SYSTEM OVERVIEW.
IRIS is a dual-platform voice-first execution engine available for both Desktop (IRIS-AI) and Mobile (IRIS-MX). Utilizing low-latency Gemini 3.1 Live API integration with bidirectional WebRTC/PCM audio, IRIS translates natural spoken intent directly into native system actions, app intents, and hardware controls.
What is Voice-First?
Traditional AI models are text-first: you type, wait, read. IRIS operates bidirectionally with real-time WebRTC audio streaming, bringing latency under 500ms. Speak naturally, interrupt anytime—IRIS listens, thinks, and executes dynamically.
What makes IRIS different?
- •Proprietary Orchestration: Protected, high-performance agent loops utilizing LangGraph state machines.
- •Pure Local Execution: Unlike web wrappers, IRIS executes CLI commands, manipulates desktop windows, and operates hardware.
- •System-Level Access: Sandboxed but deep access to directories, galleries, active processes, and ADB bridges.
Open Core Architecture
IRIS follows an open-core licensing model. The public repository controls the frontend shell, electron layout, and standard UI widgets. The core voice engine, neural orchestration loops, and system-level actions are packaged as protected main process modules to secure intellectual property.
Daily Usage Limits & Resets
By default, the Free Tier provides 10 AI Turns and 5 Tool Executions. (Note: 1 turn = 1 user input + 1 AI output that makes a proper conversational turn). To ensure fair usage across our global infrastructure, these limits are strictly synchronized and automatically reset every day between 12:00 PM - 2:00 PM, regardless of when you joined or used the application. Upgrading to the Pro Tier removes all limits entirely, granting you unlimited, unrestricted access to the engine at any time.