docker/embeddings.Dockerfile aktualisiert

docker/embeddings.Dockerfile

@@ -1,29 +1,41 @@
 FROM python:3.11-slim
 
-ENV TRANSFORMERS_CACHE=/srv/.hf_cache \
+ENV HF_HOME=/srv/.hf_cache \
+    TRANSFORMERS_CACHE=/srv/.hf_cache \
     HF_HUB_DISABLE_TELEMETRY=1 \
     PIP_NO_CACHE_DIR=1 \
-    PYTHONUNBUFFERED=1
+    PYTHONUNBUFFERED=1 \
+    OMP_NUM_THREADS=1
 
-# Minimal nötige Systemlibs für numpy/scipy/sklearn/torch
+# Runtime-Libs für NumPy/SciPy/Sklearn/Torch
 RUN apt-get update && apt-get install -y --no-install-recommends \
     libgomp1 libstdc++6 libgfortran5 ca-certificates wget && \
     rm -rf /var/lib/apt/lists/*
 
 WORKDIR /srv
 
-# CPU-Torch fest (kein CUDA, kleine Wheels)
+# 1) Pip & feste Numerik/Torch (CPU)
 RUN python -m pip install --upgrade pip && \
-    pip install --no-cache-dir --index-url https://download.pytorch.org/whl/cpu "torch==2.2.2"
+    pip install --no-cache-dir --index-url https://download.pytorch.org/whl/cpu "torch==2.2.2" && \
+    pip install --no-cache-dir "numpy==1.26.4" "scipy==1.11.4" "scikit-learn==1.3.2"
 
-# Satztransformer + Uvicorn/FastAPI
+# 2) App-Dependencies (ziehen KEIN numpy-Upgrade mehr)
 RUN pip install --no-cache-dir \
     fastapi==0.111.0 \
     uvicorn[standard]==0.30.1 \
     sentence-transformers==3.0.1
 
-# App rein
+# 3) App-Code
 COPY ../app/embed_server.py /srv/embed_server.py
 
+# 4) Build-Time-Sanity-Check (bricht Build ab, falls was nicht stimmt)
+RUN python - <<'PY'
+from sentence_transformers import SentenceTransformer
+m = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
+e = m.encode(["ping","hallo welt"], normalize_embeddings=False)
+assert len(e)==2 and len(e[0])==384, "embedding shape mismatch"
+print("Sanity OK: shape", len(e), len(e[0]))
+PY
+
 EXPOSE 8990
 CMD ["uvicorn", "embed_server:app", "--host", "0.0.0.0", "--port", "8990"]