ci: pilot per-arch split + manifest merge for faster-whisper and llama-cpp-quantization (#9727)

ci: pilot per-arch split for faster-whisper and llama-cpp-quantization

Convert two backends from QEMU-emulated multi-arch (linux/amd64,linux/arm64
on a single ubuntu-latest) to native per-arch + manifest-list merge:
- amd64 leg on ubuntu-latest
- arm64 leg on ubuntu-24.04-arm (native, ~5-10x faster than emulated)
- merge job assembles both digests under the final tag via
  docker buildx imagetools create

Backends piloted:
- -cpu-faster-whisper (small Python, fast baseline)
- -cpu-llama-cpp-quantization (heavier compile path, stress test)

Infrastructure changes that the rest of Phase 2 (Tasks 2.5+) will reuse:
- .github/backend-matrix.yml entries gain a `platform-tag` field
  ('amd64'/'arm64') for matrix entries that participate in the split.
  Other entries omit it; backend_build.yml already defaults missing
  values to '' (empty cache key suffix preserved as cache<suffix>-).
- backend.yml + backend_pr.yml forward `platform-tag` from matrix to
  the reusable backend_build.yml.
- scripts/changed-backends.js groups filtered entries by tag-suffix
  and emits a `merge-matrix` (plus `has-merges`) for groups of size>=2.
  Singletons aren't merged.
- backend.yml + backend_pr.yml gain a `backend-merge-jobs` job that
  consumes merge-matrix and calls backend_merge.yml after backend-jobs.
  PR variant is also event-gated so the no-op-on-PR merge job doesn't
  even start.

The other 34 multi-arch entries are unchanged in this PR -- Task 2.5
fans out the same shape to them once the pilot is observed green.

Assisted-by: Claude:claude-opus-4-7

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>

.github/backend-matrix.yml

@@ -118,7 +118,8 @@ include:
   - build-type: ''
     cuda-major-version: ""
     cuda-minor-version: ""
-    platforms: 'linux/amd64,linux/arm64'
+    platforms: 'linux/amd64'
+    platform-tag: 'amd64'
     tag-latest: 'auto'
     tag-suffix: '-cpu-faster-whisper'
     runs-on: 'ubuntu-latest'
@@ -128,6 +129,20 @@ include:
     dockerfile: "./backend/Dockerfile.python"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: ''
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/arm64'
+    platform-tag: 'arm64'
+    tag-latest: 'auto'
+    tag-suffix: '-cpu-faster-whisper'
+    runs-on: 'ubuntu-24.04-arm'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'true'
+    backend: "faster-whisper"
+    dockerfile: "./backend/Dockerfile.python"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: ''
     cuda-major-version: ""
     cuda-minor-version: ""
@@ -157,7 +172,8 @@ include:
   - build-type: ''
     cuda-major-version: ""
     cuda-minor-version: ""
-    platforms: 'linux/amd64,linux/arm64'
+    platforms: 'linux/amd64'
+    platform-tag: 'amd64'
     tag-latest: 'auto'
     tag-suffix: '-cpu-llama-cpp-quantization'
     runs-on: 'ubuntu-latest'
@@ -167,6 +183,20 @@ include:
     dockerfile: "./backend/Dockerfile.python"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: ''
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/arm64'
+    platform-tag: 'arm64'
+    tag-latest: 'auto'
+    tag-suffix: '-cpu-llama-cpp-quantization'
+    runs-on: 'ubuntu-24.04-arm'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'true'
+    backend: "llama-cpp-quantization"
+    dockerfile: "./backend/Dockerfile.python"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: ''
     cuda-major-version: ""
     cuda-minor-version: ""

.github/workflows/backend.yml

@@ -34,8 +34,10 @@ jobs:
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
       matrix-darwin: ${{ steps.set-matrix.outputs.matrix-darwin }}
+      merge-matrix: ${{ steps.set-matrix.outputs['merge-matrix'] }}
       has-backends: ${{ steps.set-matrix.outputs.has-backends }}
       has-backends-darwin: ${{ steps.set-matrix.outputs.has-backends-darwin }}
+      has-merges: ${{ steps.set-matrix.outputs['has-merges'] }}
     steps:
       - name: Checkout repository
         uses: actions/checkout@v6
@@ -70,6 +72,7 @@ jobs:
       cuda-major-version: ${{ matrix.cuda-major-version }}
       cuda-minor-version: ${{ matrix.cuda-minor-version }}
       platforms: ${{ matrix.platforms }}
+      platform-tag: ${{ matrix.platform-tag || '' }}
       runs-on: ${{ matrix.runs-on }}
       base-image: ${{ matrix.base-image }}
       backend: ${{ matrix.backend }}
@@ -88,6 +91,22 @@ jobs:
       max-parallel: 8
       matrix: ${{ fromJson(needs.generate-matrix.outputs.matrix) }}
 
+  backend-merge-jobs:
+    needs: [generate-matrix, backend-jobs]
+    if: needs.generate-matrix.outputs.has-merges == 'true'
+    uses: ./.github/workflows/backend_merge.yml
+    with:
+      tag-latest: ${{ matrix.tag-latest }}
+      tag-suffix: ${{ matrix.tag-suffix }}
+    secrets:
+      dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+      dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+      quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+      quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+    strategy:
+      fail-fast: false
+      matrix: ${{ fromJson(needs.generate-matrix.outputs['merge-matrix']) }}
+
   backend-jobs-darwin:
     needs: generate-matrix
     if: needs.generate-matrix.outputs.has-backends-darwin == 'true'

.github/workflows/backend_pr.yml

@@ -13,8 +13,10 @@ jobs:
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
       matrix-darwin: ${{ steps.set-matrix.outputs.matrix-darwin }}
+      merge-matrix: ${{ steps.set-matrix.outputs['merge-matrix'] }}
       has-backends: ${{ steps.set-matrix.outputs.has-backends }}
       has-backends-darwin: ${{ steps.set-matrix.outputs.has-backends-darwin }}
+      has-merges: ${{ steps.set-matrix.outputs['has-merges'] }}
     steps:
       - name: Checkout repository
         uses: actions/checkout@v6
@@ -46,6 +48,7 @@ jobs:
       cuda-major-version: ${{ matrix.cuda-major-version }}
       cuda-minor-version: ${{ matrix.cuda-minor-version }}
       platforms: ${{ matrix.platforms }}
+      platform-tag: ${{ matrix.platform-tag || '' }}
       runs-on: ${{ matrix.runs-on }}
       base-image: ${{ matrix.base-image }}
       backend: ${{ matrix.backend }}
@@ -61,6 +64,22 @@ jobs:
       fail-fast: true
       max-parallel: 8
       matrix: ${{ fromJson(needs.generate-matrix.outputs.matrix) }}
+  backend-merge-jobs:
+    needs: [generate-matrix, backend-jobs]
+    # backend_merge.yml's push-side steps are all gated on
+    # github.event_name != 'pull_request', so on a PR the merge job would
+    # do nothing. Skip it entirely to avoid spinning up an empty runner.
+    if: github.event_name != 'pull_request' && needs.generate-matrix.outputs.has-merges == 'true'
+    uses: ./.github/workflows/backend_merge.yml
+    with:
+      tag-latest: ${{ matrix.tag-latest }}
+      tag-suffix: ${{ matrix.tag-suffix }}
+    secrets:
+      quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+      quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+    strategy:
+      fail-fast: false
+      matrix: ${{ fromJson(needs.generate-matrix.outputs['merge-matrix']) }}
   backend-jobs-darwin:
     needs: generate-matrix
     uses: ./.github/workflows/backend_build_darwin.yml