The Rasa integrations requires the usage of the hf command line tool. See how to install it here.


In order to work with a Rasa project, you can perform commands from within the project's directory. If not, replace hf rasa by hf rasa --dir /path/to/rasa/project in the examples below. Both Rasa1 (using markdown files) and Rasa2 (using yml files) are supported.


When importing intents from rasa, you have the choice whether you'd like to completely replace the workspace's contents by the contents of bot. This is recommended, or else duplicate intents will be present in case they were already there. Beware that this overwrites any changes made within HumanFirst Studio.

To import, run: hf rasa import --workspace [workspace id] --clear


$ hf rasa import --workspace playbook-Z6T2R76YXJB65LA62CHOTTPT --clear
Using bot in directory: '/rasa1/bot'
Bot data successfully imported into workspace 'playbook-Z6T2R76YXJB65LA62CHOTTPT'

You may now open the workspace in your browser.


To export the data back, run: hf rasa export --workspace [workspace id]. Note that this merges the changes back into the rasa project.


$ hf rasa export --workspace playbook-Z6T2R76YXJB65LA62CHOTTPT
Using bot in directory: '/rasa1/bot'
Workspace 'playbook-Z6T2R76YXJB65LA62CHOTTPT' successfully exported and merged into bot data


hf allows conversations to be imported from the Tracker Store of an existing Rasa deployment. In order for this feature to work, a few dependencies and configuration are needed.


  • The rasa command needs to be accessible from your shell.
  • The Rasa deployment needs to be configured and accessible for the machine running the command line tool. More precisely, you need a endpoints.yml file with a properly configured tracker_store section.
  • The python rasa package needs to be accessible when executing a python program in the current shell. This means that if Rasa got installed in a virtualenv, this virtual environment needs to be activated prior to executing the hf command line.


Importing conversations from Rasa’s tracker into an existing workspace is as simple as executing:

$ hf rasa import-conversations --workspace playbook-QPGRXYTO45A4PLWNCNO6BJFC

in the directory in which the endpoints.yml file is. It is also possible to specify the endpoints file from the command line:

$ hf rasa import-conversations --workspace playbook-QPGRXYTO45A4PLWNCNO6BJFC --endpoints /path/to/endpoints.yml

Note: All conversations are imported into your HumanFirst workspace at once. We don't support incremental import at the moment, but if it's a feature you are interested in, come talk to us on our Slack community.