[OPIK-6673] [CI] feat: add hadolint Dockerfile linting to CI

.pre-commit-config.yaml

@@ -317,4 +317,20 @@ repos:
         # when a workflow that `uses:` them is linted. Matches actionlint's own
         # default files pattern.
         files: ^\.github/workflows/.+\.(yml|yaml)$
+
+  # ---------------------------------------------------------------------------
+  # Dockerfiles — hadolint. The docker-image hook needs only Docker (no local
+  # binary), matching CI runners. Runs hadolint's default rule set (failure
+  # threshold `info`); the few deliberate exceptions are `# hadolint ignore=`
+  # comments inline in the Dockerfiles, so no repo-level config file is needed.
+  # ---------------------------------------------------------------------------
+  - repo: https://github.com/hadolint/hadolint
+    rev: v2.14.0
+    hooks:
+      - id: hadolint-docker
+        name: 🐳 hadolint — dockerfiles
+        # Explicit files: regex (not the upstream `types: [dockerfile]`) so the
+        # CI matrix builder in scripts/precommit-detect-hooks.py can route a
+        # changed Dockerfile to this leg — it matches on paths, not content type.
+        files: (^|/)Dockerfile(\..+)?$
 ...

CONTRIBUTING.md

@@ -50,6 +50,9 @@ Please review the CLA before contributing:
 ## GitHub Actions workflows
 Workflow files in `.github/workflows/` are validated with [actionlint](https://github.com/rhysd/actionlint), which runs as a hook in the unified `🐙 Code Quality` workflow (and locally via pre-commit) on changed workflow files. The pre-commit framework builds the pinned actionlint from source automatically — no manual install needed. Run `make hooks` once per clone to enable it locally.
 
+## Dockerfiles
+Dockerfiles are linted with [hadolint](https://github.com/hadolint/hadolint), which runs as a hook in the unified `🐙 Code Quality` workflow (and locally via pre-commit) on changed Dockerfiles. It uses hadolint's default rule set; the handful of intentionally-suppressed rules are annotated inline in each Dockerfile with a `# hadolint ignore=` comment and a reason. The hook runs hadolint via its Docker image, so it needs only Docker — no manual install. To run it directly on a single file: `docker run --rm -i ghcr.io/hadolint/hadolint < path/to/Dockerfile`.
+
 ## Generated files (do not edit manually)
 - `apps/opik-backend/src/main/resources/model_prices_and_context_window.json`
 - `apps/opik-frontend/src/data/model_prices_and_context_window.json`

apps/opik-backend/Dockerfile

@@ -32,6 +32,11 @@ COPY install_rds_cert.sh /tmp/install_rds_cert.sh
 # perl is intentionally not installed: it pulls in perl-Archive-Tar, which is
 # affected by CVE-2026-9538 (ALAS2023-2026-1805) and has no patched AL2023 RPM.
 # install_rds_cert.sh uses sed/openssl instead of perl to extract the cert CN.
+# pipefail so the `sha256sum -c` failing inside the piped RUN below aborts the build.
+SHELL ["/bin/bash", "-o", "pipefail", "-c"]
+# DL3033: AL2023 core packages track a rolling security channel; pinning exact
+# RPM versions here would rot as AWS updates the base image's repos.
+# hadolint ignore=DL3033
 RUN yum update -y && \
     yum install -y --allowerasing shadow ca-certificates openssl dos2unix curl && \
     yum clean all && \

apps/opik-frontend/Dockerfile

@@ -21,10 +21,8 @@ ENV NODE_OPTIONS="--max-old-space-size=8192"
 ARG BUILD_MODE=production
 ARG BUILD_PLUGINS
 ENV VITE_FE_PLUGINS=${BUILD_PLUGINS}
-RUN npm run build -- --mode $BUILD_MODE
-
-# Create version.json for checking version
-RUN printf '{"version": "%s"}\n' "$VITE_APP_VERSION" > dist/version.json
+RUN npm run build -- --mode $BUILD_MODE \
+    && printf '{"version": "%s"}\n' "$VITE_APP_VERSION" > dist/version.json
 
 # Production stage
 # Using nginx alpine image with OpenTelemetry support

apps/opik-guardrails-backend/Dockerfile

@@ -7,6 +7,9 @@ ENV PYTHONUNBUFFERED=1 \
     OPIK_GUARDRAILS_DEVICE=cuda:0
 
 # Install Python and other dependencies
+# DL3008: Ubuntu apt packages track a rolling security channel; pinning exact
+# versions here would rot as the base image's repos are updated.
+# hadolint ignore=DL3008
 RUN apt-get update && apt-get install -y --no-install-recommends \
     python3.10 \
     python3-pip \
@@ -22,8 +25,11 @@ RUN ln -sf /usr/bin/python3.10 /usr/bin/python && \
 WORKDIR /opt/opik-guardrails-backend
 
 COPY requirements.txt .
-RUN pip install --no-cache-dir -U pip
-RUN pip install --no-cache-dir --disable-pip-version-check -r requirements.txt
+# Ubuntu 22.04 ships pip ~22; the pinned torch/transformers/hf_xet wheels need a
+# newer resolver. Float pip forward from a >=24 floor (no rot — it's a lower
+# bound, not a frozen version); dependency versions themselves are in requirements.txt.
+RUN pip install --no-cache-dir -U "pip>=24.0" && \
+    pip install --no-cache-dir --disable-pip-version-check -r requirements.txt
 
 COPY entrypoint.sh .
 RUN chmod u+x entrypoint.sh
@@ -33,13 +39,13 @@ RUN chmod u+x entrypoint.sh
 RUN mkdir /.cache/ /.local/ && chown -R 1001:1001 /opt/opik-guardrails-backend /.cache/ /.local/
 USER 1001:1001
 
-# Download spaCy model for PII detection
-RUN python -m spacy download en_core_web_lg
-
-# Download transformer model for restricted topic validation
+# Download models with the final user so files land readable by the user:
+#   - spaCy en_core_web_lg for PII detection
+#   - transformer bart-large-mnli for restricted topic validation
 # HF_HUB_DISABLE_XET=1 forces standard HTTP fallback; the Xet protocol
 # used by huggingface_hub[hf_xet] is blocked in most CI/Docker build environments.
-RUN HF_HUB_DISABLE_XET=1 python -c "from transformers import AutoModelForSequenceClassification, AutoTokenizer; AutoModelForSequenceClassification.from_pretrained('facebook/bart-large-mnli'); AutoTokenizer.from_pretrained('facebook/bart-large-mnli')"
+RUN python -m spacy download en_core_web_lg && \
+    HF_HUB_DISABLE_XET=1 python -c "from transformers import AutoModelForSequenceClassification, AutoTokenizer; AutoModelForSequenceClassification.from_pretrained('facebook/bart-large-mnli'); AutoTokenizer.from_pretrained('facebook/bart-large-mnli')"
 
 COPY --chown=1001:1001 opik_guardrails ./opik_guardrails
 

apps/opik-python-backend/Dockerfile

@@ -4,7 +4,12 @@
 #######################################################################
 FROM docker:29.5.1 AS builder
 
+# Alpine base ships /bin/ash, not bash; ash honors -o pipefail for the RUN
+# below whose pipe into xargs would otherwise mask a failing find.
+SHELL ["/bin/ash", "-o", "pipefail", "-c"]
+
 # Toolchain only for compiling native wheels (rust/cargo for pydantic-core).
+# hadolint ignore=DL3018
 RUN apk add --no-cache python3 py3-pip build-base python3-dev musl-dev gcc libffi-dev rust cargo
 
 WORKDIR /opt/opik-python-backend
@@ -26,7 +31,7 @@ RUN set -eux; \
     find "$PY_SITE" \
          -path "$PY_SITE/datasets/*" -prune -o \
          -path "$PY_SITE/opik_optimizer/*" -prune -o \
-         -name '*.py' -type f -print | xargs -r rm -f; \
+         -name '*.py' -type f -print0 | xargs -0 -r rm -f; \
     find "$PY_SITE" -type d -name '__pycache__' -prune -exec rm -rf {} +; \
     find "$PY_SITE" -name '*.pyi' -delete; \
     pip uninstall -y pip setuptools wheel || true; \
@@ -37,10 +42,11 @@ RUN set -eux; \
 #######################################################################
 FROM docker:29.5.1
 
-# Upgrade libexpat to the latest available version to mitigate CVEs.
-RUN apk add --no-cache --upgrade libexpat
-# Runtime-only OS deps: tini (PID 1 reaper) + python3 interpreter.
-RUN apk add --no-cache tini python3
+# libexpat --upgrade mitigates CVEs; tini (PID 1 reaper) + python3 are the
+# runtime-only OS deps.
+# hadolint ignore=DL3018
+RUN apk add --no-cache --upgrade libexpat && \
+    apk add --no-cache tini python3
 
 WORKDIR /opt/opik-python-backend
 

scripts/precommit-hook-descriptions.tsv

@@ -27,6 +27,7 @@ lizard	Cyclomatic-complexity gate
 vulture	Find dead code
 helm-docs	Regenerate Helm chart README
 actionlint	Lint GitHub Actions workflows
+hadolint	Lint Dockerfiles
 spotless	Format Java code
 eslint	Lint + autofix JS/TS
 typecheck	Whole-project tsc type check