Using Rasa X with the Financial Demo

This is a guided workflow for using Rasa X in local mode with the Financial Services Example bot. Rasa X is a tool for Conversation-driven development for conversational assistants. Here, you can try it out with the Financial Services Bot to see how Rasa X works with pre-populated conversation data. In addition to talking to your bot, you can annotate messages, save conversations as training and testing examples, and share your bot.

You can refer to README.md for how to get the Financial Demo Bot installed on your machine.

Starting Rasa X in Local Mode

First, you'll need to start Rasa X in local mode. You can do that by running this command in your local project directory:

rasa x

When Rasa X has started, it'll automatically open a window in your browser in the Conversations tab.

Making NLU Annotations

Looking at the conversation marked with a red flag, you can see that the message Hello friend has been incorrectly classified with the intent transfer_money. You can tag this conversation by clicking the Tags gear on the right and selecting nlu-issue. You could choose to name the tag something else too, or create a new tag.

To make the correction, click on the 'Hello friend' message, use the drop-down menu on the right to update the intent from transfer_money to greet, and click 'Save'. The conversation can now be marked as reviewed.

Saving a Conversation Test

Click on the next unmarked conversation in the left column to review. We can see that this conversation completed without issue. The end-to-end story can be retrieved by clicking the last message in the conversation Successfully transferred $500.00 to Kyle Gardner and selecting End-to-end Story in the right column.

When using Rasa X on a server connected to Git (as opposed to local mode), you'll be able to incorporate this story into your tests automatically with the Save to conversation test button. In local mode, you can save this as a conversation test by copying the story and pasting into the tests/conversation_tests.md file.

Train Your Model

Train your model locally by clicking the Train button on the left pop-up menu.

Test Your Model

Tests your model locally with your added end-to-end test by running:

rasa test

Activate Your Model

Activate your model for use in Rasa X using the Models tab via the left menu, clicking '...' to the right of the new model, and clicking 'Make Active'. You can test this new model by going into the Talk To Your Bot and typing the message "Hello friend."

Push Your Changes

If your local project directory is connected to a Git repository, you can commit the training data changes by pushing the modified file data/nlu/nlu.md.

Next Steps

You can optionally share your bot in local mode using ngrok using the following guide.

To deploy a bot in production and access additional features, including version control integration and automated CI/CD deployment, you will need to install Rasa X on a server using either Docker-Compose or Kubernetes.