mark as not working on windows

The releases from llama.cpp follow a different format on windows.
On unixy OSes, the server binary is
at
`build/bin/llama-server`
On Windows it is just at `llama-server.exe`. You would need to
update the download-release.sh script to make this work.

Author: NickCrews

.github/workflows/CI.yml

@@ -32,7 +32,9 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, windows-latest, macos-latest]
+        # windows isn't supported in the download-release.sh script from llama.cpp.
+        # See _binary.py for more info
+        os: [ubuntu-latest, macos-latest]
         python-version: ["3.8", "3.12"] # min and max versions, skip the middles
     runs-on: ${{ matrix.os }}
     steps:

README.md

@@ -32,6 +32,9 @@ For detailed API, read the source code.
 
 ## Install
 
+This only currently works on Linux and Mac. File an issue if you want a pointer on
+what needs to happen to make this work.
+
 For now, install directly from source:
 
 `python -m pip install git+https://github.com/NickCrews/llama-cpp-server-python@00cc5ece8783848139d41fb7f9c5e5c9b7a62686`