- Add llama.cpp package to Brewfile.private and Brewfile.work; remove ollama - Add fish functions for llama-server lifecycle management (start, stop, restart, status, logs) - Add com.joelazar.llama-server LaunchAgent with M4 Pro tuning and Qwen3.6-35B-A3B model - Update pi agent provider from ollama to llama-cpp with new model alias - Update README documentation to reference llama.cpp instead of Ollama - Remove com.joelazar.ollama LaunchAgent configuration