Download and caching faster whisper model only at runtime

[larsblumberg]

Hi Miguel and Jesus,

great lectures and repo!!!

I looked at this:

# Pre-download the Faster Whisper model during build
# This uses huggingface-cli which is included in the base image
RUN echo "Downloading model: ${FASTER_WHISPER_MODEL}" && \
    huggingface-cli download ${FASTER_WHISPER_MODEL} --local-dir-use-symlinks False && \
    echo "✓ Model downloaded successfully!"

in

realtime-phone-agents-course/Dockerfile.faster_whisper

Line 19 in 3c70e2e

huggingface-cli download ${FASTER_WHISPER_MODEL} --local-dir-use-symlinks False && \

and thought that this downloaded model would get lost if the container is rebuilt.

Should we add a build-time caching instruction here (via RUN ----mount=type=cache,target=....) or alternatively download the model when the container first starts via an ENTRYPOINT ["entrypoint.sh"] script in the container, mapping the download directory to a local volume via docker-compose.yaml at run-time?

I'm happy to provide a PR.