Upload folder using huggingface_hub

Dockerfile

@@ -21,33 +21,27 @@ RUN apt-get update && apt-get install -y \
 RUN curl -sS https://bootstrap.pypa.io/get-pip.py | python3.10
 RUN python3.10 -m pip install --no-cache-dir --upgrade pip
 
-# 3. Install Python requirements (Docling, Groq SDK, FastAPI, etc.)
+# 3. Install Python requirements
 WORKDIR /app
 COPY requirements.txt /app/requirements.txt
 RUN python3.10 -m pip install --no-cache-dir -r requirements.txt
 
-# 4. ✅ FIX: Pre-download all Docling/RapidOCR model weights at BUILD TIME
-#    as root so they land in a writable location.
-#    Without this, Docling tries to download them at runtime into a
-#    system dir where the non-root "ubuntu" user has no write permission.
-ENV DOCLING_ARTIFACTS_PATH=/opt/docling-models
-RUN python3.10 -c "from docling.pipeline.standard_pdf_pipeline import StandardPdfPipeline; StandardPdfPipeline.download_models_hf(force=True)" || true
+# 4. FIX: RapidOCR downloads weights at runtime into the system python dir.
+# By making the rapidocr directory globally writable, the non-root 'ubuntu' 
+# user can successfully save the weights there on the first request.
+RUN mkdir -p /usr/local/lib/python3.10/dist-packages/rapidocr/models && \
+    chmod -R 777 /usr/local/lib/python3.10/dist-packages/rapidocr
 
-# 5. Open permissions on the model cache so the runtime user can read/update them
-RUN chmod -R 777 /opt/docling-models 2>/dev/null || true && \
-    chmod -R 777 /usr/local/lib/python3.10/dist-packages/rapidocr 2>/dev/null || true
-
-# 6. Ubuntu 24.04 already has UID 1000 as user "ubuntu" — use it directly
+# 5. Ubuntu 24.04 already has UID 1000 as user "ubuntu" — use it directly
 RUN chown -R ubuntu:ubuntu /app
 USER ubuntu
 ENV HOME=/home/ubuntu \
-    PATH=/home/ubuntu/.local/bin:$PATH \
-    DOCLING_ARTIFACTS_PATH=/opt/docling-models
+    PATH=/home/ubuntu/.local/bin:$PATH
 
-# 7. Copy application code
+# 6. Copy application code
 COPY --chown=ubuntu . /app
 
-# 8. Block HF persistent user cache interference
+# 7. Block HF persistent user cache interference
 ENV PYTHONNOUSERSITE=1
 
 EXPOSE 7860