Converse Core

Overview

Create conversational agents with logic that's easy to work around. With this, you are able to create RULE based conversational agents.

How?

1. Define your Dialogue Definition File (DDO), that describes how the bot should interact

// booking.ddo.js
export = {
    fallbackText: "This is a fallback text",
    dialogues: {
        'bookDialogue': {
            start: 'ask-room',
            nodes: {
                "ask-room": {
                    text: "Unataka kupangisha chumba aina gani?",
                    matcher: 'room-selector'
                    with: {
                        room_options: [ 'standard', 'king', 'family' ]
                    }
                    goTo: "done"
                },
                "done": {
                    text: "Haya, imetosha. Ungependa kurudia?",
                    matcher: "yes_no",
                    goTo: '$'
                }
            }
        },
     },
     intentions: {
        'greet': {
            toMatch: ['salama', 'habari'],
            response: 'Habari zako kijana!',
        },
        'room-book': {
            toMatch: ["nataka kupanga chumba", "nataka chumba"],
            dialogueKey: 'bookDialogue'
        },
    },
}

2. Apply the DDO to the Agent

import Agent from 'converse-core'

const agent = new Agent(
    require('booking.ddo'), 
    { matchRules: ['greet', 'room-book'], },
    { name: "Dummy Agent" }
)

3. Add extra rules to apply logic needed to interact given a certain file

/**
 * Every input sent to the agent before preprocessing
 *  is trimmed of any boundary white space and the 
 *  characters are lower-cased
 */
agent.setMutation('preprocess', (input) => input.trim().toLocaleLowerCase())

/**
 * For the nodes that use a `matcher: 'room-selector'`, 
 *  apply this logic.
 * 
 * In this logic, if the the input entered isn't in the
 *  room options, return to the same node
 */
agent.dialogue('bookDialogue')
     .setMatcher('room-selector', (input, opts, context, fn) => {
         if (!opts.room_options.includes(input)) {
             // entered options isn't present.
             //  do something
             return Agent.DIALOGUE_GOTO_SELF
         }
     })

4. Use the Agent in your application

agent.chat("HABARI  ", {})

// Mutated to -> habari

// Output:
// >>{
// >>    output: "Habari zako kijani!",
// >>    state: { intent: "greet" }
// >>}

Notice

Project is still very broken, but development is in progress.

Contributions

They are welcome :)

License

MIT