refactor(ci): inline build steps, delete build-binary.sh

Replace build-binary.sh with discrete workflow steps: Build
mistralrs-server, Collect artifacts. Remove commented-out NCCL
check block.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.gitea/workflows/build-release.yml

@@ -39,30 +39,27 @@ jobs:
           fi
           echo "${HOME}/.cargo/bin" >> "$GITHUB_PATH"
 
-      #- name: Check for NCCL
-      #  run: |
-      #    if ! ldconfig -p | grep -q libnccl.so.2; then
-      #      echo "ERROR: libnccl not found. Install with:"
-      #      echo "  sudo dnf config-manager addrepo --from-repofile=https://developer.download.nvidia.com/compute/cuda/repos/rhel9/x86_64/cuda-rhel9.repo"
-      #      echo "  sudo dnf --repo=cuda-rhel9-x86_64 install libnccl libnccl-devel"
-      #      exit 1
-      #    fi
-
       - name: Clone mistral.rs at tag
         run: |
           git clone --depth 1 --branch "${{ inputs.tag }}" \
             https://github.com/EricLBuehler/mistral.rs.git src/
 
-      - name: Build
-        run: ./script/build-binary.sh
+      - name: Build mistralrs-server
+        run: |
+          export PATH="${{ matrix.cuda_home }}/bin:${PATH}"
+          export LD_LIBRARY_PATH="${{ matrix.cuda_home }}/targets/x86_64-linux/lib:${{ matrix.cuda_home }}/lib64:${LD_LIBRARY_PATH:-}"
+          cd src
+          cargo build --release --locked --features "${{ matrix.cargo_features }}"
         env:
-          FLAVOUR_NAME: ${{ matrix.name }}
-          CUDA_HOME: ${{ matrix.cuda_home }}
-          CARGO_FEATURES: ${{ matrix.cargo_features }}
           CUDA_COMPUTE_CAP: ${{ matrix.compute_caps }}
           CARGO_BUILD_JOBS: ${{ matrix.build_jobs }}
           NVCC_THREADS: ${{ matrix.nvcc_threads }}
-          SRC_DIR: src
+
+      - name: Collect artifacts
+        run: |
+          mkdir --parents artifacts
+          cp src/target/release/mistralrs-server "artifacts/mistralrs-server-${{ matrix.name }}"
+          echo "built: $(artifacts/mistralrs-server-${{ matrix.name }} --version 2>&1 | head -1)"
 
       - name: Upload binary artifact
         uses: actions/upload-artifact@v3

CLAUDE.md

@@ -27,7 +27,6 @@ Defined in `flavours.yml`. Each flavour specifies a name, `cuda_home`, `cargo_fe
 - `rpm/mistralrs.spec` — RPM spec (binary-only package, no rebuild)
 - `rpm/systemd/mistralrs@.service` — templated systemd unit (`@BINARY@` and `@FLAVOUR@` are sed-replaced during rpmbuild)
 - `rpm/systemd/mistralrs@.conf.example` — example env file for instances
-- `script/build-binary.sh` — compiles mistralrs-server with cargo (requires `FLAVOUR_NAME`, `CUDA_HOME`, `CARGO_FEATURES`, `CUDA_COMPUTE_CAP`, `SRC_DIR` env vars)
 - `script/setup/` — one-time infra setup scripts (DNS, TLS cert, nginx, GPG) for `rpm.lair.cafe` on host `oolon`
 
 ## Commands

script/build-binary.sh

@@ -1,25 +0,0 @@
-#!/usr/bin/env bash
-set -euo pipefail
-
-: "${FLAVOUR_NAME:?}"
-: "${CUDA_HOME:?}"
-: "${CARGO_FEATURES:?}"
-: "${CUDA_COMPUTE_CAP:?}"
-: "${SRC_DIR:?}"
-
-export PATH="${CUDA_HOME}/bin:${PATH}"
-export LD_LIBRARY_PATH="${CUDA_HOME}/targets/x86_64-linux/lib:${CUDA_HOME}/lib64:${LD_LIBRARY_PATH:-}"
-
-cd "${SRC_DIR}"
-
-# --locked ensures Cargo.lock is respected; fails loud if it's out of sync
-# rather than silently resolving to different versions.
-cargo build --release --locked --features "${CARGO_FEATURES}"
-
-mkdir -p ../artifacts
-cp target/release/mistralrs-server "../artifacts/mistralrs-server-${FLAVOUR_NAME}"
-
-# Also grab the other binaries if you want them
-cp target/release/mistralrs "../artifacts/mistralrs-${FLAVOUR_NAME}" 2>/dev/null || true
-
-echo "Built $(../artifacts/mistralrs-server-${FLAVOUR_NAME} --version 2>&1 | head -1)"