chore: bump to v0.1.6 and add CHANGELOG.md

CHANGELOG.md

@@ -0,0 +1,43 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+## [0.1.6] - 2026-05-19
+
+### Added
+- `linux-arm64` prebuilt static libraries (CPU) via `Dockerfile.libs-arm64`
+  using the aarch64-linux-gnu cross-compile toolchain.
+- `build-libs-linux-arm64` Makefile target.
+- `ARCH_SUFFIX` build-arg on the CPU stage of `Dockerfile.libs` so the same
+  Dockerfile can output `linux-amd64` or `linux-arm64` paths.
+
+### Changed
+- Bumped llama.cpp from `b8772` to `b9222`.
+- CGO LDFLAGS now link `-lllama-common -lllama-common-base` instead of
+  `-lcommon` across all platforms — upstream split the legacy
+  `libcommon.a` into `libllama-common.a` + `libllama-common-base.a`.
+  Consumers that build with `-tags llamacpp` (or the cuda/vulkan/whispercpp
+  variants) only need to rebuild; no source change required.
+- CUDA + Vulkan stages of `Dockerfile.libs` switched to `gcc-12` — gcc-13
+  hits a reproducible internal-compiler-error on `fattn-mma-f16` templates
+  (CUDA) and `arg.cpp` (Vulkan) at b9222.
+- Vulkan stage now installs `spirv-headers` (new b9222 dep —
+  `ggml-vulkan` calls `find_package(SPIRV-Headers)`).
+
+### Fixed
+- `Dockerfile.libs` and `Dockerfile.android` now copy mtmd headers from
+  `tools/mtmd/` (`mtmd.h`, `mtmd-helper.h`). Previously the headers were
+  committed manually because the Dockerfiles didn't track them.
+
+### Rebuilt
+All seven prebuilt platforms refreshed against b9222:
+`linux-amd64`, `linux-amd64-cuda`, `linux-amd64-vulkan`, `linux-arm64`,
+`android-arm64`, `darwin-amd64`, `darwin-arm64`.
+
+[Unreleased]: https://github.com/FootprintAI/go-nativeml/compare/v0.1.6...HEAD
+[0.1.6]: https://github.com/FootprintAI/go-nativeml/compare/v0.1.5...v0.1.6

version.go

@@ -1,7 +1,7 @@
 package gonativeml
 
 const (
-	Version           = "v0.1.7"
+	Version           = "v0.1.6"
 	LlamaCppVersion   = "b9222"
 	WhisperCppVersion = "v1.8.3"
 )