Dockerfile

# Use the official Python image.
FROM python:3.12-slim

# Set the working directory in the container.
WORKDIR /llava-fastapi

# Copy the current directory contents into the container at /llava-fastapi.
COPY . /llava-fastapi

# Create the cache directory and set permissions.
RUN mkdir -p /workspace/cache && \
    chmod -R 777 /workspace/cache

# Set the environment variable for the transformers cache directory.
ENV TRANSFORMERS_CACHE=/workspace/cache

# Install any needed packages specified in requirements.txt.
RUN pip install --no-cache-dir -r requirements.txt

# Make port 8000 available to the world outside this container.
EXPOSE 8000

# Define environment variable for the host.
ENV HOST=0.0.0.0

# Run uvicorn server on container startup.
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]