Actions - Introduction

Overview

An action is a single ability or a task that the bot performs when you give an instruction as a part of your conversation. For example, you can ask the bot to fetch your appointments for the day or ask it to schedule a meeting in your calendar.

To complete an action, the bot can prompt a few questions that are necessary to fulfil the action. For example, the conversation of a typical schedule meeting action looks like:

User : Schedule a meeting to discuss about the Zylker deal by 11 AM tomorrow.
ConvoKraft Bot : Sure, when would the meeting get over?
User : By 12 PM.
ConvoKraft Bot : Where is the meeting happening?
User : At the headquarters.
ConvoKraft Bot : Okay, I have added the meeting to your calendar.

Hence, an action is not necessarily a single message exchange between the user and the bot. It can have multiple messages exchanged between the user and the bot to achieve a single task. In the above mentioned example, the bot would schedule a meeting by executing the configured business logic.

Types of Actions

Based on the intention, actions can be classified into two types :

Directly answer a question

These actions give a direct answer to a static question asked without performing any operation or executing a business logic. They would typically get over with a single conversational exchange and they will be universal, such that their answers would be the same across users, time zones, regions or any parameters.

Example Question : What is ConvoKraft?

The response to this question would be straightforward and the same to any user asking the question, without involving any specific business logic.

You can configure these actions by defining relevant sample sentences for the actions and the answers to specific questions. When the user interacts with the ConvoKraft bot, the bot would look for these configured actions and if there is a match with the user’s query, it would respond with the direct configured answer.

Execute a business logic

These actions will instruct the bot to dynamically construct the answer for the user input by fetching relevant data from the specified data sources. This involves executing business logic specific to the action, and the answer may vary based on the data and the logic executed. The business logic execution can involve performing specified operations, typically to fetch, add, modify, or delete some data in a particular data source. The conversation may also extend to multiple messages to gather more inputs, depending on the complexity of the business logic.

You can configure these actions by defining applicable sample sentences and the relevant params for the actions of your ConvoKraft bot.

Example Questions :

1. What is in my schedule today?

To answer this question, the bot would fetch data from your calendar and would construct its answer depending on the data being fetched.

2. Schedule a meeting

To schedule a meeting, the bot would prompt and get the relevant input data from the user and would create an event in your calendar.

Action execution flow

1. Whenever a bot receives a user input, ConvoKraft tries to understand it to detect the intention of the user input. When you train a bot, the configured actions will be used by ConvoKraft to form new NLP models and enhance the existing models that help boost the intellectual ability of the bot to comprehend the intention of every action.

2. The bot would look for the configured sample sentences defined for the action. If the intention of the user input matches with any of the intentions of the configured actions, then the bot might prompt the user for more information to associate values for the params that are required to fulfil that action through a business logic, or might provide a direct response in the form of a conversation.

3. If the user asks to perform a task which doesnt have a match with any of the actions of the bot, then it would respond based on the logic configured in the Fallback message handler function.

4. Based on the configured business logic, ConvoKraft will either invoke the Webhook URL, the Deluge Function, or the Catalyst Integration Function associated with the particular action, which in turn will execute the logic.

5. Finally, the response to the action is sent as a reply to the user in the ConvoKraft bot conversationally.