Advanced AI assistant system integrating multiple language models and API routes (Groq, Cohere, Hugging Face, Gemini demo endpoint) with voice recognition, natural language processing, task automation, real-time web search, image generation, and intelligent decision-making. Features speech-to-text, text-to-speech, and comprehensive system control.
Inspired by the fictional JARVIS from Iron Man, this project implements a real-world AI assistant capable of understanding voice commands, executing system tasks, generating content, searching the web, creating images, and maintaining natural conversations through advanced language models.
Integrates three desktop AI systems plus a secured Gemini-backed browser demo route: Groq's Llama3-70B for conversational intelligence and real-time search, Cohere's Command-R-Plus for decision-making and query classification, Hugging Face's Stable Diffusion XL for desktop image generation, and Gemini for hosted portfolio-demo answers and image requests.
Voice recognition with multi-language support, natural language understanding, automated task execution (app control, web browsing, YouTube playback), content generation (essays, code, letters), real-time information retrieval, and AI-powered image creation.
Experience a web-based demonstration with text answers, live Sydney weather, voice input, spoken replies, and Gemini image request support.
AI/ML: groq, cohere, requests, Gemini API route
Voice: edge-tts, pygame, mtranslate
Web: selenium, webdriver-manager
Automation: AppOpener, pywhatkit, keyboard
Parsing: beautifulsoup4, googlesearch-python
Async: asyncio (built-in)
Utils: python-dotenv, Pillow, rich
Groq Cloud: Llama3-70B-8192, Mixtral-8x7B-32768
Cohere: Command-R-Plus (decision-making)
Hugging Face: Stable Diffusion XL Base 1.0
Gemini: Hosted browser demo answers and image requests
Google Search: Real-time information
Microsoft Edge TTS: Natural voice synthesis
Web Speech API: Browser-based STT
Architecture: Modular, async-capable
Storage: JSON chat logs, file-based state
Security: .env for API keys
Error Handling: Auto-recovery mechanisms
Performance: Parallel task execution
Scalability: Streaming responses
Successfully integrated multiple AI systems and routes (Groq, Cohere, Hugging Face, Gemini) into a unified architecture. Implemented intelligent routing between models based on query type, leveraging each AI's strengths: Llama3 for conversation, Command-R-Plus for decision-making, Stable Diffusion for desktop image generation, and Gemini for the hosted browser demo.
Developed comprehensive voice I/O system combining Web Speech API for recognition with edge-tts for natural synthesis. Supports multi-language input with automatic translation, intelligent query formatting, and context-aware voice responses with smart truncation for long-form content.
Built robust automation framework capable of controlling system applications, managing web browsers, playing YouTube content, generating professional documents, and executing system commands. Implements parallel task execution using AsyncIO for efficient multi-task handling.
Natural language conversations with context retention. Can discuss topics, answer questions, provide recommendations, and maintain multi-turn dialogues with persistent chat history.
Real-time web search with AI-powered synthesis. Fetches current information from Google, processes multiple sources, and generates comprehensive answers with proper attribution.
Generate text documents (essays, applications, letters, code) and AI images. Text generation uses Mixtral-8x7B for high-quality content, saved automatically to Notepad. The desktop image module produces Stable Diffusion XL variations, while the hosted demo can render Gemini image responses when enabled.
Voice-controlled system automation. Open/close applications, control YouTube playback, manage volume, execute Google searches, and navigate the web—all through natural voice commands.