Global settings for actions
Use Global settings to configure features across all actions.
On the Actions page, click Global settings 
.
Global settings provide options, configurations, and tasks for:
- Conversation routing
- Generative AI
- Autocorrection
- Display formats
- Algorithm version
- Autolearning
- Upload/Download
Conversation routing
In the Conversation routing tab, you have the following settings:
Clarifying questions
On the Clarifying questions tab, you can customize how an action asks clarifying questions.
In the Ask clarifying questions section, you can:
- Enable or disable if your assistant disambiguates (asks a clarifying question).
- Modify the text that your assistant uses to introduce the clarification list or when no action matches.
- Enable or disable response modes, and modify the text that your assistant uses with a response mode. If you enable response modes, you can use the Customize modes section to choose a response mode for each action to set how it behaves.
In the No matches section, you can affect how often your assistant routes customers to the No matches action when input is unrecognized.
For more information, see:
Change conversation topic
The Change conversation topic feature enables your assistant to handle digressions, dynamically responding to the user by changing the conversation topic as needed. For more information, see Allowing your customers to change the topic of the conversation.
If necessary, you can disable changing the topic for all actions:
-
On the Change conversation topic tab, set the switch to Off.
-
Click Save, and then click Close.
Allow change of topic between actions and dialog
If you are using actions and dialog, you can ensure that customers can change topics between an action and a dialog node.
This setting is available if you activate dialog in Assistant settings. For more information, see Activating dialog and migrating skills.
-
Set the toggle Change topics from actions to dialog to On.
-
Click Save, and then click Close.
Generative AI
On the Generative AI tab, you can enable or disable the generative AI capabilities of the assistant such as information gathering and conversational search.
When you enable the information gathering feature, your assistant uses a large language model (LLM) in watsonx.ai to intelligently recognize multiple pieces of information in the customer responses and fill the corresponding steps to avoid multiple prompts in a session.
When you enable conversational search, your assistant uses IBM watsonx generative AI model to provide conversational replies to the customer query.
For more information, see Using watsonx.ai for generative AI capabilities.
Confirmation to return to previous topic
By default, new assistants are set to ask a "yes or no" question to confirm that the customers want to return to the previous action. You can modify these settings in the Confirmation section. For more information, see Confirmation to return to previous topic.
Autocorrection
Autocorrection fixes misspellings that users make in their requests. The corrected words are used to match to an action.
Autocorrection is enabled automatically for all English-language assistants. It is also available in French-language assistants, but is disabled by default. The autocorrection setting isn't available for any other languages.
For more information, see Autocorrecting user input.
Display formats
Use display formats for variables that use date, time, numbers, currency, or percentages. You can also choose a default locale to use if one isn't provided by the client application. You can ensure that the format of a variable in the web chat is what you want for your assistant. For example, you can choose to have the output of a time variable appear in HH:MM format instead of HH:MM:SS.
Variables are formatted by using a system default unless you specify otherwise.
|
|
Description |
|
|---|---|---|
| Locale | Choose a default locale for the assistant if one can't be determined. The locale that you choose uses formats specific to that country and language. If you choose a locale, the date, time, number, currency, and percentage format fields
change to show choices specific to that locale. The system default is English - United States (en-US). |
|
| Date | For calendar dates, choose short, medium, long, full, YY/MM/DD, or YYYY/MM/DD |
|
| Time | For times, choose short or medium |
|
| Number fraction digits | Use the system default (up to 14 digits) or two digits | 10.12 |
| Number delimiter | Use the system default, none, comma, or period |
|
| Currency fraction digits | Use the system default (up to 14 digits) or two digits | 10.99 |
| Currency symbol | Use the system default or choose a global symbol | $10.99 |
|
|
Use the system default (up to 14 digits) or two digits | 10.75% |
Autolearning
Autolearning enables your assistant to learn from interactions with your customers and improve responses. For more information, see Using autolearning to improve assistant responses.
Algorithm version
Choose which watsonx Assistant algorithm to apply to your future trainings. For more information, see Algorithm version.
Upload/Download
You can upload or download actions.
Downloading
To back up actions, download a JSON file and store it. On the Upload/Download tab, click the Download button.
Uploading
To reinstate a backup copy of actions that you exported from another service instance or environment, import the JSON file of the actions you exported.
If the watsonx Assistant service changes between the time you export the actions and import it, due to functional updates that are regularly applied to instances in cloud-hosted continuous delivery environments, your imported actions might function differently than before.
On the Upload/Download tab, drag a JSON file onto the tab or click to select a file from your local system, then click Upload.
The imported JSON file must use UTF-8 encoding, without byte order mark (BOM) encoding. The JSON file cannot contain tabs, newlines, or carriage returns.