prefetch HF

Author: Radovan Fuchs

deploy/lightspeed-stack/Containerfile

@@ -123,6 +123,16 @@ RUN mkdir -p /opt/app-root/src/.llama/storage /opt/app-root/src/.llama/providers
 RUN mkdir -p /opt/app-root/src/.cache/huggingface && \
     chown -R 1001:1001 /opt/app-root/src/.cache
 
+# Pre-cache embedding model so HF_HUB_OFFLINE=1 works at runtime.
+# ARG defaults to 0: hermetic Konflux builds have no outbound network.
+# docker-compose passes PREBAKE_HF_MODELS=1 for local/CI builds.
+ARG PREBAKE_HF_MODELS=0
+RUN [ "${PREBAKE_HF_MODELS}" != "1" ] || \
+    ( HF_HOME=/opt/app-root/src/.cache/huggingface \
+      /app-root/.venv/bin/python -c \
+          "from sentence_transformers import SentenceTransformer; SentenceTransformer('sentence-transformers/all-mpnet-base-v2')" && \
+      chown -R 1001:1001 /opt/app-root/src/.cache/huggingface )
+
 # Add executables from .venv to system PATH
 ENV PATH="/app-root/.venv/bin:$PATH"
 

deploy/llama-stack/test.containerfile

@@ -37,6 +37,11 @@ RUN mkdir -p /opt/app-root/src/.llama/storage \
     chown -R 1001:0 /opt/app-root && \
     chmod -R 775 /opt/app-root
 
+# Pre-cache embedding model so HF_HUB_OFFLINE=1 works at runtime.
+RUN HF_HOME="/opt/app-root/src/.cache/huggingface" \
+    /opt/app-root/.venv/bin/python -c \
+        "from sentence_transformers import SentenceTransformer; SentenceTransformer('sentence-transformers/all-mpnet-base-v2')"
+
 # Copy enrichment scripts for runtime config enrichment
 COPY src/llama_stack_configuration.py /opt/app-root/llama_stack_configuration.py
 COPY scripts/llama-stack-entrypoint.sh /opt/app-root/enrich-entrypoint.sh

docker-compose-library.yaml

@@ -4,6 +4,8 @@ services:
     build:
       context: .
       dockerfile: deploy/lightspeed-stack/Containerfile
+      args:
+        PREBAKE_HF_MODELS: "1"
     platform: linux/amd64
     container_name: lightspeed-stack
     ports: