Update Dockerfile

Dockerfile

@@ -10,18 +10,17 @@ ENV OPENAI_API_KEY=EMPTY
 ENV GRADIO_SERVER_NAME=0.0.0.0
 ENV GRADIO_SERVER_PORT=7860
 
-ENV VLLM_PORT=8001
-ENV MAX_MODEL_LEN=18000
+ENV VLLM_PORT=8000
+ENV MAX_MODEL_LEN=8192
 ENV GPU_MEMORY_UTILIZATION=0.90
 
-ENV NUEXTRACT_MAX_TOKENS=15000
+ENV NUEXTRACT_MAX_TOKENS=5000
+ENV NUEXTRACT_EXAMPLE_DIR=/home/user/app/examples
 ENV NUEXTRACT_ASSETS_DIR=/home/user/app/assets
 
-# Use writable cache dirs. Do not use /data unless persistent storage permissions are fixed.
-ENV HF_HOME=/home/user/.cache/huggingface
-ENV HUGGINGFACE_HUB_CACHE=/home/user/.cache/huggingface/hub
-ENV TRANSFORMERS_CACHE=/home/user/.cache/huggingface
-ENV VLLM_CACHE_ROOT=/home/user/.cache/vllm
+ENV HF_HOME=/data/.cache/huggingface
+ENV TRANSFORMERS_CACHE=/data/.cache/huggingface
+ENV VLLM_CACHE_ROOT=/data/.cache/vllm
 
 RUN pip install --no-cache-dir \
     gradio \
@@ -35,20 +34,15 @@ WORKDIR /home/user/app
 COPY --chown=user:user app.py /home/user/app/app.py
 COPY --chown=user:user start.sh /home/user/app/start.sh
 COPY --chown=user:user assets /home/user/app/assets
+COPY --chown=user:user examples /home/user/app/examples
 
-RUN mkdir -p \
-      /home/user/.cache/huggingface/hub \
-      /home/user/.cache/vllm \
-      /home/user/app/assets && \
-    chown -R user:user /home/user && \
+RUN chown -R user:user /home/user/app && \
     chmod +x /home/user/app/start.sh
 
 USER user
 
 EXPOSE 7860
 
-# Required because vllm/vllm-openai has its own entrypoint.
-# Without this, it treats start.sh as the model path.
 ENTRYPOINT []
 
 CMD ["/bin/bash", "/home/user/app/start.sh"]