Merge pull request #134 from NotYuSheng/fix/embedder-model-preload

fix: pre-download sentence-transformers model in embedder-service

Author: NotYuSheng

embedder_service/Dockerfile

@@ -11,6 +11,18 @@ COPY embedder_service/requirements.txt .
 
 RUN pip install --no-cache-dir -r requirements.txt
 
+# Allow model to be configured at build time
+ARG EMBEDDING_MODEL="all-MiniLM-L6-v2"
+ENV EMBEDDING_MODEL=${EMBEDDING_MODEL}
+
+# Set a predictable cache directory for sentence-transformers models
+ENV SENTENCE_TRANSFORMERS_HOME=/opt/models_cache
+
+# Pre-download the sentence transformers model to avoid runtime download and cache permission issues
+RUN echo 'import os\nfrom sentence_transformers import SentenceTransformer\nSentenceTransformer(os.environ["EMBEDDING_MODEL"])' > download_model.py && \
+    python download_model.py && \
+    rm download_model.py
+
 # Copy utils folder from root context into /app/utils
 COPY shared_utils ./shared_utils