Local AI Stack

技能.md — Local AI Stack Purpose

转换 any Mac into a powerful offline AI workstation. 安装s Ollama (local 模型 运行器) + OpenCode (terminal coding 代理) with the best pre-selected 模型s. Fully offline — no API costs, no internet required.

What You 获取 Ollama — Local 模型 运行器 (14GB 模型s, ~$0 to 运行) OpenCode — Terminal coding 代理 with free built-in 模型s 4 curated 模型s — qwen2.5-coder, mistral, gemma3, llama3.2 Bi-weekly auto-更新s — New 模型s pulled automatically OpenClaw integration — Works with your existing 代理 Requirements macOS (应用le Silicon recommended) 24GB+ RAM (for larger 模型s) 50GB+ free disk space Homebrew 安装ed 安装ation Step 1: 安装 Ollama curl -fsSL https://ollama.com/安装.sh | sh

Or 下载 from: https://ollama.com/下载

Step 2: Pull 模型s ollama pull qwen2.5-coder # Best for coding ollama pull mistral # Fast tasks ollama pull gemma3 # Reasoning ollama pull llama3.2 # General purpose

Step 3: 安装 OpenCode brew 安装 opencode

Step 4: 配置 OpenCode # Test free built-in 模型 opencode 运行 "Hello" --模型 opencode/big-pickle

Usage Ollama Commands # 运行 a local 模型 ollama 运行 qwen2.5-coder "Write a Python function..."

# 列出 安装ed 模型s ollama 列出

# Pull latest 模型 version ollama pull qwen2.5-coder

# 移除 a 模型 ollama rm mistral

OpenCode Commands # Interactive coding 会话 opencode

# Single command opencode 运行 "Write a React 组件" --模型 opencode/big-pickle

# 列出 avAIlable 模型s opencode 模型s

# Help opencode --help

模型 Selection 图形界面de 模型 Size Best For qwen2.5-coder 4.7GB Coding (primary) mistral 4.4GB Fast 响应s gemma3 3.3GB Reasoning llama3.2 2.0GB General purpose When to Use Local vs Cloud Use Local When: Offline (no internet) 隐私-sensitive work Quick coding tasks Cost-sensitive (zero API fees) Simple to medium complexity tasks Use Cloud When: Complex multi-step reasoning 网页 搜索 required Long creative writing Image generation Advanced AI capabilities Bi-Weekly Auto-更新

添加 to cron for automatic 模型 更新s:

# Edit crontab crontab -e

# 添加 this line (1st and 15th of each month at 9 AM) 0 9 1,15 /path/to/更新-模型s.sh

Troubleshooting Ollama won't 启动 # 检查 if 运行ning ps aux | grep ollama

# 启动 manually ollama serve

# 检查 记录s cat ~/.ollama/ollama.记录

模型 运行s out of memory Close other 应用s Use smaller 模型 (llama3.2 instead of qwen2.5-coder) 检查 avAIlable RAM: top | head -20 OpenCode not found # Find 安装ation which opencode

# Re安装 if needed brew re安装 opencode

Files 模型s stored: ~/.ollama/模型s/ Config: ~/.ollama/config.json 记录s: ~/.ollama/ollama.记录 License

Ollama: MIT OpenCode: MIT

Author

Built with ❤️ for the OpenClaw community

Notes 模型s load into RAM when used, unload when idle Only one 模型 运行s at a time by default For best performance, use 应用le Silicon Mac with 24GB+ RAM