Create Cognigy-4 Extension

This repository will become the new 'custom-modules' repository for Cognigy.AI 4.0.0. We are no longer using the name 'custom-module' but the new term 'extension'.

Currently, extensions can only contain flow-nodes and connections, but we want to add more functionality later on.

• Download – How to install a template project.
• Development – How to code a new Extension.

This script works on macOS, Windows, and Linux.

Creating an Extension

npm init @cognigy/extension my-extension basic
cd ./my-extension

Run those two simple commands to install a new basic template project and get started with development. You can use additional arguments

Get Started Immediately

You do not need to install or configure any additional tools. They are preconfigured and hidden so that you can focus on the code.

installing with npm init @cognigy/extension <arguments> <my-extension-name>

Use this command and you are ready to go. Here are the arguments you can use:

• example gives you an example extension
• empty gives you an (almost) empty project
• basic gives you a basic project with only very few, basic elements
• no argument also gives you the basic project

building with npm run build

Builds the app for production and recreates a .tar.gz archive to be seamlessly uploaded.

deploying with npm run deploy

(not yet working!)

advanced Use the Cognigy API to upload your extension without using the website.

Development

How to get started?

If you want to build an extension, please first have a look at the 'example' project template as it contains the code for an extension which includes multiple flow-nodes and uses most of the important concepts we have introduced with the new 'extensions' functionality in Cognigy.AI 4.0.0. Install the 'example' project template with npm init @cognigy/extension example example-project

@cognigy/extension-tools

createNodeDescriptor

The method createNodeDescriptor is one of the most central methods exposed by the 'extension-tools' package. It is recommended to only use one createNodeDescriptor per file (unless it has children), since it is now possible to include multiple descriptory files per Extension.

Defining a node essentially looks like this (you can find this example in 'example/src/nodes/reverseSay):

import { createNodeDescriptor, INodeFunctionBaseParams } from "@cognigy/extension-tools";

export interface IReverseSayParams extends INodeFunctionBaseParams {
	config: {
		text: string;
	}
}

export const reverseSay = createNodeDescriptor({
	type: "reverseSay",

	defaults: {
		label: "Simple Reverse Say",
		comment: "Reverses the given string and sends it.",
		config: {
			text: "{{ci.text}}"
		}
	},

	fields: [
		{
			key: "text",
			label: "The text you want to reverse.",
			type: "cognigyText"
		}
	],

	function: async ({ cognigy, config }: IReverseSayParams) => {
		const { api } = cognigy;
		const { text } = config;

		const reversedText = text.split("").reverse().join();

		api.say(reversedText);
	}
});

Let's analyze the different pieces we can see in the code-window above:

import { createNodeDescriptor, INodeFunctionBaseParams } from "@trash-planet/extension-tools";

Here we are just importing the createNodeDescriptor method from the extension-tools package. We are also importing a Typescript interface we want to use.

export interface IReverseSayParams extends INodeFunctionBaseParams {
	config: {
		text: string;
	}
}

In this part of the node-code we are crafting a new Typescript interface for the config (the arguments) of the new flow-node. We advise you to do this for your nodes as you will have a better developer experience when you are actually creating the code (aka the function) of your flow-node. Make sure that you extend on our INodeFunctionBaseParams interface - this will give you access to:

• cognigy object with sub-properties:
  • api: This exposes the APi your flow-nodes can use.
  • input: The Cognigy Input Object, this is a proxy so you can set/get values.
  • context: The Cognigy Context Object, this is a proxy so you can set/get values.
  • profile; The Cognigy Contact Profile Object, this, too, is a proxy so you can set/get values.
• nodeId the actual id of the node that is being executed
• config this is the config/arguments of your node - that's why you should overwrit this in your own Interface
• childConfigs[] in case your node has children, this array contains those children's configs

export const reverseSay = createNodeDescriptor({
	type: "reverseSay",
	defaultLabel: "Reverse Say",
	fields: [
		{
			key: "text",
			label: "The text you want to reverse.",
			type: "cognigyText",
			defaultValue: "{{input.text}}"
		}
	],
	preview: {
		type: "text",
		key: "text"
	},

	function: async ({ cognigy, config }: IReverseSayParams) => {
		const { api } = cognigy;
		const { text } = config;

		const reversedText = text.split("").reverse().join();

		api.say(reversedText);
	}
});

Now comes the actual definition & implementation of your flow-node. You essentially pass in an object into createNodeDescriptor and it will create a full node-descriptor for you. The method will fill-up on properties you don't set with compatible default values. Let's discuss some of the properties you have to set:

• type: This is the type of your node. Every node needs to have its own type.
• defaultslabel: The default label of your node is the name that is displaed in the flow by default.
• fields: This section defines the user interface which will get generated for your node-config. You have to add a field definition per key in your config object. You reference the key in your config using the key property in the field definition. The label is used in the UI as well. The type gives you a variety of possibilities - you can e.g. say that your field should be of type cognigyText or e.g. of type json or toggle.
• preview: (optional) Displays the value of one of the node's config-fields.
• function: This actually contains the code of your flow-node. Ensure that you add the async keyword. The execution engine will always await the execution of your node. We have full Typescript support, please e.g. check the typings of the cognigy object as we don't have full documentation, yet.

createExtension

Whereas the createNodeDescriptor method essentially just fills-up default-values for your flow-nodes, it's the createExtension methods job to bundle all nodes and connection definitions into one large object which will then be passed into the Cognigy.AI system.

Make sure that you are using the default export syntax for the createExtensions return-value as we need to be able to find the complete object.

It is highly recommended to always create a module.ts file which essentially looks like the one in example/src/module.ts. Please only import all nodes and connections into this file and assign them to the payload you pass into the create extension method.

import { createExtension } from "@cognigy/extension-tools";

/* import all nodes */
import { reverseSay } from "./nodes/reverseSay";
import { executeCognigyApiRequest } from "./nodes/executeCognigyApiRequest";
//if multiple nodes are in one script, import all of them
import { randomPath, randomPathLeft, randomPathRight } from "./nodes/randomPath";
import { fullExample } from "./nodes/fullExample";

/* import all connections */
import { apiKeyConnection } from "./connections/apiKeyConnection";

export default createExtension({
    //define the nodes to be exported
	nodes: [
		reverseSay,
		executeCognigyApiRequest,

		randomPath,
		randomPathLeft,
		randomPathRight,
		fullExample
	],

    //define the nodes to be exported
	connections: [
		apiKeyConnection
	]
});

Connections

In Cognigy 3 we have introduced the concept of Secrets which allow you to securely store configurations. Secrets do no longer exist in Cognigy 4, but were replaced with the Connections.

Extension now have a stringer relationship with their cocrresponding flow-nodes. They can be assigned using configuration fields in the flow node.

If you want to use a connection within a node, you have to do the following:

• add a field with type: connection to your flow-node, you can see this in 'example/src/nodes/executeCognigyApiRequest.ts'

  {
      key: "connection",
      label: "The api-key connection which should be used.",
      type: "connection",
      params: {
          connectionType: "api-key" // this needs to match the connections 'type' property
      }
  }

• create a connection definition, see 'example/src/connections/'
• add the connection into your createExtensions call - if you miss it there, it will not be there.

You have to ensure that the params.connectionType in your flow-node field definition maps to a connection and a proper type of the connection. You can define multiple connections within a single extension and the type field is used to find the correct connections that satisfy the fields your node requires.