Dockerfile

# Use NVIDIA CUDA base image with Python support (ONLY for GPU Spaces)
FROM nvidia/cuda:12.1.1-runtime-ubuntu22.04

# Set working directory
WORKDIR /app

# Install Python and system dependencies
RUN apt-get update && apt-get install -y \
    python3.10 \
    python3-pip \
    python3.10-venv \
    python3.10-dev \
    build-essential \
    cmake \
    git \
    curl \
    wget \
    && rm -rf /var/lib/apt/lists/*

# Make python3.10 the default
RUN update-alternatives --install /usr/bin/python python /usr/bin/python3.10 1
RUN update-alternatives --install /usr/bin/pip pip /usr/bin/pip3 1

# Upgrade pip
RUN pip install --upgrade pip

# Install prebuilt llama-cpp-python CUDA wheel
RUN CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install --no-cache-dir llama-cpp-python==0.2.90 --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu121

# Copy requirements.txt and install remaining dependencies
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# Create a non-root user
RUN useradd -m -u 1000 user

# Create cache directory with proper permissions
RUN mkdir -p /app/.cache/huggingface
RUN chown -R user:user /app

# Copy app code
COPY . .
RUN chown -R user:user /app

# Create models directory
RUN mkdir -p models && chown -R user:user models

# Switch to non-root user
USER user

# Environment variables
ENV GRADIO_SERVER_NAME="0.0.0.0"
ENV GRADIO_SERVER_PORT=7860
ENV CUDA_VISIBLE_DEVICES=0
ENV HF_HOME=/app/.cache/huggingface
ENV TRANSFORMERS_CACHE=/app/.cache/huggingface

# Expose Gradio port
EXPOSE 7860

# Start app
CMD ["python", "app.py"]