ai-zen

Progressive Framework for Agentic EA

手里有图 心里有谱 脚下有路

• 边界 知识域（Knowledge Domain）给Agent一张地图
• 工具 运行时 (Runtime Sandbox) 赋予Agent感知力
• 教学 白盒抽象(White-box Abstraction) 消除上下文黑盒
• 流程 机构防腐(Architectural Anti-corruption) 机械化强制执行
• 闭环 合流哲学(Merge Philosophy) 高吞吐量合并机制

aTeam with V-RPC

oMLX

sw_vers -productVersion

https://github.com/jundot/omlx/releases

hdiutil attach oMLX-0.2.20-macos26-tahoe.dmg
ls "/Volumes/oMLX/oMLX.app" 
cp -R "/Volumes/oMLX/oMLX.app" /Applications/
hdiutil detach "/Volumes/oMLX"

xattr -cr /Applications/oMLX.app

# homebrew 方式（未测试）
export HOMEBREW_NO_AUTO_UPDATE=1

export HOMEBREW_BREW_GIT_REMOTE="https://mirrors.ustc.edu.cn/brew.git"
export HOMEBREW_CORE_GIT_REMOTE="https://mirrors.ustc.edu.cn/homebrew-core.git"
export HOMEBREW_BOTTLE_DOMAIN="https://mirrors.ustc.edu.cn/homebrew-bottles"

# 然后重新尝试
brew tap jundot/omlx https://github.com/jundot/omlx

brew install omlx

# Upgrade to the latest version
brew update && brew upgrade omlx

# Run as a background service (auto-restarts on crash)
brew services start omlx

ollama 4 brew

export HOMEBREW_NO_AUTO_UPDATE=0
brew upgrade ollama
brew services restart ollama
ollama -v
ollama version is 0.13.5
  
vi homebrew.mxcl.ollama.plist
    OLLAMA_FLASH_ATTENTION
    1
    OLLAMA_KV_CACHE_TYPE
    q8_0
    OLLAMA_NUM_PARALLEL
    16
    OLLAMA_MAX_LOADED_MODELS
    8
    OLLAMA_MAX_QUEUE
    512
    OLLAMA_KEEP_ALIVE
    72h
    OLLAMA_NUM_THREADS
    32
    OLLAMA_CONTEXT_LENGTH
    20480
    OLLAMA_HOST
    0.0.0.0:11434

#配置检查
plutil ./homebrew.mxcl.ollama.plist
diff ./homebrew.mxcl.ollama.plist  ~/Library/LaunchAgents/homebrew.mxcl.ollama.plist
cp ./homebrew.mxcl.ollama.plist  ~/Library/LaunchAgents/homebrew.mxcl.ollama.plist

#重载服务（不要用 brew services restart，否则可能覆盖环境变量）：
launchctl unload ~/Library/LaunchAgents/homebrew.mxcl.ollama.plist
launchctl load ~/Library/LaunchAgents/homebrew.mxcl.ollama.plist

#检查生效
lsof -iTCP:11434 -sTCP:LISTEN
ps eww $(pgrep ollama) | grep OLLAMA  

qwen3 /no_think

ollama create qwen3n -f qwen3n.Modelfile