Dockerfile

# Use an official Python runtime as a parent image
FROM python:3.9-slim

# Set the working directory in the container
WORKDIR /app

# Install system dependencies required by some Python packages
RUN apt-get update && apt-get install -y --no-install-recommends build-essential curl

# Copy the requirements file into the container
COPY requirements.txt .

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

# Download the local LLM model during the build process to avoid runtime downloads and egress costs.
RUN mkdir -p /app/backend/models && \
    curl -L "https://gpt4all.io/models/gguf/orca-mini-3b-gguf2-q4_0.gguf" -o /app/backend/models/orca-mini-3b-gguf2-q4_0.gguf

# Copy the rest of the application's code into the container
COPY . .

# Create directories for runtime data and set permissions for the non-root user (Hugging Face uses user 1000)
RUN mkdir -p /app/papers /app/backend/jobs /app/backend/sessions && \
    chown -R 1000:1000 /app/papers /app/backend/jobs /app/backend/sessions

# Set the HOME directory for the non-root user to ensure correct cache path resolution
ENV HOME=/app

# Set Hugging Face cache directory to a writable location
ENV HF_HOME /app/.cache/huggingface
RUN mkdir -p ${HF_HOME} && chown -R 1000:1000 ${HF_HOME}

# Create and set permissions for the GPT4All cache directory
RUN mkdir -p /app/.cache/gpt4all && \
    chown -R 1000:1000 /app/.cache/gpt4all

# Tell the container to listen on the port provided by Cloud Run
EXPOSE 7860

# Run app.py when the container launches using a production-grade server
CMD ["gunicorn", "--bind", "0.0.0.0:7860", "--workers", "1", "--threads", "8", "--timeout", "0", "backend.app:app"]