Update Dockerfile

Dockerfile

@@ -3,10 +3,11 @@ FROM nvidia/cuda:12.1.1-cudnn8-runtime-ubuntu22.04
 ENV DEBIAN_FRONTEND=noninteractive
 ENV PATH="/root/.local/bin:$PATH"
 
-RUN echo "user:x:1000:1000::/home/user:/bin/bash" >> /etc/passwd
-RUN mkdir -p /home/user && chown 1000:1000 /home/user
+# Fix for getpwuid() error when no user is present
+RUN echo "user:x:1000:1000::/home/user:/bin/bash" >> /etc/passwd && \
+    mkdir -p /home/user && chown 1000:1000 /home/user
 
-# System dependencies
+# Install dependencies
 RUN apt-get update && apt-get install -y \
     python3.10 \
     python3.10-venv \
@@ -22,20 +23,18 @@ RUN apt-get update && apt-get install -y \
 RUN ln -sf /usr/bin/python3.10 /usr/bin/python && \
     ln -sf /usr/bin/pip3 /usr/bin/pip
 
-# Install uv and ensure it's available in this layer
+# Install uv and latest vllm[audio] from nightly index
 RUN curl -Ls https://astral.sh/uv/install.sh | bash && \
-    /root/.local/bin/uv pip install --system 'vllm[audio]' --extra-index-url https://wheels.vllm.ai/nightly
+    ~/.local/bin/uv pip install --system "vllm[audio]" --extra-index-url https://wheels.vllm.ai/nightly
 
-
-
-# Optional: install Hugging Face tools
+# Install Hugging Face tools
 RUN pip install huggingface_hub
 
-# Preload model (optional, helps with cold start)
+# Preload model weights (optional but recommended)
 RUN python -c "from huggingface_hub import snapshot_download; snapshot_download('mistralai/Voxtral-Mini-3B-2507')"
 
-# Expose vLLM API server port
+# Expose the OpenAI-compatible API port
 EXPOSE 8000
 
-# Default command to run the server
-CMD ["vllm", "serve", "mistralai/Voxtral-Mini-3B-2507", "--tokenizer_mode mistral", "--config_format mistral", "--load_format mistral"]
+# Final run command (corrected)
+CMD ["vllm", "serve", "mistralai/Voxtral-Mini-3B-2507", "--tokenizer_mode", "mistral", "--config_format", "mistral", "--load_format", "mistral"]