Split repo into apiserver/ and webserver/, add auth error handling and nginx proxy timeouts

apiserver/Dockerfile

@@ -0,0 +1,26 @@
+# Uses the exact same verified environment as your Dev Container
+FROM nvcr.io/nvidia/pytorch:25.02-py3
+
+# Install API and Flux-specific dependencies
+# Note: We skip the torch versions in your list because the NVIDIA base 
+# already includes the optimized sm_120 version we need.
+RUN pip3 install --no-cache-dir \
+    fastapi \
+    uvicorn \
+    pydantic \
+    transformers==4.46.1 \
+    diffusers==0.35.2 \
+    accelerate==0.30.1 \
+    sentencepiece \
+    protobuf \
+    bcrypt
+
+# Set up the application
+WORKDIR /app
+COPY api.py /app/api.py
+
+# Ensure GPU visibility (matching your devcontainer.json)
+ENV NVIDIA_VISIBLE_DEVICES=all
+ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
+
+CMD ["python3", "api.py"]