Skip to content

Main classes

The project consists of various modules, the main modules concern the tock-bot-engine engine

Story

A Story is a piece of conversation about a specific topic. It is linked to at least one intent (intent) - the StarterIntent

No "story selection" service as there can be a scxml reading library for the state machine, story routing is part of the engine in general.

Pre-defined story slot

These are StoryDefinition called at various times in the bot outside the classic flow, generally for a specific action.

  • unknownStory: Default story if no intent is detected
  • keywordStory: If a keyword is recognized in the user message, bypass the NLP and launch this story directly
  • helloStory: Launched when the bot starts
  • goodbyeStory: Launched when the bot exits
  • noInputStory: ? called if the user is inactive
  • userLocationStory : Story used for the SendLocation action
  • handleAttachmentStory : Story used for the SendAttachment action
  • keywordStory : Story used to bypass NLP with keywords

Attachments

Tock takes a specific behavior for attachments. In the case of receiving an attachment, the bot expects a SendAttachment action from the connector. NLP is then bypassed

SwitchStory

It is possible to automatically switch from one Story to another from a story using BotBus::switchStory(StoryDefinition). The story is added to the dialog as the last story and its main intent is defined as the current intent. Switching from one Story to another does not make sense for the state machine, changes are made, by definition, through a transition, never from state to state. By implementing the internal event system it is possible to have a similar behavior with the state machine, the event triggers the transition in the state machine which triggers the corresponding Story.

Intent = Story Id

The Bot uses the current intent to make the link directly with the StoryDefinition to execute, the NlpController uses the story list to check if the intent is supported by the bot,

Bot

Controller for the behavior of the bot. Calls the NLP part (if necessary) to find the intent and entities from a message and executes the story corresponding to the intent. To find the Story a direct link is made between Story and intent.

Nlp (NlpController impl)

Controller for the NLP part. Calls the NLP to identify the intent and entities of a message and saves them in the Dialog. Checks with BotDefinition::findIntent if an intent returned by the NLP is known to the bot, transmits Intent::unknown if it is not the case.

Technical-functional Tock

UserTimeline

Contains the dialog information and user data. Contains the last Action of the dialog (bot) and the last UserAction (user) Action

Dialog

Represents the conversation between the user and the bot(s). Has a DialogState object that seems interesting to introduce the state of the state machine in order to be backwards compatible.

DefinitionBuilders

Groups utility functions to instantiate new Bot and Story definitions. For Stories uses the IntentAware interface to link various pre-defined intents to the StoryDefinition that will be executed. Uses intents to retrieve the corresponding story. Dokka

  • Bot Api Client ClientDefinitionBuilders
  • Bot Engine DefinitionBuilders

Maybe useful to create simple FAQ definitions or scenarios that will be instantiated on the client side.

The definition classes

  • Bot engine:
    These are the abstractions that define the main objects (defined in the engine) of the Tock chatbot and used in the Dialog Manager, including StoryDefinition, BotDefinition This is the most interesting if you want to add new retroactive features to the entire chatbot. The default implementations are BotDefinitionBase and StoryDefinitionBase.
  • Bot Api Client : These are the implementations used when instantiating a bot Api Client : ClientStoryDefinition, ClientBotDefinition which creates StoryConfiguration and BotConfiguration when instantiating them.
  • NOTE : The definitions between the engine and the client are different. The engine (in integrated mode) has more predefined story slots in BotDefinitionBase, see above)

  • Otherwise it may be useful to override BotApiDefinition, which implements a specific BotDefinitionBase.

  • NLP Front Shared: Definition of the objects in the front: ApplicationDefinition, IntentDefinition, EntityDefinition

StoryDefinition

Interface for the objects that carry the business code. A StoryDefinition defines the actions performed when a story is executed. Holds the list of the primary intents supported by a story, and the complete list of the intents supported by the story. Checks whether an intent is supported, or whether it is a primary intent. Exposes the reference intent.

StoryDefinitionBase

Abstract implementation of StoryDefinition.

StoryStep

A step in the execution of a story. Steps define the different behaviors of a story over successive executions or different intents. Steps use the same intents as their story and have a similar structure, with a list of primary and secondary intents and an optional main intent.

BotBus

Carries the flow of information of a bot execution following a user message. A bus is instantiated for each message. Carries all the data relevant to the execution of the workflow, including the Dialog, the UserTimeline, the current story, the entities and the user action. Exposes the bot's answer API. Extended by the connectors to add specific answers.

StoryStep mechanism

The Tock engine provides the step mechanism: steps correspond to a stage in the execution of a story, and a story can go from one step to another according to arbitrary criteria, such as the current intent or the number of executions of the story. Steps are structured like stories, with primary, secondary and main intents, but these are optional. If intents are defined for a step, it is selected automatically; if a main intent is defined, the bot automatically switches to the corresponding story. Steps can also contain child steps.

Chat with Tock
×