Chat-Completions Enhancements: Updated Examples + Code Improvements (#92)

* Completed step 2.2. Link python and install requirements in setup (Makefile for local runs, Dockerfile for building)

* Removed unneeded files

* - example made compact
- naming/structuring refactoring for convenience and readability
- general refactoring

Signed-off-by: Maroon Ayoub <[email protected]>

* Update examples/kv_events/README.md

Co-authored-by: Nir Rozenbaum <[email protected]>
Signed-off-by: Maroon Ayoub <[email protected]>

Apply suggestion from @nirrozenbaum

Co-authored-by: Nir Rozenbaum <[email protected]>
Signed-off-by: Maroon Ayoub <[email protected]>

* - upgraded python version
- minor fixes
- docker fix
- addressed review comments

Signed-off-by: Maroon Ayoub <[email protected]>

---------

Signed-off-by: Guy Girmonsky <[email protected]>
Co-authored-by: Maroon Ayoub <[email protected]>
Co-authored-by: Nir Rozenbaum <[email protected]>

Author: 3 people

.dockerignore

@@ -0,0 +1,21 @@
+# Git
+.git
+.gitignore
+
+# Build artifacts
+bin
+build
+
+# IDE and OS files
+.idea
+.vscode
+*.DS_Store
+
+# Local virtual environments
+venv
+
+# Python cache files
+__pycache__
+
+# Docker files
+Dockerfile

.github/workflows/ci-pr-checks.yaml

@@ -13,10 +13,19 @@ jobs:
       - name: Checkout source
         uses: actions/checkout@v4
 
-      - name: Install system dependencies (ZeroMQ)
+      - name: Cache system (apt) dependencies
+        uses: actions/cache@v4
+        with:
+          path: /var/cache/apt/archives
+          key: ${{ runner.os }}-apt-${{ hashFiles('**/.github/workflows/ci-pr-checks.yaml') }}
+          restore-keys: |
+            ${{ runner.os }}-apt-
+
+      - name: Install system dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install -y libzmq3-dev pkg-config
+          sudo add-apt-repository ppa:deadsnakes/ppa -y
+          sudo apt-get install -y libzmq3-dev pkg-config python3.12 python3.12-dev python3.12-venv
 
       - name: Sanity check repo contents
         run: ls -la
@@ -33,14 +42,33 @@ jobs:
       - name: Install dependencies
         run: go mod download
 
-      - name: Install golangci-lint
+      # This step is still useful to verify the Python config before subsequent steps
+      - name: Run detect-python
+        run: make detect-python
+
+      - name: Cache Python (pip) dependencies
+        uses: actions/cache@v4
+        with:
+          path: ~/.cache/pip
+          # This key is based ONLY on the requirements file
+          key: ${{ runner.os }}-pip-${{ hashFiles('**/requirements.txt') }}
+
+      - name: Install golangci-lint without running it
         uses: golangci/golangci-lint-action@v8
         with:
           version: v2.1.6
-          args: ""
+          args: "--help"
 
-      - name: Run precommit checks
-        run: make precommit
+      - name: Run precommit checks with CGo environment
+        run: |
+          # Export the CGO flags required for the linter to parse CGo files.
+          # This duplicates some logic from the Makefile but is necessary for the CI step.
+          export CGO_CFLAGS="$(python3.12-config --cflags) -I./lib"
+          export CGO_LDFLAGS="$(python3.12-config --ldflags --embed) -L./lib -ltokenizers -ldl -lm"
+          export CGO_ENABLED=1
+          
+          # Now run the linting command from the Makefile
+          make precommit
 
       - name: Run make build
         shell: bash

.golangci.yml

@@ -50,14 +50,13 @@ linters:
     - reassign
     - revive
     - staticcheck
-    - tagliatelle
+    # - tagliatelle
     - testableexamples
     - testpackage
     - thelper
     - tparallel
     - unconvert
     - unparam
-    - unused
     - usestdlibvars
     - varnamelen
     - whitespace
@@ -120,4 +119,4 @@ formatters:
     paths:
       - third_party$
       - builtin$
-      - examples$
+      - examples$

Dockerfile

@@ -19,11 +19,13 @@ ARG TARGETARCH
 
 WORKDIR /workspace
 
+# Install system-level dependencies first. This layer is very stable.
 USER root
 # Install EPEL repository directly and then ZeroMQ, as epel-release is not in default repos.
+# Install all necessary dependencies including Python 3.12 for chat-completions templating.
 # The builder is based on UBI8, so we need epel-release-8.
 RUN dnf install -y 'https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm' && \
-    dnf install -y gcc-c++ libstdc++ libstdc++-devel clang zeromq-devel pkgconfig && \
+    dnf install -y gcc-c++ libstdc++ libstdc++-devel clang zeromq-devel pkgconfig python3.12-devel python3.12-pip && \
     dnf clean all
 
 # Copy the Go Modules manifests
@@ -33,6 +35,12 @@ COPY go.sum go.sum
 # and so that source changes don't invalidate our downloaded layer
 RUN go mod download
 
+# Copy only the requirements file.
+COPY pkg/preprocessing/chat_completions/requirements.txt ./requirements.txt
+# Install Python dependencies. This layer will be cached unless requirements.txt changes.
+RUN python3.12 -m pip install --upgrade pip setuptools wheel && \
+    python3.12 -m pip install -r ./requirements.txt
+
 # Copy the go source
 COPY examples/kv_events examples/kv_events
 COPY . .
@@ -43,13 +51,17 @@ ARG RELEASE_VERSION=v1.22.1
 RUN curl -L https://github.com/daulet/tokenizers/releases/download/${RELEASE_VERSION}/libtokenizers.${TARGETOS}-${TARGETARCH}.tar.gz | tar -xz -C lib
 RUN ranlib lib/*.a
 
-# Build
-# the GOARCH has not a default value to allow the binary be built according to the host where the command
-# was called. For example, if we call make image-build in a local env which has the Apple Silicon M1 SO
-# the docker BUILDPLATFORM arg will be linux/arm64 when for Apple x86 it will be linux/amd64. Therefore,
-# by leaving it empty we can ensure that the container and binary shipped on it will have the same platform.
+# Set up Python environment variables needed for the build
+ENV PYTHONPATH=/workspace/pkg/preprocessing/chat_completions:/usr/lib64/python3.9/site-packages:/usr/lib/python3.9/site-packages
+ENV PYTHON=python3.9
 
-RUN CGO_ENABLED=1 GOOS=${TARGETOS:-linux} GOARCH=${TARGETARCH:-amd64} go build -ldflags="-extldflags '-L$(pwd)/lib'" -a -o bin/kv-cache-manager examples/kv_events/online/main.go
+# Build the application with CGO enabled.
+# We export CGO_CFLAGS and CGO_LDFLAGS using python3.12-config to ensure the Go compiler
+# can find the Python headers and libraries correctly. This mirrors the fix from the Makefile.
+RUN export CGO_CFLAGS="$(python3.12-config --cflags) -I/workspace/lib" && \
+    export CGO_LDFLAGS="$(python3.12-config --ldflags --embed) -L/workspace/lib -ltokenizers -ldl -lm" && \
+    CGO_ENABLED=1 GOOS=${TARGETOS:-linux} GOARCH=${TARGETARCH:-amd64} \
+    go build -a -o bin/kv-cache-manager examples/kv_events/online/main.go
 
 # Use distroless as minimal base image to package the manager binary
 # Refer to https://github.com/GoogleContainerTools/distroless for more details
@@ -59,8 +71,25 @@ WORKDIR /
 # The final image is UBI9, so we need epel-release-9.
 USER root
 RUN dnf install -y 'https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm' && \
-    dnf install -y zeromq
+    dnf install -y zeromq libxcrypt-compat python3.12 python3.12-pip && \
+    dnf clean all
+
+
+
+# Install Python dependencies in the final image.
+COPY --from=builder /workspace/requirements.txt /tmp/requirements.txt
+RUN python3.12 -m pip install --upgrade pip setuptools wheel && \
+    python3.12 -m pip install --no-cache-dir -r /tmp/requirements.txt \
+    && rm -rf /tmp/requirements.txt
+
+# Copy this project's own Python source code into the final image
+COPY --from=builder /workspace/pkg/preprocessing/chat_completions /app/pkg/preprocessing/chat_completions
+
+# Set the PYTHONPATH. This mirrors the Makefile's export, ensuring both this project's
+# Python code and the installed libraries (site-packages) are found at runtime.
+ENV PYTHONPATH=/app/pkg/preprocessing/chat_completions:/usr/lib64/python3.12/site-packages
 
+# Copy the compiled Go application
 COPY --from=builder /workspace/bin/kv-cache-manager /app/kv-cache-manager
 USER 65532:65532