Chatbots are a giant step in transforming the way users are able to communicate directly with a website in a conversational way. They offer the ability for users to uncover answers to questions which may otherwise be hidden deep within the content or structure e of a site. The introduction of AI powered engines has improved the utility of chatbots immensely, taking them beyond providing canned answers to a set list of questions.
The management of the content which drives the chatbot has traditionally been done external to the CMS. Tyically the chatbot knowledge base would be managed within the chatbot application. Typically this interface is available to adminstrators in control of the cahtbot, making it harder for editors to update the content.
Morpht has addressed this problem by building the Quickchat Drupal module which makes it easy for editors to manage the content and sync it across to the chatbot which can then use the knowledge base to train the large language model.
What’s QuickChat?
Quickchat is a fully conversational and multilingual AI assistant powered by OpenAI's GPT-3, a 175 billion parameters state-of-the-art language model. Quickchat chat bots can recognize and speak your native language.
You can use it to automate your customer support, online applications, searching through internal knowledge base and many other tasks.
Where does Drupal fit in?
The key insight is that editors should be able to manage their content in one place - the CMS. This is foundation principle at Morpht where we conceive of the CMS is the central place for managing content for a variety of systems including the website and other tools such as chatbots. The CMS is conceived of as a Content eXperience Platform (CXP) where content driven experiences can be made available on a number of platforms and devices.
With this mindset it is natural to use Drupal as a way to manage the knowledge bases and then to provide an integration which can sync the knowledge base across to Quickchat. The Quickchat Drupal module therefore provides a set of tools which makes this process configurable and relatively straightforward to setup.
Introducing the Quickchat Drupal module
The Quickchat Drupal module has the following key concepts:
Knowledge base fields
There is nothing particularly special about a knowledgebase field. It is just a longtext field which can be attached to entities such as nodes and taxonomy terms. Generally it would be added as an "unlimited" an "optional" field on content types where editors were likely to be doing their content management. For example, knowledge base fields could be added to articles, pages, FAQs and glossaries. It is standard Drupal at this point.
With this structure it is possible to set up multiple knowledge bases if so desired. For example, content could be marked up with a taxonomy or similar to indicate that the knowledgebase content is meant for a certain chatbot model. It is therefore possible to drive many chatbots from a single Drupal instance.
Views
The Views module provides a flexible system for listing content. It is possible to filter Views in a number of ways to get the content you need. The Quickchat module therefore uses Views to allow site builders to define one or more listings for knowledge bases. Sitebuilders can define one or more Views to get the data they need to populate the models which have been defined on Quickchat.
The module, out of the box, does define a demonstration View to give you an idea of how this can be configured. You can choose to use this view if it suits your purpose, however, you may want to define something more specific to suit your needs.
Quickchat Configuration
The module also provides a way to connect Drupal to Quickchat. Multipel Models can be defined within Drupal so that you can wire up the models as they are defined in Quickchat. All you need to do is copy across the API information provided in Quickchat and then configure that into the model in Drupal.
Update and Rebuild
The module provides a permission which can be set to allow editors to have access to the Update and Rebuild functionality. Giving users this permission will allow them to push the knowledge base to Quickchat (update) and the retrain the GPT3 model (rebuild). Once these steps have been carried out the chatbot will have been updated and can then be tested live on the site.
Documentation
We have provided very detailed instructions on the Quickchat documentation pages to walk you through these steps. The team at Quickchat have also written a how to article on how the Quickchat Drupal module can be used.
Demo
The module was recently showcased at the Sydney Drupal meetup by Murray Woodman.
Thanks to Quickchat
During the development of the module Morpht worked closely with the Quickchat team who were very responsive to what we were doing. Quickchat extended their API to make the integration possible and also provided a special Drupal integration section to their service to help with the configuration. We thank them for being so open to working with us to make the integration possible.
Conclusion
Chatbots are changing how businesses communicate with their consumers and employees – and even how products communicate with users. We hope you could benefit from this module and increase conversion rate, cut down customer service costs, complement customer service agents, provide 24⁄7 support and improve internal communication.
We'd love to hear of your experiences with the module and the aways in which you are using it.
