Creating your AI Chatbot Agent

Modified on Fri, 26 Apr 2024 at 02:35 AM

Yes, we're excited to get you going with creating your AI Chatbot Agent.


Remember, every AI chatbot agent has 3 building blocks:

  1. Your Knowledge Base
  2. Assistant aka the Brain of your AI chatbot agent
  3. A Widget to put in front of your users


Here, we will take you through each step in detail so you can seamlessly create and empower your AI chatbot agent with advanced functionalities.


Creating Your AI Chatbot Agent


1. Preparing your knowledge base

Every AI chatbot agent is powered by your knowledge base to serve a user's query. It can contain multiple Data Sources to feed information to the Assistant (please note: currently, AI Phone Agents don't use the knowledge base to feed themselves so there's no need to create one).


Here are the steps to create a single Data Source. 


          a. After logging in, click on the AI Agents.

     b. Click on Add Data Source. 

       c. Open the Data Source Type dropdown, and select one of the options. Provide a name and description for your reference. We support 4 formats i.e. PDF, HTTP (website crawl), images (under development), and text blob. 



  • PDF: You can upload a single or multiple PDF files at once.


  • HTTP: Provide your website URL to start a crawl, and the max number of pages to be crawled (default: 100)

  • Text: Paste the blob text from any source to add it to your Data Source. 

    d. Once you click the Submit button, it will create the Data Source and start its extraction & indexing. For this guide, we picked the HTTP option.

    e. You can see the added Data Source in the table. The status reflects the current state and will signal once it is ready to use:
  • Empty: The highlighted icon in the screenshot below shows that it's an empty Data Source. 


  • Indexing: Once the data is extracted or crawled, it will start the indexing. You'll see the status to be updated: 

  • Ready to use: Once it is ready and ready to use, it will be reflected as:

  • You can note that it was able to crawl 9 pages as shown under the Data Source Files column and a total word count of 6,419.

Now your Data Source is ready to use and feed an Assistant aka the brain of your AI Chabot Agent.


2. Setting up the Assistant


An Assistant is the main controller of your AI Agent. It is responsible for comprehending the users' requests and fulfilling them based on the available resources at its disposal.


Here are the steps to create an Assistant. 


        a. Navigate to Assistants. Make sure you have selected AI Agents from the sidebar. 



         b. Click on Add an Assistant.



           c. Here you can set up the Assistant as needed. 




  • Name: Give a name to your Assistant.


  • Description: Describe what the Assistant would do. It is for your own reference only.


  • LLM Model: Pick one of the LLMs that would generate the response. As of now, we support OpenAI GPT 3.5 and GPT 4 models. A separate guide will be published soon to cover the differences. 


In brief, GPT-3.5 is suitable for handling most queries efficiently and economically, offering faster responses. On the other hand, GPT-4 stands out in comprehending queries and producing accurate answers, albeit at a slower pace and a significantly higher cost, approximately 20 times more expensive.



  • Prompt: It is one of the most important fields that provides your Assistant with the basic instructions that control its personality, and quality of answers. It can also set the language it is expected to return the responses.

    It can also be used to overwrite any answers and to set phrases for specific cases related to your business.

    For example: You may want to respond with a particular phrase when the Assistant doesn't find relevant information in its knowledge base.


  • Enable Voice: Other than text, your users can interact with your AI Agents through voice. You can select up to 4 languages that you want the Assistant to recognize.

    For each selected language, you can then select a voice that it should use to speak back the response.

    You can also use your own voice to respond to your users by setting it up using the given link.

  • Show Images: If any images are extracted while building the knowledge base, they could be shown along with the response.


      d. Once you click the Submit button, it will create the Assistant.


Now you can build the knowledge base by linking Data Sources to your newly created Assistant. Click on the database icon as shown: 




         e. Select the Data Sources that you'd like to form the knowledge base for your Assistant.


       f. Advanced features like Form and Intent shall be covered in a separate guide.


      g. Now your Assistant is ready to use and deployed in the form of a Widget.



 3. Setting up a Widget


A Widget is the interface for your AI Agent through which your user interacts with it. 


Here are the steps to create a Widget and deploy it on your website. 


        a. Navigate to Widgets. Make sure you have selected AI Agents from the sidebar. 



          b. Click on Add a Widget.


           c. Here you can set up the Widget as needed. 



  • Name: Give a name to your Widget.


  • Description: Describe what the Assistant would do. It is for your own reference only.


  • Select Assistant: Pick one of the Assistants that would be used to act as the interface. 


  • Widget Type: You can select one out of Chat, HTML, or WhatsApp. 


Chat bubbles can be embedded on any page of your website in the bottom right corner.



HTML variants are embeddable directly in an iframe or can be accessed directly through a URL.



WhatsApp is deployed for your customers to interact through their own WhatsApp numbers. A separate guide is provided to set its configuration.



           d. Once you click the Submit button, it will create the Widget.


           e. You can find out basic information about the Widget and the Assistant it is built on. 


Also, it gives you the option to get the code, and QR, customize the design, or open the Widget that you just created in a new tab.

 


  • Get Code: This option gives you the code for your widget for both Chat bubble type and iframe by selecting from the dropdown menu. 


    Once you copy and paste the code into the respective section of your website, your AI Agent will be live. You can also generate a QR Code and use it anywhere including physical places such as at a clinic, restaurant, event and so many other places. 
  • Customize Widget: Our AI Chatbot Agents can be fully customized to suit the look and feel of where you intend to deploy them. Refer to this manual for more details.


  • Visit Demo: Click on this link to open the AI Chatbot Agent in a new tab. You can also use the same URL anywhere to make it accessible to your users.


Now your Widget is ready to use and deployed in the forms explained above. You can also try it out from the Playground section of your dashboard.






Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select atleast one of the reasons
CAPTCHA verification is required.

Feedback sent

We appreciate your effort and will try to fix the article