gateflow v1.1.1

GateFlow

GateFlow is a light weight library to manage state and enforce order for flows that span multiple multi-requests.

Installation

Install gateflow from npm.

npm install gateflow

Usage

GateFlow and GateFlowStore works out of the box with redis, but will work with storage that implements the Store interface exported by the package.

Import

import {GateFlow, GateFlowStore} from "gateflow";
import redis from "redis";

Configure GateFlowStore

Create a random secret for keys, and a ttl that represents how long a flow should be active for before it expires (in seconds.)

const client = redis.createClient({ /* your configuration */ });
const secret = "your_random_secret";
const ttl = 300; // seconds

const store = new GateFlowStore(client, secret, ttl);

For testing purposes you can a local storage implementation to follow along.

Note: Local storage is not recommend for not for production use.

import { Store } from 'gateflow';

class LocalStore implements Store {
  private values: Record<string, string> = {};

  del(key: string, cb: Callback<number> = () => undefined): boolean {
    delete this.values[key];
    cb(null, 1);
    return true;
  }

  get(key: string, cb: Callback<string | null> = () => undefined): boolean {
    const value = this.values[key];
    if (value !== undefined) {
      cb(null, value);
    } else {
      cb(null, null);
    }
    return true;
  }

  set(key: string, value: string, mode: string, duration: number, cb: Callback<"OK" | undefined> = () => undefined): boolean {
    this.values[key] = value;
    cb(null, 'OK');
    return true;
  }

  quit() {}
}

Configure GateFlow Service

GateFlow takes an instance of GateFlowStore, and a Schema and returns a service for creating and managing flows.

Add the schema either inline or by using the static method buildSchema().

const gateFlow = new GateFlow(store, [
	["login", ["login"]],
	["send_mfa", ["login",  "send_mfa"]],
	["verify_mfa", ["login",  "send_mfa",  "verify_mfa"]],
]);

const schema = GateFlow.buildSchema([
	["login", ["login"]],
	["send_mfa", ["login", "send_mfa"]],
	["verify_mfa", ["login", "send_mfa", "verify_mfa"]],
]);
const gateFlow = new GateFlow(store, schema);

Schema

Building a schema for your flow requires planning which routes should be accessible and which routes should not. A flow consists of an array of "gates," with each "gate" consisting of a "name" (the first element), and an array of available gates (the second element) at a given "gate."

[
	["login", ["login"]],
	["send_mfa", ["login", "send_mfa"]],
	["verify_mfa", ["login", "send_mfa", "verify_mfa"]],
]

In this example, each "gate" is a request a user must pass through, in order, to progress through an MFA flow on login. The flow might go something like this:

1. A user logs into your application; they start at gate login. If they successfully provide a username/password, the user progresses to the "next" gate send_mfa, and returns a key to the frontend.

2. The user is presented with destinations for MFA (phone or email) and sends a request with their selection and the key. If this is valid, the application sends a code to their destination and the user progresses to the "next" gate verify_mfa.

3. The user is presented with a form to enter the code sent to their destination. They send a request with the code and the key. If the code is valid, the user progresses to the "next" gate, which there is none, so the flow is complete and GateFlow invalidates the key.

Looking at the schema, we can see the first element (the gate's "name") aligns closely with the requests made by the user. The second element represents the other gates available to change to at a given gate.

For example, in an MFA flow, it is important that a user not be able to jump from the send_mfa gate directly to the complete_mfa gate. Additionally, if the user is at the gate verify_mfa, it would be valid for the user to go back to the send_mfa gate if they needed to resend their code again. This is how GateFlow enforces order across requests.

Example

Using the flow presented in the above section Schema.

const gateFlow = new GateFlow(store, [
	["login", ["login"]],
	["send_mfa", ["login", "send_mfa"]],
	["verify_mfa", ["login", "send_mfa", "verify_mfa"]],
	["complete_mfa", ["login", "send_mfa", "verify_mfa", "complete_mfa"]],
]);

Login

1. A user logs into your application, they start at gate login. If they successfully providing a username/password, the user progresses to the "next" gate send_mfa, and returns a key to the frontend.

After validating the user's credentials, we collect the available destinations for their MFA to be sent. Here, we can store the destinations in StepFlow so they can be verified on the next request. We can also store the token that would have gone back to them, if not for MFA.

async function(req, res, next) {
	let token = "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...";
	let destinations = [{
		type: "phone",
		value: "(503) 555-7392"
	},{
		type: "email",
		value: "gateflow@tinkertamper.com"
	}];
	
	let gatekeeper = await gateFlow.create({token, destinations});
	await gatekeeper.next();
	res.send({
		key: gatekeeper.key,
		destinations: destinations.map(dest => dest.type)
	});
}

We create a new gatekeeper, and optionally initialize it with the token we will be giving the user at the end of the flow and the list of available destinations the user will be choosing from in the next request. After all processes have been successful, we call next() to progress the "gate" from login to send_mfa - the next gate in the flow. Lastly, we return the key to the client for use on subsequent requests.

Verify MFA

2. The user is presented with destinations for MFA (phone or email) and sends a request with their selection and the key. If this is valid, the application sends a code to their destination and the user progresses to the "next" gate verify_mfa.

On the next request, we will receive the key and an MFA destination passed as the desired index in the destination's array. We must validate the destination passed, send that destination an MFA code, and move the gatekeeper forward.

async function(req, res, next) {
	const key = req.body.key;
	const index = req.body.destination;

	let gatekeeper = await gateFlow.continue(key);

	let [isValid, errors] = await gatekeeper.test("send_mfa");
	if (!isValid) {
		throw new Error(errors.join(", "));
	}

	await gatekeeper.knock("send_mfa");

	let {destinations, token} = await gatekeeper.getData();
	let code = await sendCode(destinations[index], token);
	await gatekeeper.setData("code", code);
	
	await gatekeeper.next();
	
	res.send({});
}

On this request, and all subsequent requests, we will be using the continue() method on the gateFlow service to get our gatekeeper object.

Before we begin servicing the request, we can use test() to determine if the gate associated with the request is valid, given the current state of the flow. Also available is the knock() method, which allows you to set the current gate, which also does a test and will throw if the gate is not valid. To illustrate why knock() and test() are useful, let's look at our schema.

The current gate is send_mfa (set via the next() call at the end of the first route). In our "schema" we defined mfa_send as the following:

["send_mfa", ["login", "send_mfa"]]

Calling test(status) on send_mfa verify status is either login or send_mfa. Calling knock(status) will verify status is either login or send_mfa and then set the current gate to that status.

Back to the code, we retrieve the destinations we stored from the first request, and that "trusted" data to send an MFA code to the selected destination. From that function, we get the code that was sent, which we then give to the gatekeeper for use on the next request.

Lastly, and after everything was successful, we progress to the next gate by calling next().

Verify MFA

3. The user is presented with a form to enter the code sent to their destination. They send a request with the code and the key. If the code is valid, the user progresses to the "next" gate complete_mfa, the user progresses to the "next" gate, which there is none, so the flow is complete and GateFlow invalidates the key.

On this request, the user supplies the code they received and the key from the original request.

async function(req, res, next) {
	const key = req.body.key;
	const code = req.body.code;

	let gatekeeper = await gateFlow.continue(key);
	let [isValid, errors] = await gatekeeper.test("verify_mfa");
	if (!isValid) {
		throw new Error(errors.join(", "));
	}

	await gatekeeper.knock("verify_mfa");

	let cache = await gatekeeper.getData();
	if (cache.code !== code) {
		throw new Error("Invalid MFA code");
	}

	let [isComplete, {token}] = await gatekeeper.next();

	if (isComplete) {
		res.send({token});
	} else {
		res.sendStatus(401);
	}
}

Similar to the last request, we use continue to get a gatekeeper object and then test that the current gate requested is valid with test() and knock(). In this request, verify that the code passed matches the code we have stored.

At the end, the request with a call to next to progress the next gate. The next method returns a boolean for isComplete and the session object we have stored as a part of the flow. Because this is the last gate in our schema, the gatekeeper will invalidate the key and its cache from our store. We then can pass back the token we stored from the original request to complete our MFA flow.