Dockerfile

FROM nvidia/cuda:12.1.1-devel-ubuntu22.04

# Model choices
ARG TEXT_MODEL=llama-v3-2-3b-instruct-q6-k
ARG TRANSCRIBE_MODEL=distilled-whisper-large-v3
ARG SPEECH_MODEL=xtts-v2

# Create user
RUN useradd -m -u 1000 anachrovox

# Set home and work directory
ENV HOME=/app \
    PATH=/app/.local/bin:$PATH
WORKDIR /app

# Copy configuration
COPY config/nginx.conf /app/nginx.conf
COPY config/dispatcher.yaml /app/dispatcher.yaml
COPY config/overseer.yaml /app/overseer.yaml

# Copy WWW contents
ADD www /app/www

# Copy Anachrovox application code
ADD src/anachrovox /app/anachrovox

# Create log directory
RUN mkdir -p /app/logs

# Expose port
EXPOSE 7860

# Install packages including spaces dev mode requirements
RUN apt-get update && \
    apt-get install -y \
    bash \
    git git-lfs \
    curl wget procps \
    htop vim \
    python3-pip python3-dev \
    nginx && \
    rm -rf /var/lib/apt/lists/*

# Adjust permissions
RUN chown -R 1000 /var/log/nginx /var/lib/nginx /app

# Drop privileges
USER 1000

# Install taproot
RUN pip3 install --no-cache-dir taproot[tools,console,av]

# Install models - spaces doesn't seem to care about layer size, so we minimize overall size instead
RUN taproot install \
    audio-transcription:${TRANSCRIBE_MODEL} \
    text-generation:${TEXT_MODEL} \
    speech-synthesis:${SPEECH_MODEL} \
    --debug \
    --optional

# Copy run script
COPY --chown=anachrovox --chmod=755 run.sh /app/run.sh

# Run the application
CMD ["/app/run.sh"]