pin PepperWizard image build to the package versions and base images …

.github/workflows/publish.yml

@@ -57,8 +57,7 @@ jobs:
           cache-to: type=gha,mode=max,scope=stt
 
   backfill:
-    # On private->public flip or manual re-run: build every existing v* tag,
-    # both images per tag.
+    # On private->public flip or manual re-run: build each v* tag that has no published image; a published image is never overwritten.
     if: github.event_name == 'public' || github.event_name == 'workflow_dispatch'
     runs-on: ubuntu-latest
     steps:
@@ -80,17 +79,24 @@ jobs:
             echo "No v* tags found, nothing to backfill."
             exit 0
           fi
+          build_missing() {
+            if docker manifest inspect "$1" >/dev/null 2>&1; then
+              echo "$1 is published, skipped"
+            else
+              docker build -t "$1" -f "$2" .
+              docker push "$1"
+            fi
+          }
           for tag in $tags; do
             echo "::group::$tag"
             git checkout "$tag"
-            docker build -t "$WIZARD_IMAGE:$tag" -f Dockerfile .
-            docker push "$WIZARD_IMAGE:$tag"
-            docker build -t "$STT_IMAGE:$tag" -f stt-service/Dockerfile .
-            docker push "$STT_IMAGE:$tag"
+            build_missing "$WIZARD_IMAGE:$tag" Dockerfile
+            build_missing "$STT_IMAGE:$tag" stt-service/Dockerfile
             echo "::endgroup::"
           done
           latest=$(echo "$tags" | tail -1)
-          docker tag "$WIZARD_IMAGE:$latest" "$WIZARD_IMAGE:latest"
-          docker push "$WIZARD_IMAGE:latest"
-          docker tag "$STT_IMAGE:$latest" "$STT_IMAGE:latest"
-          docker push "$STT_IMAGE:latest"
+          for image in "$WIZARD_IMAGE" "$STT_IMAGE"; do
+            docker pull "$image:$latest"
+            docker tag "$image:$latest" "$image:latest"
+            docker push "$image:latest"
+          done

Dockerfile

@@ -1,8 +1,8 @@
 # Stage 1: Get the library
-FROM ghcr.io/action-prediction-lab/pepper-box:latest AS bridge-source
+FROM ghcr.io/action-prediction-lab/pepper-box:v2.3.1@sha256:cb296a2ac483ac7d5789bc0367227464b191c6da9110b9db5c565bb81938d8f8 AS bridge-source
 
 # Stage 2: Build the Wizard
-FROM python:3.9-slim
+FROM python:3.9-slim@sha256:2d97f6910b16bd338d3060f261f53f144965f755599aab1acda1e13cf1731b1b
 
 # Set the working directory in the container
 WORKDIR /usr/src/app
@@ -17,6 +17,10 @@ RUN apt-get update && apt-get install -y \
 # Copy the requirements file into the container
 COPY requirements.txt .
 
+# Every pip install takes package versions from constraints.txt, the package list of the v2.3.1 image
+COPY constraints.txt .
+ENV PIP_CONSTRAINT=/usr/src/app/constraints.txt
+
 # Install base requirements
 RUN pip install --no-cache-dir -r requirements.txt
 

constraints.txt

@@ -0,0 +1,67 @@
+Flask==3.1.3
+Jinja2==3.1.6
+MarkupSafe==3.0.3
+PyYAML==6.0.3
+Werkzeug==3.1.8
+annotated-types==0.7.0
+anthropic==0.105.2
+anyio==4.12.1
+av==15.1.0
+blinker==1.9.0
+certifi==2026.5.20
+charset-normalizer==3.4.7
+click==8.1.8
+distro==1.9.0
+docstring_parser==0.18.0
+exceptiongroup==1.3.1
+filelock==3.19.1
+fsspec==2025.10.0
+h11==0.16.0
+hf-xet==1.5.0
+httpcore==1.0.9
+httpx==0.28.1
+huggingface_hub==0.36.2
+idna==3.18
+importlib_metadata==8.7.1
+itsdangerous==2.2.0
+jiter==0.15.0
+mpmath==1.3.0
+networkx==3.2.1
+numpy==2.0.2
+nvidia-cublas-cu12==12.8.4.1
+nvidia-cuda-cupti-cu12==12.8.90
+nvidia-cuda-nvrtc-cu12==12.8.93
+nvidia-cuda-runtime-cu12==12.8.90
+nvidia-cudnn-cu12==9.10.2.21
+nvidia-cufft-cu12==11.3.3.83
+nvidia-cufile-cu12==1.13.1.3
+nvidia-curand-cu12==10.3.9.90
+nvidia-cusolver-cu12==11.7.3.90
+nvidia-cusparse-cu12==12.5.8.93
+nvidia-cusparselt-cu12==0.7.1
+nvidia-nccl-cu12==2.27.3
+nvidia-nvjitlink-cu12==12.8.93
+nvidia-nvtx-cu12==12.8.90
+opencv-python-headless==4.13.0.92
+packaging==26.2
+prompt_toolkit==3.0.52
+pydantic==2.13.4
+pydantic_core==2.46.4
+pyzmq==27.1.0
+regex==2026.1.15
+requests==2.32.5
+safetensors==0.7.0
+scipy==1.13.1
+sentencepiece==0.2.1
+sniffio==1.3.1
+sympy==1.14.0
+tokenizers==0.22.2
+torch==2.8.0
+tqdm==4.67.3
+transformers==4.57.6
+triton==3.4.0
+typing-inspection==0.4.2
+typing_extensions==4.15.0
+urllib3==2.6.3
+wcwidth==0.7.0
+zipp==3.23.1

requirements.txt

@@ -6,7 +6,7 @@ prompt_toolkit
 numpy
 scipy
 transformers
-torch --extra-index-url https://download.pytorch.org/whl/cpu
+torch
 sentencepiece
 anthropic
 av