Role-Based Access Control

When you initially set up Rasa Enterprise, you created an admin user with the command:

$ cd /etc/rasa
$ sudo python rasa_x_commands.py create admin USER PASSWORD

You can use the rasa_commands.py script to create as many users as you want. These users will sign in with the username and password you created for them, and they can change their password any time they log in.

New users are automatically created when they first sign in via single sign-on.

Permission management

By default, there are three user roles defined: admin, annotator and tester. Admin users can customize these roles and assign users to them.

Use the User menu to navigate to the Manage Roles page. Here you can customize the permissions associated with each role, and also create new roles. Roles are defined by the set of permissions they grant. Here’s an overview of all permissions that can be assigned in Rasa X:

Category

“view” permissions

“modify” permissions

Test conversation

have a conversation with the assistant and see the user’s own conversation history

-

Chat logs

view all users’ conversation history and apply filters; view conversation evaluations

flag conversations

NLU training Data

view and download NLU training data and suggestions

create, delete, replace, update NLU training data; create, delete, update temporary intents and user goals

Stories

view and download stories

create, delete, replace, update stories

Responses

view response templates

create, delete, update response templates

Models

view and pull models

train models; update model training config; create, delete, update model tags; create, delete Rasa NLU evaluations

Analytics dashboard

view analytics dashboard

-

Deployment environment

view deployment environments config

modify deployment environments config

Default roles

Use the User menu to navigate to the Manage Users page to assign users to roles.

These are the permissions granted to the three default roles:

Default role

Permissions

admin

all

annotator

Test conversation: “view”; NLU training data: “view” and “modify”; Responses: “view” and “modify”; Stories: “view” and “modify”

tester

Test conversation: “view”