Skip to content

How Canoryn Works

Canoryn is built on a structured architecture that connects system inputs, AI reasoning, and native action execution.

1. Input & Context (Triggers)

Canoryn uses multimodal input to understand context:

  • Vision Loop: Periodically captures screen state (text, UI elements, images).
  • Audio Stream: Listens for wake words ("Hey Canoryn") and voice commands.
  • System Events: Monitors file changes, app launches, and hotkeys.
  • Clipboard: Watches for copied text or images.

How Canoryn processes workflows:

├─────────────────────────────────────────────────────────┤
│  INPUTS & TRIGGERS    AI REASONING      SYSTEM ACTIONS  │
│  ─────────────────    ────────────      ──────────────  │
│  See & Hear           Think & Decide    Do & Execute    │
└─────────────────────────────────────────────────────────┘

1. Inputs & Context

How Canoryn captures system events and user intent:

Input TypeImplementation
VisionScreen analysis, UI element detection
HearingVoice commands, audio context
ContextActive app, time, location

2. Reasoning & Logic (AI Processing)

How Canoryn processes instructions and decides next steps:

ComponentFunction
LLM ReasoningNatural language NLU
Blueprint LogicVisual workflow execution
Memory DatabaseRetrieval of relevant past context

3. Actions & Integrations (Execution)

How Canoryn runs workflows and controls system elements:

SystemCapabilities
macOS APIsApp control, file management
AccessibilityUI automation, clicks, typing
IntegrationsSpotify, Calendar, etc.

Data Flow

User speaks "Play something chill"


   ┌─────────┐
   │ TRIGGER │ ← Voice transcription
   └────┬────┘


   ┌─────────┐
   │ PROCESS │ ← LLM understands intent
   └────┬────┘   Memory recalls: "User likes Lo-fi"


   ┌─────────┐
   │ ACTION  │ ← Spotify: Play Lo-fi playlist
   └─────────┘

Local-First Architecture

Everything runs on your Mac:

  • LLM: Ollama (local) or cloud providers (your choice)
  • Memory: SQLite database, encrypted
  • Processing: Native Swift, no web views

Further Reading