v1.2.129
- Start managed llama.cpp from its binary directory.\n- Prepend the llama.cpp binary directory to process search paths so backend libraries such as CUDA/Vulkan DLLs can load.\n- Add regression coverage for llama.cpp process launch environment.
Full Changelog: v1.2.128...v1.2.129