chore: init

rpm/mistralrs.spec

@@ -0,0 +1,100 @@
+%global _build_id_links none
+%global debug_package %{nil}
+%global __strip /usr/bin/true
+
+# Passed in via --define at rpmbuild time
+%{!?mistralrs_version: %global mistralrs_version 0.7.0}
+%{!?mistralrs_flavour: %global mistralrs_flavour cuda13}
+
+Name:           mistralrs-server-%{mistralrs_flavour}
+Version:        %{mistralrs_version}
+Release:        1%{?dist}
+Summary:        Fast, flexible LLM inference server (mistral.rs, %{mistralrs_flavour} flavour)
+
+License:        MIT
+URL:            https://github.com/EricLBuehler/mistral.rs
+
+# Pre-built binary (produced in the build job, not rebuilt here)
+Source0:        mistralrs-server-%{mistralrs_flavour}
+Source1:        mistralrs@.service
+Source2:        mistralrs@.conf.example
+
+ExclusiveArch:  x86_64
+
+# Runtime requirements. We link against the CUDA runtime; consumers must have
+# a matching CUDA installation or the rpmfusion nvidia driver's cuda-libs.
+# We don't hard-require it at the RPM level because consumers may have CUDA
+# from multiple sources (nvidia direct, rpmfusion, etc.) — failing to load
+# libcuda.so at runtime gives a clearer error than RPM dep resolution would.
+Requires:       systemd
+
+# Flavours are mutually exclusive with other flavours of themselves at the
+# same install path, but you can install cuda13, cuda13-fa, cuda13-fa-nccl
+# side by side — they all get separate /opt paths.
+Provides:       mistralrs-server = %{version}-%{release}
+
+%description
+mistral.rs is a blazingly fast LLM inference engine written in Rust.
+This package provides the %{mistralrs_flavour} flavour, built with features:
+cuda, cudnn, and optionally flash-attn and nccl depending on flavour name.
+
+Binary installs to /opt/mistralrs/%{mistralrs_flavour}/bin/ and can coexist
+with other flavours. Use `update-alternatives --config mistralrs-server` to
+select the default /usr/bin/mistralrs-server symlink target.
+
+%prep
+# Nothing to unpack; Source0 is the binary itself
+cp %{SOURCE0} .
+cp %{SOURCE1} .
+cp %{SOURCE2} .
+
+%build
+# Already built
+
+%install
+install -D -m 0755 mistralrs-server-%{mistralrs_flavour} \
+    %{buildroot}/opt/mistralrs/%{mistralrs_flavour}/bin/mistralrs-server
+install -D -m 0644 mistralrs@.service \
+    %{buildroot}%{_unitdir}/mistralrs-%{mistralrs_flavour}@.service
+install -D -m 0644 mistralrs@.conf.example \
+    %{buildroot}%{_sysconfdir}/mistralrs/%{mistralrs_flavour}.conf.example
+
+# Patch the unit to point at this flavour's binary
+sed -i "s|@BINARY@|/opt/mistralrs/%{mistralrs_flavour}/bin/mistralrs-server|g" \
+    %{buildroot}%{_unitdir}/mistralrs-%{mistralrs_flavour}@.service
+sed -i "s|@FLAVOUR@|%{mistralrs_flavour}|g" \
+    %{buildroot}%{_unitdir}/mistralrs-%{mistralrs_flavour}@.service
+
+%post
+# Register this flavour as an alternative for /usr/bin/mistralrs-server.
+# Priority = 10 for cuda13, 20 for cuda13-fa, 30 for cuda13-fa-nccl so that
+# "more featureful" wins by default. Consumers can override with
+# `update-alternatives --config mistralrs-server`.
+priority=10
+case "%{mistralrs_flavour}" in
+    *nccl*) priority=30 ;;
+    *fa*)   priority=20 ;;
+esac
+update-alternatives --install /usr/bin/mistralrs-server mistralrs-server \
+    /opt/mistralrs/%{mistralrs_flavour}/bin/mistralrs-server "${priority}"
+
+%systemd_post mistralrs-%{mistralrs_flavour}@.service
+
+%preun
+%systemd_preun mistralrs-%{mistralrs_flavour}@.service
+
+%postun
+if [ $1 -eq 0 ]; then
+    update-alternatives --remove mistralrs-server \
+        /opt/mistralrs/%{mistralrs_flavour}/bin/mistralrs-server
+fi
+%systemd_postun_with_restart mistralrs-%{mistralrs_flavour}@.service
+
+%files
+/opt/mistralrs/%{mistralrs_flavour}/
+%{_unitdir}/mistralrs-%{mistralrs_flavour}@.service
+%{_sysconfdir}/mistralrs/%{mistralrs_flavour}.conf.example
+
+%changelog
+* Thu Apr 23 2026 Robin Thijssen <grenade@lair.cafe> - %{mistralrs_version}-1
+- Automated build for %{mistralrs_flavour} flavour