llama.cpp b9620 cleans up server and bundles UI assets for easier deployment

The ggml-org team has released llama.cpp b9620, a maintenance update for the widely-used open-source C++ library that enables local running of large language models. The core changes involve server-side code cleanup for static assets handling, including simplifying file name management and standardizing static file name usage across the codebase. Additionally, the CMake and UI build process has been updated to bundle UI assets into a single archive, making it easier to distribute and deploy the web interface alongside the server.

This release continues llama.cpp's tradition of extensive platform support, with pre-built binaries available for macOS (Apple Silicon and Intel), Linux (multiple architectures with Vulkan, ROCm, OpenVINO, SYCL), Windows (CPU, CUDA 12 and 13, Vulkan, SYCL, HIP), Android (arm64), and iOS XCFramework. Notably, the Apple Silicon build with KleidiAI is currently disabled. The UI assets are also provided as a separate download. For developers and power users running local LLMs, this update simplifies server management and UI deployment, maintaining llama.cpp's position as the go-to tool for local AI inference.