First Project Guide

This page shows the necessary steps to create a simple project in Knovvu Virtual Agent for the first time user.

Login to Interface

Users can login Knovvu Virtual Agent platform from the link below.

https://eu.va.knovvu.com/

In order to login to the Knovvu VA interface, customers must enter their tenant name, user name and password.

Create a Virtual Agent

The user logging into the platform will come to the Virtual Agent page.

1. The first thing the user has to do is to create a virtual agent by clicking on the “+ New Virtual Agent” button and filling all the mandatory fields.

:::

After saving the virtual agent, “Intents” page which contains intents belong to the created virtual agent, is opened. By default, “fallback” and “welcome” intents are created within the virtual agent. In addition to these, the users can add as many intent as they want by clicking on “+ New Intent” button. Users who do not want to create intents from scratch can also use our pre-built templates.

2. In this document, as an example, a virtual agent has been created on the weather. The figure below shows the creation of the intent named 'weather'
   

3. After creating an intent the users must enter utterances, which are the training sentences about the topic.
   

4. Every intent has a purpose and a meaning, therefore the actions that can be done with this intent should be defined in the “Actions” tab. By default, the “Say” node, which gives the intent name to the end user, is placed.
   

Nodes can be dragged and dropped to change the flow and be edited by double clicking on the related node. Changes is saved automatically or user can save by clicking the “Save” button on the node edit section.

5. In this weather example, if the weather service provider wants to know the location as further information, it needs to ask to user. To ask a question to the end user, “Ask” node should be used.

6. There are different response types which designer can select from inside of the node. Such as text and quick reply.
   In this example, we want user to select within provided button options (quick reply). The user is presented with 3 options: Istanbul, Ankara and Izmir.
   

7. Next, VA’s answer according to the user’s selection should be determined. The “Say” node will be used accordingly. To customize the response, variables can be used.
   

8. The second path of the flow should also be determined in for situations where the user does not answer in the requested time. In this case the flow continues from the timeout output of the node. Timeout time is customizable from inside of the node.
   

Note that every node on the page should be linked to be able to execute the flow.
9. After completing these steps, the virtual agent is created and ready to be tested. The virtual agent can be tested by clicking on “Test” button with the speech bubble icon.
The final conversation between the user and the virtual agent can be shown below.

10. When the ”Train” on the right hand corner on the Virtual Agent page is orange, it means Virtual Agent needs to be trained. After intents, utterances and actions are added and tested, VA should now be trained by clicking on orange “Train” button.
    

It can take up to few minutes for the training to be over. Once the training is over, the virtual agent can now be used in a project.

Create a Project

1. On the Projects page, the user can create a new project by clicking on “+ New Project” button and filling all the mandatory fields.
   From this page Virtual Agent and Integrations can be added or created (check Project)

2. By default, “Start” and “End” nodes, "Condition" , "Ask" and "Virtual Agent" nodes are placed.
   First condition node checks whether this is the user's first interaction in a session (The period that encompasses the interaction between the user and the chatbot is called a "session." When a user engages with a chatbot, a session begins, and all interactions that occur during this session are related to each other). If it is the first interaction and it is not null (second condition), flow directs to the ask node and writing the welcome message.

3. Since this project is based on the scenario of a person asking about the weather, it will involve the virtual agent, trained with weather intent, to understand what the user is asking.

4. The timeout output should be determined as well in case the customer does not respond in the requested time.
   

5. If the virtual agent understands the user’s input, it will generate the response. In this case, the closing sentence to be presented to the end user is expressed with the “Say” node.
   

6. However, if the virtual agent cannot undertsand the user's input, the flow will continue from the fallback output
   

7. The final state of the flow and an example conversation that will pass between the bot and the end user can be seen below.