Warning: This document is for an old version of Rasa X.

Enable Workflows

Once you’ve created a minimum viable assistant that can handle the most important happy paths, you want real users to start talking to your assistant as soon as possible. You should capture these real conversations to create additional training data to improve both NLU and Core accuracy. Rasa X enables you to do this with workflows to easily capture conversations for review.

Talking to Your Assistant

In Rasa X, you talk to your assistant on the Talk to your bot screen. This is similar to the rasa shell command used with Rasa Open Source, except all messages not included in the training data are collected on the Annotate new data screen for later annotation. You also have the option to save your conversation as a new story in Rasa X.

Start by conversing with your assistant on the Talk to your bot screen. If you click Save, the current story will be saved with the training data and the session will reset:


The messages that are not yet part of the training data will end up on the Annotate new data screen under NLU training:


On the Talk to your bot screen, you can also switch on Interactive Learning mode, which lets you teach your assistant while you talk to it. Next to each message, you can click Correct this, which will let you correct a certain input or action and continue the conversation from the corrected event. In Interactive Learning mode, the messages with labeled intents are automatically added to your training data instead of going to the Annotate new data screen.

Once you have a story you’re happy with, click Save to add it to the training data:


You can learn more about Interactive Learning in the Rasa Docs.

Conversations with Test Users

Once you have tested and improved your assistant by talking to it yourself, you are ready to share it with test users and have them try it out.

In Rasa X, you can create a link to share your assistant with specific people for testing before connecting your assistant to an external channel and making it live to everyone.

Your test users will be able to paste this link into their web browser and immediately start talking to your assistant. All conversations with your test users will be captured on the Conversations screen for you to review and further improve your assistant based on what you learn.

1. To share your assistant with someone, click on the Share with guest testers icon on the Models screen.


2. In the popup, you can edit what testers will see. Then you can send the link to testers who can try out your assistant.


3. Once you share this link, testers will be able to talk to your assistant using a tester view.


4. Every conversation your tester has with your assistant is automatically collected by Rasa X, and you can view them on the Conversations screen. Each tester will be assigned a unique ID that you can use to differentiate their conversations.



If you want to use Share with guest testers yourself, just open the link in an incognito tab.

Conversations with Real Users

Once your assistant has matured, you will deploy it to a wider group of users. You can connect your assistant to external channels such as Slack, Twilio, Facebook Messenger, your own website, and more. View the full list of channels supported (and how to create your own custom channel if you don’t find an existing one that suits your needs) here.

By using Rasa X you are able to capture and view the conversations that these real users have with your assistant via these external channels. All conversations can be viewed on the Conversations screen of Rasa X. The screenshot below shows an example of conversations over a SocketIO channel:


To capture conversations that happen on these channels, connect your channel to Rasa as usual by editing the credentials.yml file with the credentials from the channel you want to use. You can find the relevant credentials by selecting the channel here.

In order for Rasa X to collect the conversations, you need to change the endpoint you configure with the external channel to point to your Rasa X server. This is the URL that the channel posts to:

  • change from : http://<HOSTNAME>:<PORT>/webhooks/<CHANNEL>/webhook

  • to : http://<RASA X HOSTNAME>/core/webhooks/<CHANNEL>/webhook