How does the LLM routing system improve performance?

Ky routes trivial tasks to local models like tinyllama for 50ms responses, simple questions to Mistral, and only uses Claude for complex analysis, saving API costs and reducing latency.

Can I use Ky Assistant hands-free?

Yes, it includes the Kymera Voice layer with wake word detection, streaming text-to-speech, and echo cancellation for a full voice-interactive experience.

What are the hardware requirements for this skill?

To run locally, you need Python 3.11+, Ollama for local LLMs, and Docker to manage the Neo4j and Qdrant databases used for memory.

What makes the Ky memory system unique?

It uses the World Weaver tri-mode memory architecture, which stores experiences (episodic), knowledge graphs (semantic), and learned patterns (procedural) for high-fidelity context.

Ky AI Voice Assistant

Name: Ky AI Voice Assistant
Author: astoreyai

byastoreyai

•

生産性とワークフロー

Powers a Jarvis-like personal AI assistant with voice interaction, persistent memory, and intelligent task orchestration.

概要

Ky is a comprehensive AI orchestration layer that transforms Claude into a proactive personal assistant through the 'Closure' brain architecture. It intelligently routes tasks between local LLMs for low-latency responses and Claude for complex reasoning, while maintaining a tri-mode memory system (World Weaver) for long-term context. With its Kymera voice layer providing STT/TTS capabilities and deep integrations into calendars and emails, this skill provides a centralized command center for users seeking an automated, context-aware digital partner.

主な機能

Tri-mode memory system (World Weaver) covering episodic, semantic, and procedural knowledge
1 GitHub stars
Intelligent LLM routing for ultra-low latency local responses and high-reasoning Claude tasks
Kymera Voice layer featuring wake word detection, streaming TTS, and GPU-accelerated STT
Proactive task management including automated morning briefings and skill learning
Resilient system architecture with circuit breakers and automated health monitoring

ユースケース

Automating complex multi-step workflows across calendar, email, and local services
Creating a hands-free, voice-controlled environment for development and task management
Building a persistent AI knowledge base that remembers past interactions and preferences

概要

主な機能

Tri-mode memory system (World Weaver) covering episodic, semantic, and procedural knowledge
1 GitHub stars
Intelligent LLM routing for ultra-low latency local responses and high-reasoning Claude tasks
Kymera Voice layer featuring wake word detection, streaming TTS, and GPU-accelerated STT
Proactive task management including automated morning briefings and skill learning
Resilient system architecture with circuit breakers and automated health monitoring

ユースケース

Automating complex multi-step workflows across calendar, email, and local services
Creating a hands-free, voice-controlled environment for development and task management
Building a persistent AI knowledge base that remembers past interactions and preferences