Creating a virtual agent
Last updated
Last updated
Before you can think of a conversation, you must have two beings: somebody to send a message and someone to receive. A virtual agent, as a part of a conversation, will receive an input from a user and answer it.
To talk to the user, the virtual agent needs an ear to receive the message, a brain to interpret and a mouth to answer. The channel is the ear and mouth, where the virtual agent will receive the message and answer it. The cognitive engine, or NLP, is the brain that will interpret that message.
In this chapter, you will learn how to create a new virtual agent or edit an existing one. You will see how to choose a channel and integrate a virtual agent to a natural language processor.
Before you start to create a virtual agent, you have to choose:
A channel
A language
A cognitive engine
If you want to integrate external analytics tools
There is a step-by-step integration guide with cognitive engines and chatbase in the appendices of this document.
Virtual agent selection screen
After you log in, if you choose to create a new virtual agent, you will be taken to the virtual agent creation page.
To create a virtual agent, select “add new bot”.
Virtual agent Creation screnn
You will need to name the virtual agent. To do so, write it in the "bot name” field. Try not to name your virtual agent by the channel you are going to expose it, for example “Facebook”, because later on you might use the same virtual agent on other channels and this name will be wrong. Instead, try naming it as “B2C”, “Providers” or other names that explain briefly what people you are targeting.
You may also give a face to the virtual agent. You can upload a jpg or png file (up to 1 MB) to be the face of the virtual agent you are building, this image will only available in the Cockpit.
It is not recommended changing the virtual agent name, as it is used to integrate it to channels.
The image might be changed without any concerns.
Important: virtual agents cannot share a name.
Channel selection list
It is not recommended to change channel names. Your virtual agent might malfunction if you rename the channel.
Then, you have to choose the language that your virtual agent will interact with its users.
Important: channels cannot share a name.
You have to select a NLP (Natural Language Processing) for your virtual agent.
The NLP function is to understand what the users are saying and each one has its strengths, weaknesses and different integration steps. A step-by-step integration guide can be found at this document appendix.
Natural Language Processing engine selection
eva can integrate with the following cognitive engines:
NTT DATA Clever
IBM Watson Assistant
Microsoft Luis
Google Dialogflow
Microsoft QnA Maker (does not support dialogs)
A developer can integrate other NLPs to eva, but this requires knowledge in programming and about eva.
eva supports external tools to monitor virtual agent performance. Currently, only Chatbase is supported.
Once you filled all the required fields, selected a channel and integrated a NLP engine, click “save bot” to create your virtual agent.
Created virtual agents
Virtual agent selection screen
When you log into eva, you can choose between creating a virtual agent or accessing an existing one.
Edit virtual agent
When you click the virtual agent menu, in the lower right corner of each agent, a few options are shown: delete and edit. All information shown in the card is about the virtual agent you are currently editing.
By clicking on this option, you can edit any of the virtual agent characteristics, such as its name (although it is not recommended), thumbnail, or its language. Remember that changing those characteristics might impact the agent’s operation. If you click the card, you will go to the agent’s control panel, which contains the Dialog Manager, Channels, Dashboard and Automated Tests.
Clicking on this option will make you delete a virtual agent.
Important: Deleting a virtual agent is an irreversible action. You will lose all virtual agent’s data when deleting it.
Here you can see the video tutorial for the virtual assistant creation:
Now you have to select the first channel that the virtual agent will use. A channel is the connection between a virtual agent and a communication app. Check the chapter to learn more. When you click, a dropdown with channels appear. Just select the channel. This step is required, since a virtual agent must have at least one channel for it to be used, but you will be able to add channels or change them afterwards. Every channel must have a name.
To integrate a NLP, check in the appendices .
To integrate with Chatbase, check the appendix . After you get Chatbase API key, insert it in the field.
