How itachi-1824/jarvis works
Project type and purpose
• Jarvis is a real-time AI personal assistant written in Python. It provides voice/video interaction, handles weather lookups, web searches, and email management by orchestrating multiple external AI and communication services.
• It is primarily a single-process backend application with a real-time communication frontend (via LiveKit) and integrations with Google AI, DuckDuckGo, and Gmail.
File structure
• Top-level files:
• agent.py → Core runtime: connects to LiveKit, handles incoming voice/video, orchestrates commands.
• prompts.py → Defines instruction and personality prompts for the AI model.
• tools.py → Implements “tool” functions: weather API calls, search, email sending.
• .env.example → Template for API keys and secrets.
• requirements.txt → Lists Python dependencies (LiveKit, google-api, requests, dotenv).
• No separate “frontend” folder because the LiveKit client acts as the frontend; the application acts as a server and client in one.
README for additional insights
• Architecture is described in prose: LiveKit handles real-time voice/video streaming, Google AI powers speech recognition/synthesis and natural language understanding, DuckDuckGo and Gmail are used via REST.
• Environment variables configure connections to external services.
Detailed capture of architectural elements
• Module with key libraries (e.g., “livekit-sdk” in the LiveKit Connector).