From e50e4ea69bad1856d054f82b6fd0a4d78a39f69e Mon Sep 17 00:00:00 2001
From: Andres Marafioti <andimarafioti@gmail.com>
Date: Fri, 25 Sep 2026 09:41:34 +0200
Subject: [PATCH] Prepare release 0.4.1 with macOS PyPI wheels

---
 README.md                   | 30 ++++++++++++++++++++----------
 pyproject.toml              |  2 +-
 src/qwentts_cpp/__init__.py |  2 +-
 3 files changed, 22 insertions(+), 12 deletions(-)

diff --git a/README.md b/README.md
index 181666e..a5bed5e 100644
--- a/README.md
+++ b/README.md
@@ -34,7 +34,8 @@ CPU builds are still useful for development and smoke tests.
 
 ## Installation
 
-The default PyPI package is built for CUDA 12.8:
+Starting with 0.4.1, PyPI provides Linux CUDA 12.8 wheels and macOS 14+
+Apple Silicon Metal wheels. Pip selects the appropriate wheel automatically:
 
 ```bash
 pip install qwentts-cpp-python
@@ -74,13 +75,10 @@ revision; validation artifacts are not reused for publishing.
 
 ### Apple Silicon (Metal)
 
-The PyPI `0.4.0` release contains Linux CUDA wheels. For **macOS 14 or newer,
-arm64 Python 3.10+**, use the `+metal` wheel published to Hugging Face. After
-the **Publish Hugging Face Wheels** workflow completes, install it with:
+For **macOS 14 or newer, arm64 Python 3.10+**, install directly from PyPI:
 
 ```bash
-python -m pip install --only-binary=qwentts-cpp-python "qwentts-cpp-python==0.4.0+metal" \
-  -f https://huggingface.co/datasets/andito/qwentts-cpp-python-wheels/tree/main/whl/metal
+python -m pip install --only-binary=qwentts-cpp-python "qwentts-cpp-python>=0.4.1"
 python -c "from qwentts_cpp import QwenLibrary; print(QwenLibrary().version())"
 ```
 
@@ -91,14 +89,26 @@ from macOS. The build applies a shader-only workaround for the pinned ggml's
 invalid scalar-to-BF16-vector fill cast, restoring the source checkout afterward;
 the C ABI is unchanged. Use a native arm64 Python, rather than an Intel Python
 under Rosetta.
-The `+metal` version identifies the backend on Hugging Face; it is never sent to
-PyPI and cannot collide with the Linux CUDA or other backend variants.
+Downstream packages can declare `qwentts-cpp-python>=0.4.1` as a regular
+dependency, without an extra index or a direct wheel URL. The macOS platform
+tag distinguishes Metal wheels from Linux CUDA wheels on PyPI. Intel Macs and
+macOS versions older than 14 are not supported by this Metal wheel.
+
+The earlier `0.4.0` Metal wheel remains available on Hugging Face:
+
+```bash
+python -m pip install "qwentts-cpp-python==0.4.0+metal" \
+  -f https://huggingface.co/datasets/andito/qwentts-cpp-python-wheels/tree/main/whl/metal
+```
+
+The `+metal` version identifies the backend on Hugging Face; PyPI wheels use
+the plain public version without this suffix.
 
 Before publication, download the `hf-wheel-metal-macosx-arm64` artifact from
 the PR's **Apple Silicon Metal wheel** check, unzip it, and install its `.whl`
-with `python -m pip install /path/to/qwentts_cpp_python-0.4.0+metal-*.whl`.
+with `python -m pip install /path/to/qwentts_cpp_python-0.4.1+metal-*.whl`.
 This macOS check also runs for PRs; the Linux matrix remains dispatch-only.
-The Hugging Face publisher rebuilds and checks its own Metal wheel.
+Both publishers rebuild and check their own Metal wheels.
 
 Run a streaming smoke test using local GGUF weights (weights are not bundled):
 
diff --git a/pyproject.toml b/pyproject.toml
index 4a0f3c1..58edc32 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "qwentts-cpp-python"
-version = "0.4.0"
+version = "0.4.1"
 description = "Python ctypes bindings and wheel packaging for qwentts.cpp"
 readme = "README.md"
 license = "MIT"
diff --git a/src/qwentts_cpp/__init__.py b/src/qwentts_cpp/__init__.py
index e4d97a1..a82494e 100644
--- a/src/qwentts_cpp/__init__.py
+++ b/src/qwentts_cpp/__init__.py
@@ -16,7 +16,7 @@
 )
 from .models import GGUF_REPO, resolve_gguf_paths
 
-__version__ = "0.4.0"
+__version__ = "0.4.1"
 
 __all__ = [
     "GGUF_REPO",