centauri-os

name	centauri-os
description	Centauri OS — Wade's custom Android-based operating system with Caroline AI as its core agent. Use for: designing or developing Centauri OS features, defining voice command interfaces, integrating Caroline into Android system services, building the AI-Enhanced CodeAssist app, and planning the Centauri OS architecture.

Centauri OS Skill

Centauri OS is Wade's custom Android-based operating system with the Caroline AI as its core agent. It represents the convergence of all his AI and mobile development work into a single, deeply integrated platform.

Core Architecture

Layer	Component	Technology
AI Core	Caroline AI	Dolphin Mistral on RunPod
Voice Interface	STT/TTS	Faster-Whisper + ElevenLabs
OS Base	Android	Custom AOSP build
App Layer	Expo/React Native	SDK 54
Memory	Long-term context	Wade Global State
Security	Encrypted comms	TLS + local-first

Core Workflows

1. Voice Command Interface Design

Define and implement voice commands for the OS:

• Wake word: "Hey Caroline" triggers the AI assistant.
• System commands: "Caroline, call [client name]", "Caroline, what's my schedule today?"
• Construction commands: "Caroline, start the timer for tile work", "Caroline, add 2 hours to the bathroom project."

2. AI-Enhanced CodeAssist

The CodeAssist app is a key component of Centauri OS:

• Voice I/O: Dictate code and hear explanations read back.
• Knowledge Graph: Interconnected information about Wade's codebase.
• Autonomous Evolution: The AI learns from coding sessions and improves its suggestions.

Repository: tywade1980/ai-agent-android

3. System Integration Points

Integration	Description
Phone calls	Caroline intercepts and screens calls via the telephony skill
Calendar	Caroline reads and writes to Google Calendar
Location	Caroline infers work context from location data
Camera	Caroline analyzes job site photos for progress tracking
Notifications	Caroline filters and prioritizes notifications

Key References

• Full Design Proposal: See references/centauri_os_design.md for the complete architecture document.
• Voice Command Spec: See references/voice_commands.md for the full command vocabulary.

Best Practices

• Local-first: Prioritize on-device processing to minimize latency and protect privacy.
• Offline capable: Core functions must work without internet (RunPod may be offline on job sites).
• Battery aware: Voice processing is power-intensive; implement smart wake word detection.
• Caroline continuity: All OS interactions feed back into Caroline's personality evolution.