone of my capability. --> I wanted to see if i could connect multiple systems - speech, logic, automation and voice - into a single, usable assistant.
--> AND YES I DID IT 💪🏻

🪪LICENSE

Personal use. Modify. Improve. Break it. Learn from it. This project was made to learn, not to sell.

📋Credits

1. OpenAI Whisper & GPT
2. Porcupine (by Picovoice)
3. Microsoft Edge TTS
4. Playwright
5. Priyanshu