Graphlogic.ai documentation

This task will help you to learn the platform and its functionality.

Registration

First of all, create an account on the Platform.

Creating a Simple Agent

First of all, you need to create a simple survey Agent. This Agent will ask the Bot user about their satisfaction with management.

Creating Agents is done on the Dashboard tab. Create a new Folder for your Agent with the name “Onboarding Task".

Create a Project and then create an Agent in it.

To create the Agent's Script, go to the Bot Builder.

Create an Agent Script.

The Agent Script should look like this: the Agent asks a question, provides buttons for answering the question, and after the Bot user clicks a button, asks the next question.

To create this Script, study the Platform Slots description and choose suitable Slots.

Use the following questions and answers:

Questions for Agent

Question 1

Text: Hello! Are you ready to take a survey on satisfaction with management?

Answers:

Yes Agent Action after Button click: Proceed to the next question

No Agent Action after Button click: Text "Alright. You can always come back later." and waiting for the Bot user's response

Bot user enters text instead of clicking a button Agent Action: Text "Please click one of the buttons." and return to the button menu

Question 2

Text: How effective is your manager's work (based on your personal assessment)?

Answers:

Extremely effective Agent Action after Button click: Proceed to the next question

Moderately effective Agent Action after Button click: Proceed to the next question

Not effective at all Agent Action after Button click: Proceed to the next question

Bot user enters text instead of clicking a button Agent Action: Text "Please click one of the buttons." and return to the button menu

Question 3

Text: How professional is your manager's behavior?

Answers:

Extremely professional Agent Action after Button click: Proceed to the next question

Moderately professional Agent Action after Button click: Proceed to the next question

Not professional at all Agent Action after Button click: Proceed to the next question

Bot user enters text instead of clicking a button Agent Action: Text "Please click one of the buttons." and return to the button menu

Question 4

Text: How attentive is your manager to details?

Answers:

Extremely attentive Agent Action after Button click: Proceed to the next question

Moderately attentive Agent Action after Button click: Proceed to the next question

Not attentive at all Agent Action after Button click: Proceed to the next question

Bot user enters text instead of clicking a button Agent Action: Text "Please click one of the buttons." and return to the button menu

Question 5

Text: How clear are the goals set by your manager?

Answers:

Completely clear Agent Action after Button click: Proceed to the next question

Moderately clear Agent Action after Button click: Proceed to the next question

Not clear at all Agent Action after Button click: Proceed to the next question

Bot user enters text instead of clicking a button Agent Action: Text "Please click one of the buttons." and return to the button menu

Question 6

Text: How willing is your manager to admit their mistakes?

Answers:

Very willing Agent Action after Button click: Text "Thank you for your answers and have a great day!" and waiting for the Bot user's response

Moderately willing Agent Action after Button click: Text "Thank you for your answers and have a great day!" and waiting for the Bot user's response

Not willing to admit Agent Action after Button click: Text "Thank you for your answers and have a great day!" and waiting for the Bot user's response

Bot user enters text instead of clicking a button Agent Action: Text "Please click one of the buttons." and return to the button menu

Training and Testing the Agent in Debug Mode

After creating the Agent Script, you need to train the Agent and test it.

Train your Agent.

If there are any errors in the Script, find the problematic Slots and fix the errors. Then train the Agent again.

After successfully training the Agent, open the Debug widget by clicking the Debug button.

Test the Script by interacting with the Agent in the Debug widget.

Deploying the Agent to Telegram

After the Agent Script is tested, deploy the Agent to Telegram.

Publish your Agent.

Create a Bot in Telegram and connect your Agent to it.

Go to your newly created Bot in Telegram and test its functionality in the messenger.

Creating an NLU-powered Bot

Next, you need to create a more advanced Agent that can understand natural language.

Create a new Project in the same Folder where the Survey Bot is located.

Create an Agent in it.

Create Intents for the new Agent.

Create 9 Intents from this

table⁠

In addition, create an Intent named "Onboarding" and provide your own Training phrases for it.

Then go to the Agent Designer tab and create the Agent Script.

The Agent Script should look like this: the Bot user asks a question, the Agent recognizes one of the Intents in the Bot user's message, provides a text response and waits for further messages from the Bot user.

To create this Script, study the Platform Slots description and choose suitable Slots.

The responses to the Intents are included in the

table⁠

Training and Testing the Agent in Debug Mode

After creating the Agent Script, train the Agent and test it in the Debug widget.

Deploying the Agent to Telegram

After testing in the Debug widget, publish the Agent, deploy it to Telegram and interact with it through the messenger.

Creating a Bot from Excel

The Platform allows you to import an Agent Script from an Excel file.

Create an Agent Script in Excel that is similar to what you have already created in the Creating an NLU-powered Bot section.

Import the file into the Platform.

Training and Testing the Agent in Debug Mode

After creating the Agent Script, train the Agent and test it in the Debug widget.

Deploying the Agent to Telegram

After testing in the Debug widget, publish the Agent, deploy it to Telegram and interact with it through the messenger.

Working with Analytics

After you have created three Agents, trained them, deployed them to Telegram, and tested them, look through the analytics of their interaction with Bot user.

One hour after the end of the interactions, go to the Analytics tab.

For the simple survey bot, export:

Chat history from the Telegram channel;

Slot stats, and highlight the number of times the Slot containing the first question was completed.

For the NLU-powered bot, export:

Chat Context, and highlight the name of the last recognized Intents in Chats;

Agent Usage Report, and highlight the total number of Chats in the period.

For the bot from Excel, export:

Stop slots report.

Registration

Creating a Simple Agent

Questions for Agent