ci: download pre-built wheels for llama-cpp-python on macOS (#2235)

* ci: add extra index for llama-cpp-python MacOs

* retry

* fix

* fix pyproject

* retry

* again

* improvements

* typos

Author: anakin87

.github/workflows/llama_cpp.yml

@@ -30,8 +30,22 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
-        python-version: ["3.9", "3.13"]  
+        include:
+          - os: ubuntu-latest
+            python-version: "3.9"
+          - os: ubuntu-latest
+            python-version: "3.13"
+          - os: windows-latest
+            python-version: "3.9"
+          - os: windows-latest
+            python-version: "3.13"
+
+        # on macOS, we test with 3.10 and 3.12 because pre-built wheels are only available for these Python versions
+        # and installing from source might fail due to missing dependencies (CMake fails with "OpenMP not found")
+          - os: macos-latest
+            python-version: "3.10"
+          - os: macos-latest
+            python-version: "3.12"
 
     steps:
       - name: Support longpaths

integrations/llama_cpp/pyproject.toml

@@ -28,6 +28,14 @@ classifiers = [
 ]
 dependencies = ["haystack-ai>=2.16.1", "llama-cpp-python>=0.2.87"]
 
+# On macOS GitHub runners, we use a custom index to download pre-built wheels.
+# Installing from source might fail due to missing dependencies (CMake fails with "OpenMP not found")
+[tool.uv]
+index-strategy = "unsafe-best-match"  # this ensures that packages are not only searched in the below index but also in PyPI
+[[tool.uv.index]]
+name = "llama-cpp-python-macos"
+url = "https://abetlen.github.io/llama-cpp-python/whl/metal/"
+
 [project.urls]
 Documentation = "https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/llama_cpp#readme"
 Issues = "https://github.com/deepset-ai/haystack-core-integrations/issues"