arm wip

.github/workflows/build-multiarch.yml

@@ -0,0 +1,120 @@
+name: Build Multi-Architecture Docker Images
+
+on:
+  workflow_dispatch:
+
+jobs:
+  build:
+    strategy:
+      matrix:
+        include:
+          - platform: linux/amd64
+            runs-on: ubuntu-latest
+            arch-suffix: amd64
+          - platform: linux/arm64
+            runs-on: ubuntu-24.04-arm
+            arch-suffix: arm64
+    
+    runs-on: ${{ matrix.runs-on }}
+    
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v4
+
+    - name: Extract version from pyproject.toml
+      id: version
+      run: |
+        VERSION=$(grep '^version = ' pyproject.toml | cut -d '"' -f 2)
+        echo "version=$VERSION" >> $GITHUB_OUTPUT
+        echo "Version: $VERSION"
+
+    - name: Set up Docker Buildx
+      uses: docker/setup-buildx-action@v3
+
+    - name: Login to Docker Hub
+      if: github.event_name != 'pull_request'
+      uses: docker/login-action@v3
+      with:
+        username: ${{ secrets.DOCKER_USERNAME }}
+        password: ${{ secrets.DOCKER_PASSWORD }}
+
+    - name: Build and push backend
+      uses: docker/build-push-action@v5
+      with:
+        context: .
+        file: ./Dockerfile.backend
+        platforms: ${{ matrix.platform }}
+        push: ${{ github.event_name != 'pull_request' }}
+        tags: phact/openrag-backend:${{ steps.version.outputs.version }}-${{ matrix.arch-suffix }}
+        cache-from: type=gha
+        cache-to: type=gha,mode=max
+
+    - name: Build and push frontend
+      uses: docker/build-push-action@v5
+      with:
+        context: .
+        file: ./Dockerfile.frontend
+        platforms: ${{ matrix.platform }}
+        push: ${{ github.event_name != 'pull_request' }}
+        tags: phact/openrag-frontend:${{ steps.version.outputs.version }}-${{ matrix.arch-suffix }}
+        cache-from: type=gha
+        cache-to: type=gha,mode=max
+
+    - name: Build and push OpenSearch
+      uses: docker/build-push-action@v5
+      with:
+        context: .
+        file: ./Dockerfile
+        platforms: ${{ matrix.platform }}
+        push: ${{ github.event_name != 'pull_request' }}
+        tags: phact/openrag-opensearch:${{ steps.version.outputs.version }}-${{ matrix.arch-suffix }}
+        cache-from: type=gha
+        cache-to: type=gha,mode=max
+
+  manifest:
+    needs: build
+    runs-on: ubuntu-latest
+    if: github.event_name != 'pull_request'
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v4
+
+    - name: Extract version from pyproject.toml
+      id: version
+      run: |
+        VERSION=$(grep '^version = ' pyproject.toml | cut -d '"' -f 2)
+        echo "version=$VERSION" >> $GITHUB_OUTPUT
+
+    - name: Login to Docker Hub
+      uses: docker/login-action@v3
+      with:
+        username: ${{ secrets.DOCKER_USERNAME }}
+        password: ${{ secrets.DOCKER_PASSWORD }}
+
+    - name: Create and push multi-arch manifests
+      run: |
+        VERSION=${{ steps.version.outputs.version }}
+        
+        # Backend manifest
+        docker buildx imagetools create -t phact/openrag-backend:$VERSION \
+          phact/openrag-backend:$VERSION-amd64 \
+          phact/openrag-backend:$VERSION-arm64
+        docker buildx imagetools create -t phact/openrag-backend:latest \
+          phact/openrag-backend:$VERSION-amd64 \
+          phact/openrag-backend:$VERSION-arm64
+          
+        # Frontend manifest  
+        docker buildx imagetools create -t phact/openrag-frontend:$VERSION \
+          phact/openrag-frontend:$VERSION-amd64 \
+          phact/openrag-frontend:$VERSION-arm64
+        docker buildx imagetools create -t phact/openrag-frontend:latest \
+          phact/openrag-frontend:$VERSION-amd64 \
+          phact/openrag-frontend:$VERSION-arm64
+          
+        # OpenSearch manifest
+        docker buildx imagetools create -t phact/openrag-opensearch:$VERSION \
+          phact/openrag-opensearch:$VERSION-amd64 \
+          phact/openrag-opensearch:$VERSION-arm64
+        docker buildx imagetools create -t phact/openrag-opensearch:latest \
+          phact/openrag-opensearch:$VERSION-amd64 \
+          phact/openrag-opensearch:$VERSION-arm64

Dockerfile

@@ -11,12 +11,19 @@ RUN usermod -aG wheel opensearch
 # Change the sudoers file to allow passwordless sudo
 RUN echo "opensearch ALL=(ALL) NOPASSWD:ALL" >> /etc/sudoers
 
-# FIXME handle the machine arch better, somehow
-ARG ASYNC_PROFILER_URL=https://github.com/async-profiler/async-profiler/releases/download/v4.0/async-profiler-4.0-linux-x64.tar.gz
+# Handle different architectures for async-profiler
+ARG TARGETARCH
+RUN if [ "$TARGETARCH" = "amd64" ]; then \
+      export ASYNC_PROFILER_URL=https://github.com/async-profiler/async-profiler/releases/download/v4.0/async-profiler-4.0-linux-x64.tar.gz; \
+    elif [ "$TARGETARCH" = "arm64" ]; then \
+      export ASYNC_PROFILER_URL=https://github.com/async-profiler/async-profiler/releases/download/v4.0/async-profiler-4.0-linux-arm64.tar.gz; \
+    else \
+      echo "Unsupported architecture: $TARGETARCH" && exit 1; \
+    fi && \
+    mkdir /opt/async-profiler && \
+    curl -s -L $ASYNC_PROFILER_URL | tar zxvf - --strip-components=1 -C /opt/async-profiler && \
+    chown -R opensearch:opensearch /opt/async-profiler
 
-RUN mkdir /opt/async-profiler
-RUN curl -s -L $ASYNC_PROFILER_URL | tar zxvf - --strip-components=1 -C /opt/async-profiler
-RUN chown -R opensearch:opensearch /opt/async-profiler
 
 RUN echo "#!/bin/bash" > /usr/share/opensearch/profile.sh
 RUN echo "export PATH=\$PATH:/opt/async-profiler/bin" >> /usr/share/opensearch/profile.sh

Dockerfile.backend

@@ -1,7 +1,13 @@
 FROM python:3.13-slim
 
 # Install curl for uv installation and openssl for RSA key generation
-RUN apt-get update && apt-get install -y curl openssl && rm -rf /var/lib/apt/lists/*
+# Also install git for potential dependencies and build-essential for native compilations
+RUN apt-get update && apt-get install -y \
+    curl \
+    openssl \
+    git \
+    build-essential \
+    && rm -rf /var/lib/apt/lists/*
 
 # Install uv
 RUN curl -LsSf https://astral.sh/uv/install.sh | sh
@@ -15,9 +21,24 @@ COPY pyproject.toml uv.lock ./
 RUN uv sync
 
 # Copy sample document and warmup script for docling
-COPY documents/2506.08231v1.pdf ./
+COPY documents/warmup_ocr.pdf ./
 COPY warm_up_docling.py ./
-RUN uv run python warm_up_docling.py && rm warm_up_docling.py 2506.08231v1.pdf
+RUN uv run docling-tools models download 
+RUN uv run python - <<'PY'
+import pathlib, easyocr
+cache = pathlib.Path("/root/.EasyOCR/model")
+cache.mkdir(parents=True, exist_ok=True)
+# Prewarm the detector + recog for Docling’s default langs
+easyocr.Reader(['fr','de','es','en'],
+               download_enabled=True,
+               model_storage_directory=str(cache))
+print("EasyOCR cache ready at", cache)
+PY
+
+RUN uv run python warm_up_docling.py && rm warm_up_docling.py warmup_ocr.pdf
+
+
+#ENV EASYOCR_MODULE_PATH=~/.cache/docling/models/EasyOcr/
 
 # Copy Python source
 COPY src/ ./src/
@@ -26,4 +47,4 @@ COPY src/ ./src/
 EXPOSE 8000
 
 # Start backend in foreground
-CMD ["uv", "run", "python", "src/main.py"]
+CMD ["uv", "run", "python", "src/main.py"]

docker-compose.yml

@@ -67,7 +67,6 @@ services:
       - ./documents:/app/documents:Z
       - ./keys:/app/keys:Z
     gpus: all
-    platform: linux/amd64
 
   openrag-frontend:
     image: phact/openrag-frontend:latest

pyproject.toml

@@ -27,10 +27,11 @@ dependencies = [
 [tool.uv.sources]
 #agentd = { path = "/home/tato/Desktop/agentd" }
 torch = [
-  { index = "pytorch-cu128" },
+  { index = "pytorch-cu128", marker = "sys_platform == 'linux' and platform_machine == 'x86_64'" },
+  # macOS & other platforms use PyPI (no index entry needed)
 ]
 torchvision = [
-  { index = "pytorch-cu128" },
+  { index = "pytorch-cu128", marker = "sys_platform == 'linux' and platform_machine == 'x86_64'" },
 ]
 
 [[tool.uv.index]]

warm_up_docling.py

@@ -4,10 +4,10 @@ print('Warming up docling models...')
 
 try:
     # Use the sample document to warm up docling
-    test_file = "/app/2506.08231v1.pdf"
+    test_file = "/app/warmup_ocr.pdf"
     print(f'Using {test_file} to warm up docling...')
     DocumentConverter().convert(test_file)
     print('Docling models warmed up successfully')
 except Exception as e:
     print(f'Docling warm-up completed with: {e}')
-    # This is expected - we just want to trigger the model downloads
+    # This is expected - we just want to trigger the model downloads