← Yumu Hub
Home LLM Architecture
Kevin's home AI infrastructure — April 2026
Access Methods
💬
iMessage
Direct messaging via BlueBubbles bridge
Native
🌐
Web Chat
LLM web UI + OpenClaw dashboard
CloudFlare Tunnel
🏠
Local Network
Direct LAN access to all services
LAN
Hosts
💻
MacBook Air
Online
Chip
Apple M2 Air
Memory
16 GB Unified
Storage
512 GB SSD
🦞
OpenClaw Gateway
AI agent — sessions, tools, skills
:8080
🏠
Yumu Hub
Local web apps — benchmark, games, tools
:8888
🖥️
Desktop
Online
CPU
Ryzen 9 3900X
Mobo
MSI X570 Gaming Edge WiFi
RAM
4×8 GB DDR4 Trident Z RGB
GPU 1
RTX 5070 Ti (x16)
GPU 2
RTX 5070 (x4)
NVMe
Corsair MP600 1 TB
SSD
WD 500 GB
🦙
Llama Server
Local LLM inference engine
Email OTP
API Token
:8080
📊
Metrics Server
GPU, token throughput, request stats
:8082
Network Connections
iMessage → OpenClaw
BlueBubbles bridge — iMessage ↔ OpenClaw Gateway
BlueBubbles → :8080
CloudFlare Tunnel → OpenClaw Gateway
macbook-tunnel · HTTP/2 · Quick Tunnel
yumu.kevin-zhu.com → :8080
CloudFlare Tunnel → Llama Server
Email OTP for humans · API token for code
llm.kevin-zhu.com → :8080
CloudFlare Tunnel → Yumu Hub
Local apps dashboard
yumu.kevin-zhu.com → :8888
OpenClaw → Llama Server (LAN)
Local network — same home network
192.168.1.155:8080
OpenClaw ↔ Yumu Hub (Local)
Same host — localhost
localhost:8080 ↔ :8888
iMessage (Native)
CloudFlare Tunnel
LAN Connection
Localhost