fix: Vicuna template instead of ChatML

fix: Ollama v0.3.4 for GGUF n_rot compat

fix: use Ollama v0.1.30 (correct URL)

feat: switch to Ollama (zero compilation, no OOM)

fix: remove CMAKE_ARGS to use pre-built wheel instead of compiling

fix: add MAX_JOBS=1 to prevent OOM during compilation

fix: use pre-built wheel to prevent OOM during Docker build

init: WHOAMISec model API