Architecture and modules

Messaging channels

Makes bot able to response on any available messaging channel.

• Connector (<Facebook>, <BotService>, ...) - transforms an incomming event
• Sender (<FacebookSender>, <BotServiceSender>, ...) - sends an outgoing event

Conversational interface

Transforms incomming messaging event to bot response and is responsible for managing conversation state.

Modules (Services)

Provides additional abilities like logging, NLP and Notifications.

• Chat Logger - Stores the history of conversations

  Available at lib/chatLogStorage.js and connected to Messaging Channel Connector at bot/index.js

• Anonymization - Filters any sensitive data in text messages

  Available at bot/anonymize.js, connected to <Ai> at bot/bot.js and to Analytics at bot/onAction.js

• Chat Analytics - By default as an integration to Google Analytics

  Available at bot/onAction.js connected to Processor at bot/Processor.js to catch incomming events and connected to Router at bot/bot.js to catch all visited interactions.

• NLP - Natural language processing service

  Wingbot NLP is built in core wingbot NPM module. It's configured in bot/bot.js in global wingbot.ai service. You can register own NLP modulte with ai.register(module) method.

• Plugins - Business logic and API integrations

  Available at bot/plugins/index.js to be able to register all required modules. Connected at bot/bot.js to BuildRouter.

• Notifications

  Complex subsystem consists of Notifications service at lib/notifications.js which uses NotificationsStorage at the same place. Notifications has own "interval" script in routes and are connected as a processor plugin to Processor at bot/processor.js.

GraphQL API

Makes wingbot.ai or other applications able to access bots services. Has own route in routes/api.js. Allows to attach or detach any API you want.