Use a data source with your Interfaces AI Chatbot

Beta

Zapier Interfaces is a beta product. It’s available for use but still in active development and may change.   

With a data source, you can connect your Interfaces AI Chatbot to your own knowledge sources and tailor the responses for your business or project. You can restrict your bot from using its training information to provide answers and set custom responses when information doesn't exist.

Note

Connecting a data source to your chatbot is a feature only available to Interfaces Premium subscribers. 

Add a data source

1. On any page of an Interfaces project, click the plus sign at the bottom of the page to add a new component.
2. In the component menu, select Chatbot.
3. In the dialog box that appears, click Yes to confirm. The sidebar will open.
4. Click the Data tab.
5. Click + Add source.
6. Click Browse files, then find and select the source file you want to use. 
7. Click Add source. It may take a moment to process, depending on the file size.

Note

You can learn how to set up other features of a chatbot in our AI Chatbot help guide.

Limitations

• Maximum file size: 1MB.
• Accepted file formats: .txt, .csv, .json, .pdf, .doc, .docx
• Maximum number of files: one

Adjust chatbot behavior for missing answers

If the chatbot uses a data source, you can define how it will respond when the data source has no results that match a specific question. 

To change how the chatbot behaves:

1. On a project page, click your chatbot component. The sidebar will open.
2. Click the Data tab.
3. Click  the When results are not found from the data source field, then select an option from the dropdown menu:

• Generate an AI response without data source: the chatbot will generate a response using the connected OpenAI model. It will not use additional context provided by the connected data source. This is the default setting.
• Show a custom message: write your own message that will always be displayed as the response if the chatbot cannot find relevant knowledge within the connected data source. 

Tip

You can use a custom message to direct your visitor to other sources of help, such as a contact form. 

Best practices for writing directives when using a data source

A directive is how you instruct the chatbot to interact with your visitors. You can improve the chatbot's responses by adding information such as names of fields in the file or how to identify an answer.

Currently, the chatbot looks for a double space - or line break - to indicate separate pieces of information (different groups of questions and answers, for example) in a data source file. So when you create a file, remember that the AI will process it better if related information appears together on the same line or within the same paragraph.

When using a .csv file, it automatically uses line breaks to separate rows, so each row is seen as a separate set of information.

If you are using .doc or .txt files, you must separate all related information using one line break (enter key).

Note

Zapier has begun incorporating certain OpenAI features/functionalities into its product. How OpenAI handles data sent via these features/functionalities is governed by Zapier’s separate contract with OpenAI.

To protect our customers, we have obtained OpenAI's agreement not to fine-tune/model train with customer content.

Provide feedback and get help

You can make a feature request, provide feedback on existing features and get help from the Interfaces team.