dockerize (#283)

.dockerignore

@@ -0,0 +1,6 @@
+ragna/_version.py
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class

.github/workflows/docker.yml

@@ -0,0 +1,108 @@
+name: docker
+
+on:
+  pull_request:
+    paths:
+      - ".github/workflows/docker.yml"
+      - ".dockerignore"
+      - "Dockerfile"
+      - "environment-dev.yml"
+      - "pyproject.toml"
+      - "ragna-docker.toml"
+      - "requirements-docker.lock"
+  push:
+    branches:
+      - release/*
+  workflow_dispatch:
+
+concurrency:
+  group: ${{ github.ref }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
+env:
+  REGISTRY: quay.io
+  PROJECT: quansight/ragna
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        shell: bash -el {0}
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Setup environment
+        uses: ./.github/actions/setup-env
+        with:
+          optional-dependencies: "false"
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v3
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Set up Docker Build Cache
+        id: cache
+        run: |
+          CACHE_DIR="${HOME}/.cache/docker"
+          mkdir --parents "${CACHE_DIR}"
+          echo "cache-dir=${CACHE_DIR}" | tee --append $GITHUB_OUTPUT
+
+      - name: Restore docker cache
+        uses: actions/cache@v3
+        with:
+          path: ${{ steps.cache.outputs.cache-dir }}
+          key: docker-${{ hashFiles('Dockerfile','requirements-docker.lock') }}
+          restore-keys: docker-
+
+      - name: Setup metadata
+        id: metadata
+        run: |
+          IMAGE_NAME="${{ env.REGISTRY }}/${{ env.PROJECT }}"
+          echo "image-name=${IMAGE_NAME}" | tee --append $GITHUB_OUTPUT
+
+          VERSION=$(python -m setuptools_scm)
+          echo "version=${VERSION}" | tee --append $GITHUB_OUTPUT
+
+          TAG=$(echo "${VERSION}" | sed 's/+/-/g')
+          echo "tag=${TAG}" | tee --append $GITHUB_OUTPUT
+
+          TAGGED_IMAGE_NAME="${IMAGE_NAME}:${TAG}"
+          echo "tagged-image-name=${TAGGED_IMAGE_NAME}" | tee --append $GITHUB_OUTPUT
+
+      - name: Login to Quay.io
+        if: ${{ github.event_name == 'workflow_dispatch' }}
+        uses: docker/login-action@v3
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ secrets.DOCKER_REGISTRY_USERNAME }}
+          password: ${{ secrets.DOCKER_REGISTRY_PASSWORD }}
+
+      - name: Build and push on workflow dispatch
+        id: build
+        uses: docker/build-push-action@v5
+        with:
+          cache-from: type=local,src=${{ steps.cache.outputs.cache-dir }}
+          cache-to: type=local,dest=${{ steps.cache.outputs.cache-dir }}
+          tags: ${{ steps.metadata.outputs.tagged-image-name }}
+          build-args: |
+            SETUPTOOLS_SCM_PRETEND_VERSION_FOR_RAGNA=${{ steps.metadata.outputs.version }}
+          # Unfortunately, there currently seems to be no way to build a multiplatform
+          # image and access a single one for the host platform afterwards. Thus, we
+          # only build a multiplatform image when we also want to push to the registry.
+          # https://github.com/docker/buildx/issues/166
+          # https://github.com/moby/buildkit/issues/1555
+          # prettier-ignore
+          platforms:
+            ${{ github.event_name == 'workflow_dispatch' && 'linux/amd64,linux/arm64' || 'linux/amd64' }}
+          load: ${{ github.event_name != 'workflow_dispatch' }}
+          push: ${{ github.event_name == 'workflow_dispatch' }}
+
+      - name: Smoke test
+        run: docker run ${{ steps.metadata.outputs.tagged-image-name }} --version

Dockerfile

@@ -0,0 +1,30 @@
+FROM python:3.11
+
+WORKDIR /opt/ragna
+
+COPY requirements-docker.lock .
+RUN pip install --progress-bar=off --no-deps --no-cache --requirement requirements-docker.lock
+
+# Pre-download the default embedding model
+RUN python -c "from chromadb.utils.embedding_functions import ONNXMiniLM_L6_V2; ONNXMiniLM_L6_V2()._download_model_if_not_exists()"
+
+COPY ragna ./ragna
+COPY pyproject.toml .
+# Since we don't copy the .git folder, but still use setuptools-scm as build-backend
+# we need to make two manual changes:
+# 1. With setuptools-scm all files that are tracked by git are automatically included in
+#    the built wheel. Since we have corresponding .dockerignore file to our .gitignore,
+#    the ragna folder only includes files that we are tracking. Thus, we just include
+#    everything manually.
+# 2. We need to pass the version expliclitly as
+#    --build-arg SETUPTOOLS_SCM_PRETEND_VERSION_FOR_RAGNA=...,
+#    since setuptools-scm cannot infer the version
+RUN echo '[tool.setuptools.package-data]\n"*" = ["*"]' >> pyproject.toml
+ARG SETUPTOOLS_SCM_PRETEND_VERSION_FOR_RAGNA
+RUN pip install --progress-bar=off --no-deps .
+
+WORKDIR /var/ragna
+COPY ragna-docker.toml ragna.toml
+
+ENTRYPOINT ["ragna"]
+CMD ["ui"]

environment-dev.yml

@@ -13,6 +13,8 @@ dependencies:
       - pre-commit
       - types-aiofiles
       - sqlalchemy-stubs
+      - setuptools-scm
+      - pip-tools
       # documentation
       - mkdocs
       - mkdocs-material

ragna-docker.toml

@@ -0,0 +1,22 @@
+local_cache_root = "/var/ragna"
+document = "ragna.core.LocalDocument"
+authentication = "ragna.deploy.RagnaDemoAuthentication"
+
+[components]
+source_storages = [
+    "ragna.source_storages.Chroma",
+    "ragna.source_storages.RagnaDemoSourceStorage",
+    "ragna.source_storages.LanceDB"
+]
+assistants = [
+    "ragna.assistants.RagnaDemoAssistant"
+]
+
+[api]
+url = "http://0.0.0.0:31476"
+origins = ["http://localhost:31477"]
+database_url = "sqlite:////var/ragna/ragna.db"
+
+[ui]
+url = "http://0.0.0.0:31477"
+origins = ["http://localhost:31477"]