add Dockerfile and CI image build

.github/workflows/release.yml

@@ -4,6 +4,9 @@ on:
     tags:
       - "v*.*.*"
 
+env:
+  REGISTRY_IMAGE: holisticon/bpm-ai-inference
+
 jobs:
   publish:
     runs-on: ubuntu-latest
@@ -16,3 +19,78 @@ jobs:
           pypi_token: ${{ secrets.PYPI_TOKEN }}
           # Install project without developer requirements.
           ignore_dev_requirements: "yes"
+
+  build-push:
+    strategy:
+      matrix:
+        include:
+          - arch: amd64
+          - arch: arm64
+    runs-on: ubuntu-latest
+    outputs:
+      image: ${{ steps.image.outputs.image }}
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v3
+      - run: docker context create builders
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+        with:
+          version: latest
+          endpoint: builders
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_PASSWORD }}
+      - name: Get the version
+        id: vars
+        run: echo ::set-output name=tag::$(echo ${GITHUB_REF:10})
+      - name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          platforms: linux/${{ matrix.arch }}
+          provenance: false # otherwise a manifest list is created and create-push-manifest job fails
+          context: .
+          file: ./Dockerfile
+          build-args: |
+            PYTHON_VERSION=3.11
+          push: true
+          tags: |
+            ${{ env.REGISTRY_IMAGE }}:${{ steps.vars.outputs.tag }}-cpu-${{ matrix.arch }}
+            ${{ env.REGISTRY_IMAGE }}:latest-cpu-${{ matrix.arch }}
+      - id: image
+        run: echo "::set-output name=image::$(echo ${{ env.REGISTRY_IMAGE }}:latest-cpu)"
+
+  create-push-manifest:
+    runs-on: ubuntu-latest
+    needs: [build-push]
+    steps:
+      -
+        name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_PASSWORD }}
+      -
+        name: Get version
+        id: vars
+        run: echo ::set-output name=tag::$(echo ${GITHUB_REF:10})
+      -
+        name: Create version-manifest and push
+        run: |
+          docker manifest create \
+            ${{ env.REGISTRY_IMAGE }}:${{ steps.vars.outputs.tag }}-cpu \
+            --amend ${{ env.REGISTRY_IMAGE }}:${{ steps.vars.outputs.tag }}-cpu-amd64 \
+            --amend ${{ env.REGISTRY_IMAGE }}:${{ steps.vars.outputs.tag }}-cpu-arm64
+          docker manifest push ${{ env.REGISTRY_IMAGE }}:${{ steps.vars.outputs.tag }}-cpu
+      -
+        name: Create latest-manifest and push
+        run: |
+          docker manifest create \
+            ${{ env.REGISTRY_IMAGE }}:latest-cpu \
+            --amend ${{ env.REGISTRY_IMAGE }}:latest-cpu-amd64 \
+            --amend ${{ env.REGISTRY_IMAGE }}:latest-cpu-arm64
+          docker manifest push ${{ env.REGISTRY_IMAGE }}:latest-cpu

Dockerfile

@@ -0,0 +1,42 @@
+ARG PYTHON_VERSION="3.11"
+
+###############################################################################
+# 1. Install dependencies
+###############################################################################
+# poetry setup code based on https://github.com/thehale/docker-python-poetry (does not provide multiarch images)
+FROM python:${PYTHON_VERSION} AS build-python
+ARG POETRY_VERSION="1.6.1"
+
+ENV POETRY_VERSION=${POETRY_VERSION}
+ENV POETRY_HOME="/opt/poetry"
+ENV POETRY_VIRTUALENVS_IN_PROJECT=true
+ENV POETRY_NO_INTERACTION=1
+ENV PATH="$POETRY_HOME/bin:$PATH"
+
+RUN apt-get update && apt-get install -y --no-install-recommends curl
+# Install Poetry via the official installer: https://python-poetry.org/docs/master/#installing-with-the-official-installer
+# This script respects $POETRY_VERSION & $POETRY_HOME
+RUN curl -sSL https://install.python-poetry.org | python3 -
+# only install dependencies into project virtualenv
+WORKDIR /app
+COPY requirements.txt pyproject.toml poetry.lock ./
+RUN poetry run python -m pip install -r requirements.txt
+RUN poetry install --only main --no-root --no-cache
+
+###############################################################################
+# 2. Final, minimal image that starts the inference server daemon
+###############################################################################
+FROM python:${PYTHON_VERSION}-slim
+ARG PYTHON_VERSION
+
+ENV PYTHONUNBUFFERED=1
+
+WORKDIR /app
+COPY ./bpm_ai_inference/ ./bpm_ai_inference/
+COPY --from=build-python /app/.venv/lib/python${PYTHON_VERSION}/site-packages /usr/local/lib/python${PYTHON_VERSION}/site-packages
+RUN apt-get update  \
+    && apt-get install -y --no-install-recommends curl tesseract-ocr poppler-utils  \
+    && apt-get autoremove -y && apt-get clean && rm -rf /var/lib/apt/lists/*
+
+COPY init.py .
+CMD ["python3", "init.py", "python -m bpm_ai_inference.daemon"]

bpm_ai_inference/deamon.py → bpm_ai_inference/daemon.py

@@ -27,7 +27,7 @@
     logging.basicConfig(format='%(levelname)s: %(message)s', level=logging.INFO)
 
     daemon = create_remote_object_daemon(
-        host=os.environ.get('DAEMON_HOST', 'localhost'),
+        host=os.environ.get('DAEMON_HOST', '0.0.0.0'),
         port=int(os.environ.get('DAEMON_PORT', 6666))
     )
 

init.py

@@ -0,0 +1,29 @@
+import subprocess
+import os
+import time
+import sys
+
+
+def main():
+    # Get the list of processes from command line arguments
+    processes_params = [arg.split() for arg in sys.argv[1:]]
+    # Start all processes
+    processes = [subprocess.Popen(params) for params in processes_params]
+    print(f"[init] Started {len(processes)} processes.")
+    # Continuously monitor the processes
+    while True:
+        for i, process in enumerate(processes):
+            exit_status = process.poll()
+            if exit_status is not None:
+                print(f"[init] Process '{' '.join(processes_params[i])}' exited with status {exit_status}, terminating other processes...")
+                # Terminate all other processes
+                for j, other_process in enumerate(processes):
+                    if i != j:
+                        other_process.terminate()
+                return
+        # Sleep for a bit to avoid busy-waiting
+        time.sleep(1)
+
+
+if __name__ == "__main__":
+    main()

pyproject.toml

@@ -14,19 +14,23 @@ bpm-ai-core = "2.2.0"
 faster-whisper = "^0.10.0"
 lingua-language-detector = "^2.0.2"
 pytesseract = "^0.3.10"
-torch = [
-    { version = "=2.2.2", source="pypi", markers = "sys_platform == 'darwin'" },
-    { version = "=2.2.2+cpu", source = "torch-cpu", markers = "sys_platform != 'darwin'" },
-]
 transformers = "^4.37.2"
 sacremoses = "^0.1.1"
 sentencepiece = "^0.2.0"
 nltk = "^3.8.0"
+pyro5 = "^5.15"
+
+
+[tool.poetry.group.dev.dependencies]
+torch = [
+    { version = "=2.2.2", source="pypi", markers = "sys_platform == 'darwin'" },
+    { version = "=2.2.2+cpu", source = "torch-cpu", markers = "sys_platform != 'darwin'" },
+]
 spacy = [
     { version = "=3.7.4", markers = "sys_platform != 'darwin'" },
     { version = "=3.7.4", extras = ["apple"], markers = "sys_platform == 'darwin' and platform_machine == 'arm64'" },
 ]
-pyro5 = "^5.15"
+
 
 [[tool.poetry.source]]
 name = "torch-cpu"

requirements.txt

@@ -0,0 +1,3 @@
+spacy==3.7.4
+torch==2.2.2
+--extra-index-url https://download.pytorch.org/whl/cpu

tests/test_speech.py

@@ -8,6 +8,7 @@ async def test_faster_whisper():
 
 
 async def test_faster_whisper_url():
+    # todo flaky: av.error.EOFError: [Errno 541478725] End of file: 'audio.ogg'
     fw = FasterWhisperASR()
     result = await fw.transcribe("https://upload.wikimedia.org/wikipedia/commons/d/dd/Armstrong_Small_Step.ogg")
     assert "giant leap for mankind" in result.text.lower()