I’m using latest version of rsa container using the following Dockerfile:
#FROM rasa/rasa:3.4.4-full FROM rasa/rasa:latest #ENTRYPOINT [“python3”]
ARG AZURE_CONTAINER ENV AZURE_CONTAINER $AZURE_CONTAINER ARG AZURE_ACCOUNT_NAME ENV AZURE_ACCOUNT_NAME $AZURE_ACCOUNT_NAME ARG AZURE_ACCOUNT_KEY ENV AZURE_ACCOUNT_KEY $AZURE_ACCOUNT_KEY
ENV PATH “$PATH:/app/.local/bin”
RUN /usr/bin/python3 -m pip install azure-storage-blob azure-identity RUN /usr/bin/python3 -m pip install pandas RUN /usr/bin/python3 -m pip install jieba RUN /usr/bin/python3 -m pip install python-git RUN /usr/bin/python3 -m pip install pyinstaller RUN /usr/bin/python3 -m pip install --upgrade pip
RUN alias python=“/usr/bin/python3” RUN alias python3=“/usr/bin/python3”
USER 1001 EXPOSE 80
CMD [ “run”, “–enable-api”, “–port”, “80”, “–model”, “models/20230801-152417-adiabatic-altitude.tar.gz”, “–remote-storage”, “azure”, “–cors”, “*”, “–debug”]
I’m filling the authentication variables during docker run.
docker build --no-cache --tag test_rasa --build-arg AZURE_CONTAINER=‘xxx’ --build-arg AZURE_ACCOUNT_NAME=‘xxx’ --build-arg AZURE_ACCOUNT_KEY=‘xxxxx’ -f Dockerfile . -t testentrypoint
Docker container have the right values on env varibles but is like python can’t load the installed azure plugins imported.
I have a python script using: from azure.storage.blob import BlobServiceClient, BlobClient, ContainerClient
Using this script from rasa container I’m able to list the container with the trained models.
Best regards,
Ricardo