Switch to GPU-enabled Docker image

- Using techdaskalos/cybersecchatbot:gpu from Docker Hub
- Enable GPU acceleration with N_GPU_LAYERS=35
- Configure MODEL_POOL_SIZE=10 for concurrent users
- Requires GPU hardware in HF Space settings

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Dockerfile

@@ -1,6 +1,6 @@
-# Use pre-built image from Docker Hub
-# Build this image locally with: ./build-and-push.sh
-FROM techdaskalos/cybersecchatbot:latest
+# Use pre-built GPU image from Docker Hub
+# Build this image locally with: docker buildx build --platform linux/amd64 -f Dockerfile.base.gpu -t techdaskalos/cybersecchatbot:latest-gpu . --push
+FROM techdaskalos/cybersecchatbot:gpu
 
 # Environment variables (already set in base image, but can override)
 ENV PYTHONUNBUFFERED=1
@@ -9,6 +9,12 @@ ENV MODEL_FILENAME=phi4-mini-instruct-Q4_K_M.gguf
 ENV USE_RAG=false
 ENV CACHE_ENABLED=true
 
+# GPU configuration - offload all layers to GPU
+ENV N_GPU_LAYERS=35
+
+# Concurrent request handling - 10 model instances for 10 concurrent users
+ENV MODEL_POOL_SIZE=10
+
 # Set Hugging Face cache to /data for persistence and write permissions
 ENV HF_HOME=/data/huggingface
 
@@ -20,4 +26,4 @@ RUN mkdir -p /data /app/models /app/knowledge_db /data/huggingface/hub /data/hug
 COPY test_interface.html /app/
 
 EXPOSE 8000
-CMD ["python", "main.py"]
+CMD ["python", "main.py"]