I’ve got a completed assistant and I’m moving on to Rasa X deployment to start getting testing and training feedback.
I’ve successfully followed the docker-compose installation guide and I’m running it in development for now to debug everything - I’ve run into an issue.
Docker-compose up runs successfully - no errors.
I can successfully go to http://localhost:80/ which takes me to Rasa X login, can login and follow the steps to integrate with a test repo with a super basic rasa repo in it, and I was able to train a model - so lots of the UI and container setup appears to function very smoothly.
However, once I’m at this step and I want to talk to my assistant in the conversation tab, the system seems to hang with the triple dot “waiting for response” image from the rasa side of the conversation. I’ve tested real time, interactive mode, plus resetting the conversation - and I don’t seem to be able to get a response from the bot using the project and model I’m using (which is a pretty basic bot, just out of rasa init plus one custom action) - should be no issues with this lightweight project.
So to summarize:
- Rasa X working
- docker-compose up working
- git integration via SSH working
- basic rasa project with no known issues (rasa init & train)
- can also successfully train in the rasa UI
- cannot get a response from Rasa X conversations tab
At the same time as the first initiation of this conversation, I see the logs in the following docker-compose logs.
rasa-production_1 | /opt/venv/lib/python3.7/site-packages/rasa/shared/utils/io.py:93: UserWarning: The model at ‘None’ could not be loaded. Error: http://rasa-x:5002/api/projects/default/models/tags/production
rasa-production_1 | /opt/venv/lib/python3.7/site-packages/rasa/shared/utils/io.py:93: UserWarning: Agent could not be loaded with the provided configuration. Load default agent without any model.
rasa-production_1 | /opt/venv/lib/python3.7/site-packages/rasa/shared/utils/io.py:93: UserWarning: No policy ensemble or domain set. Skipping action prediction and execution.
rasa-production_1 | More info at Policies
rasa-production_1 | /opt/venv/lib/python3.7/site-packages/rasa/shared/utils/io.py:93: UserWarning: No policy ensemble or domain set. Skipping action prediction.You should set a policy before training a model.
I am wondering the following: Do I need to tag the model for “rasa-production” to be able to talk to the assistant in the Rasa X UI correctly? I’m not personally sure if the response comes from the rasa-worker or the rasa-production images as I don’t understand the internals of the setup yet, and I’m a bit stuck.
Or alternatively is there another reason the system is not providing a response?
Any assistance would be greatly appreciated - very close to being able to deploy Rasa X and invite test users - stuck on this hurdle!