🤖 Jarvis — AI Voice Assistant

A desktop AI voice assistant built from scratch in Python, with a dark, HUD-style interface inspired by Iron Man's J.A.R.V.I.S. Runs entirely on free/local tools — no paid APIs required.

---

✨ Features

Jarvis understands both typed and spoken commands and can:

• 💬 Chat naturally — powered by a local LLM (Ollama, llama3.2), so conversations stay private and free
• 🎙️ Listen and respond via voice (speech-to-text + text-to-speech)
• 🌦️ Check the weather for any city (Open-Meteo API, no key needed)
• 📰 Read news headlines (BBC RSS feed)
• 📖 Answer "who is / what is" questions via live Wikipedia summaries
• 🕐 Tell the time and date
• 🗺️ Open Google Maps for a place or get directions
• 📸 Take and save screenshots on command
• 📊 Report system info — CPU, RAM, and battery status
• 😂 Tell a joke
• 🚀 Launch apps and sites — Chrome, VS Code, Notepad, YouTube, GitHub, Gmail, Reddit
• 🧠 Know who its developer is — a small personality touch, aware it was built by its creator

---

🛠️ Tech Stack

Purpose	Library
LLM / chat	ollama (llama3.2, local)
GUI	customtkinter
Speech-to-text	SpeechRecognition + PyAudio
Text-to-speech	pyttsx3
Weather	Open-Meteo REST API
News	BBC RSS feed
Knowledge lookup	Wikipedia REST API
System info	psutil
Screenshots	pyautogui
Jokes	pyjokes

---

⚙️ Setup

Prerequisites

• Python 3.12+
• Ollama installed and running, with the llama3.2 model pulled:

  ollama pull llama3.2

Install

git clone https://github.com/<your-username>/jarvis.git
cd jarvis
python -m venv venv
venv\Scripts\activate        # Windows
pip install customtkinter ollama pyttsx3 SpeechRecognition pyaudio requests pyautogui psutil pyjokes

Run

python jarvis_gui.py

Type a message, or hit 🎤 Talk to speak to Jarvis directly.

Edit the DEVELOPER_NAME and DEFAULT_CITY constants at the top of jarvis_gui.py to personalize it.

---

🐞 Notable Debugging & Fixes

Building this surfaced some real Windows audio quirks worth documenting:

• WASAPI host API detection — PyAudio needed explicit WASAPI host lookup to correctly find the default input device and its native sample rate on Windows with Realtek drivers, instead of assuming device index 0.
• pyttsx3 engine reuse bug — reinitializing the TTS engine (pyttsx3.init()) fresh on every speak() call avoided a bug where the engine would silently stop working after the first few calls.
• Ollama PATH issues — Ollama needed to be running as a background service and accessible on PATH before the Python client could connect; connection failures are now caught with a friendly fallback message.

Included test scripts (test_wasapi.py, test_all_devices.py, test_mic.py, test_pyaudio.py, test_stream.py, test_record_and_save.py) were used along the way to isolate and confirm the correct working microphone and sample rate.

---

📁 Project Structure

jarvis/
├── jarvis_gui.py          # main application
├── test_wasapi.py         # WASAPI host API diagnostics
├── test_all_devices.py    # scan all input-capable devices
├── test_mic.py            # basic mic + recognition test
├── test_pyaudio.py        # list all PyAudio devices
├── test_stream.py         # raw stream open/read test
├── test_record_and_save.py# record + save + transcribe test
└── README.md

---

🔭 Roadmap / Ideas

• Wake-word activation ("Hey Jarvis")
• Persistent conversation memory across sessions
• Package as a standalone .exe for easier distribution
• Add more skills (reminders, calendar, smart home)

---

📄 License

MIT — feel free to fork and build on it.