Question 1

How to charge my customers for my ChatGPT plugin (Stripe, Plans and Quotas)

Accepted Answer

So you created a successful ChatGPT plugin but it costs you a lot in server fees?

Well, PluginLab allows you to configure some quotas and plans to restrict your members.

It takes less than 10 minutes to charge your first customers with an elegant payment portal.

For people who prefer watching a video while they are chilling on their couch, here it is:

|| Before to get started be sure you have configured your ChatGPT plugin with OAuth on PluginLab

||| Note that at the moment OpenAI is quite unclear about policies of plugin monetization. Many users are currently charging customers using our payment feature but you are using it at your own risk.

1. Connect your stripe account

Go to https://app.pluginlab.ai/plans
Click the CONNECT STRIPE button
Follow the process on Stripe

Stripe Connect

|| Note that you cannot link the same Stripe project to more than one plugin at a time. This limitation is mainly here to avoid conflict between products of different plugins.

2. Create your first plan

Now you have linked your Stripe account we can start creating our plans.

A plan is a product that will give access your users to some set of features.

A plan is made of one or more prices and can be attached to a quota.

In this example we're going to create two plans:

Basic
Premium

| You can create as many p

Question 2

Your WeatherGPT Plugin with Next, Vercel and PluginLab

Accepted Answer

Introduction

Vercel is an incredible infrastucture platform that lets you deploy Nextjs applications in a snap.

In this tutorial we'll see how we can host a ChatGPT plugin built in NextJS and named WeatherGPT.

This plugin has been by Steven Tey and it's an amazing starter kit because it offers the following features:

AI-Plugin & OpenAPI JSON specs
Returns a link to a dynamic Next.js route
Nextjs 13 App Router
Edge Route Handlers for returning weather data
Metadata API for SEO tags
ImageResponse API for dynamic OG images at the edge

What we'll see in this tutorial

How to deploy the WeatherGPT plugin on Vercel
How to connect it to PluginLab
How to install it on ChatGPT

||| UPDATE: since we released our own templating system it's now event easier to get started with WeatherGPT directly from PluginLab. More info in this video: https://www.youtube.com/watch?v=09D8NZjbMi8&t=312s

What you'll build

This tutorial is also available on Youtube if you prefer to chill on your couch ;)

${youtube}Tutorial on Youtube

Getting started

Fork and/or clone the repository

Since we'll be deploying the project using the Vercel git integration it might be interesting for you to fork the repository directlly from your git account.

Go to https://github.com/steven-tey/weathergpt
Click on Fork
Copy the new clone URL
Clone

Question 3

RESTful API

Accepted Answer

PluginLab has a RESTful API that makes it possible for developers to interact with their member's data.

Thanks to this API you can manage your members and do some daily operations such as fetching users, updating them and more.

|| Note: this guide is describing how to use our RESTful API. If you are looking to implement it using NodeJS, we suggest you use our NodeJS SDK instead.

Authenticating your requests

The first thing to use the Admin API is to authenticate your requests.

To do so, you have to include your secret key in a header named X-PluginLab-Admin-Sdk-Secret and your plugin id in a header named X-PluginLab-Plugin-Id.

Here is an example with curl:

curl --location --request GET 'https://auth.pluginlab.ai/admin/plugins/5e3a84b73822f193bd23be935cc5b3d5/members?limit=50' \
--header 'X-PluginLab-Plugin-Id: 5e3a84b73822f193bd23be935cc5b3d5' \
--header 'X-PluginLab-Admin-Sdk-Secret: ca4424c3d050878b67c208953efca2391324571be7738279a19e2c59bc3e6b97'

Here is a Nodejs example using the axios library:

        const httpClient = axios.create({
            baseURL: "https://auth.pluginlab.ai/admin/plugins/<PLUGIN_ID>/<PATH>",
            headers: {
                'X-PluginLab-Admin-Sdk-Secret': <YOUR_SECRET_KEY>,
                'X-PluginLab-Plugin-Id': <PLUGIN_ID>,
            },
        });

|| We show you how to generate a secret in the next section

Verifying user's token

If you plan to use our RESTful API, it means you will

Question 4

How to signup my ChatGPT Plugin members using Google

Accepted Answer

Introduction

OpenAI provides three different authentication mechanisms for your ChatGPT Plugins:

HTTP Level
User HTTP
OAuth

In this article we are going to focus on OAuth since it's the most convenient way to signup new users when they install your ChatGPT plugin for the first time.

PluginLab offers many different sign-in methods such as Passwordless Email and Google.

Authenticating your users with a third party such as Google has many advantages over the simpler Passwordless Email flow.

Google provides more info (firstname, lastname, pictureUrl) than just the email address
If you users are already logged-in on their Google account, they will be able to sign-up in two clicks, so it's faster

If you prefer watching a video

If you are the kind of person who prefers watching a tutorial on your couch, then be our guest. Otherwise, keep scrolling :)

${youtube}How to setup OAuth for ChatGPT Plugin (sign in with Google)

1. Create and configure your plugin without OAuth

In this tutorial we are going to use the example plugin to get started quickly.

If you want to configure your own plugin directly I suggest you have a look at How to configure my ChatGPT plugin with PluginLab.

|| If you have any trouble configuring your plugin manifest or OAS, feel free to reach out and we'll do it with you :)

![Starting with an example plugin](https://storage.crisp.chat/users/helpdesk/website/d89a

Question 5

How do I get the user id from the API calls?

Accepted Answer

|| This article applies only if your plugin on PluginLab implements a user-http or a oAuth authentication scheme. You can get more information about how to authenticate your users here.

It's a great time to remind how PluginLab works.

PluginLab acts as a proxy layer between your backend API and ChatGPT.

This allows us to proceed with various checks, track your events, provide multiple features and last but not least garantee your users are authenticated.

PluginLab proxy schema

So, as this graphic shows PluginLab will always send you at least two HTTP headers that will be useful for you:

X-PluginLab-Event-Id: the unique id of the event request
Authorization: the authorization bearer containing the ID Token you can use to verify and get the user's info.

Once decoded, the Authorization token will provide you the following info:

The user id
The user email
The user name (if any)
The user plan and price ids (if any)

Extract the Authorization Payload / Verify the token signature

||| The id token is a JWT. It means the payload is base64 encoded. However, you should not trust the content of the payload unless you verify the token's signature.

The easiest way to verify the signature is to use one of our SDK:

[NodeSDK](/en/article/nodejs-

Question 6

Python SDK

Accepted Answer

Python SDK

Altough it's possible to rebuild your own SDK by using our PluginLab RESTful API, if you're using Python as a backend technology it will be much simpler to use our Python SDK.

To use our Python SDK, follow the instuctions below or directly go to our Github Repository.

Install

pip install pluginlab_admin

Getting started

Get started by initializing the Pluginlab app:

from pluginlab_admin import App

app = App(
    secret_key="<PLUGIN_SECRET>",
    plugin_id="<PLUGIN_ID>",
)

# to manage auth-related stuff, mostly members
auth = app.get_auth()

The SDK secret should be generated from your Pluginlab dashboard.

Verify the user token

Everytime ChatGPT will talk to your backend through the PluginLab proxy, it will send the user token in the Authorization header as a bearer token. You can verify the integrity of such token by using the verify_token method:

```python auth = app.get_auth()

usually you'd get the token from the Authorization header of the HTTP request, formatted as `Bearer <token>`

token = "eyJhbGc…dQssw5c"

try: payload = auth.verify_token(token) # from that point we know the information in the token hasn't been tampered with

# we could, for instance, check if the user has a given plan and take specific action based on that
premium_plan_id = 'KWsmKyJ

Frequently Read Articles

Browse All Categories