Arthur AI

A voice-activated personal assistant for Windows, built in Python. Arthur listens for a wake word, recognises your spoken command, classifies the intent using a trained Naive Bayes model, and responds with both text output and synthesized speech.

---

Features

Capability	Example Commands
Q&A / Math (WolframAlpha)	"solve x² + 2x - 3 = 0", "what is 1 + 2"
Wikipedia search	"who is Amitabh Bachchan?", "who is Salman Khan?"
Weather	"how is the weather in Nagpur?", "is it cold outside?"
News headlines	"tell me the news", "play the current news"
Maps	"where is Denmark?", "show me the map of India"
Open websites	"open Twitter", "launch Instagram"
Play music	"play a song", "play some music"
Stop music	"stop the song", "pause music"
Tell the time	"tell me the time", "what time is it?"
Tell a joke	"tell me a joke", "make me laugh"
Take a screenshot	"take a screenshot", "capture the screen"
Type text	"type a message" (dictates text via pyautogui)
Shutdown PC	"shut down"
Greetings	"hi", "hello", "namaste"
What can you do?	"what can you do?"

---

Architecture

Arthur_AI/
├── main.py           # Entry point — wake word loop + intent dispatcher
├── Brain.py          # All capability implementations (weather, wiki, jokes, etc.)
├── sensors.py        # Speech input (STT) and speech output (TTS)
├── prediction.py     # Naive Bayes intent classifier (trained on intents.csv)
├── weather.py        # WeatherAPI wrapper — returns full weather + astronomy data
├── data/
│   ├── intents.csv   # Training data for the intent classifier
│   ├── responses.json# Randomized spoken responses per intent
│   ├── jokes.json    # Joke dataset (setup + punchline)
│   └── wakewords.txt # List of wake words that activate Arthur
└── requirements.txt  # Python dependencies

How it works:

1. sensors.py continuously listens via the microphone.
2. When a wake word from data/wakewords.txt is detected, Arthur activates.
3. The user's speech is transcribed using Google Speech Recognition.
4. prediction.py classifies the transcribed text into an intent using a TF-IDF + Multinomial Naive Bayes model trained on data/intents.csv.
5. main.py dispatches to the appropriate handler in Brain.py.
6. Arthur speaks the response back via pyttsx3.

---

Requirements

• Python 3.x
• Windows (recommended — voice engine uses Windows SAPI voices; some paths are Windows-specific)
• A working microphone
• Internet connection (for STT, WolframAlpha, Wikipedia, Weather, News, Maps)

API Keys Required

The following API keys are hardcoded in the source files. Replace them with your own before running:

Key	File	Where to get it
WolframAlpha App ID	Brain.py	developer.wolframalpha.com
WeatherAPI Key	weather.py	weatherapi.com
NewsAPI Key	Brain.py	newsapi.org

---

Setup

1. Clone the repository

git clone https://github.com/codyandersan/Arthur_AI.git
cd Arthur_AI

2. Install dependencies

pip install -r requirements.txt

Some NLTK packages are also required. Run once in a Python interpreter:

import nltk
nltk.download('punkt')
nltk.download('averaged_perceptron_tagger')
nltk.download('maxent_ne_chunker')
nltk.download('words')

3. Configure your setup

Songs folder — Update the song_folder path in both main.py and Brain.py:

song_folder = r"C:\Users\YourName\Music"  # Change this

Microphone index — If Arthur doesn't hear you, try changing device_index in sensors.py:

self.mic = sr.Microphone(device_index=0)  # Try 0 or 1

Voice selection — On Windows, self.voices[3].id selects a specific TTS voice. Change the index to match a voice installed on your system.

Create websites.json — This file is needed for the "open website" feature but is not included in the repo. Create it manually in the project root:

{}

Arthur will populate it automatically as you ask it to open websites.

4. Run Arthur

python main.py

Arthur will print "Hello Sir, What can I do for you?" and start listening for the wake word "arthur".

If Arthur hangs: Open Command Prompt and run taskkill /im python.exe /t /f

---

Adding Wake Words

Edit data/wakewords.txt and add one wake word per line:

arthur
hey arthur
assistant

---

Dependencies

Package	Purpose
SpeechRecognition	Microphone input and Google STT
pyttsx3	Text-to-speech output
pygame	Music playback
scikit-learn	Intent classification (Naive Bayes)
pandas	Loading training data CSV
nltk	Tokenization and named entity recognition
wikipedia	Wikipedia summaries
requests	HTTP calls (WolframAlpha, WeatherAPI)
newsapi-python	News headlines
google	Google search for website discovery
pyautogui	Screenshot capture and keyboard typing
numpy	scikit-learn dependency