17 Experts Share Their Experience Developing ChatGPT Plugins

Introduction

Welcome to our new blog post series! We’re taking a deep dive into the world of ChatGPT plugins and their development. We’ve had the privilege of sitting down with some of the most innovative and experienced minds in the field, who have generously shared their insights, processes, and advice for those who aspire to join this exciting realm of technology.

From understanding the workings of GPT-3 and GPT-4 APIs to conceptualizing and implementing a plugin, our interviewees have covered it all. They’ve also shared their best practices, emphasizing the importance of letting ChatGPT do the talking, and using plugins only when necessary.

So, whether you’re a seasoned developer or a curious beginner, there’s something in this post for you. Get ready to roll up your sleeves, learn from the best, and perhaps even start crafting your own ChatGPT plugins. Let’s dive in!

Plugin Development Process

Although the development process can vary depending on the type of plugin, there are a few critical points in the process. These include:

1. Developers need to understand how the GPT-3 or GPT-4 API would work with the model, depending on their preferences.
2. Before designing the plugin, developers need to provide a clear concept of it. This includes defining the input/output, charting the functionality, and planning the prompt handling capabilities.
3. Once the design is ready, the next step is plugin implementation. This is where developers write the code and use APIs if needed. OpenAI’s code interpreter and web browsing plugin with open-source code bases are good examples of how to proceed.
4. After creating the API’s manifest file and OpenAPI specification, you can connect it to the ChatGPT system via the ChatGPT UI. You have to add the plugin manifest file to yourdomain.com/.well-known/ path. Once done, you can start testing your API.

The goal is to empower ChatGPT to communicate with APIs crafted by the developers, thereby expanding its functionality and enabling it to execute a broad spectrum of tasks. Here are a few responses from developers regarding the general process of development:

“As the API was already there and was described based on the Open API specification, the development process was really short and consisted more of testing if ChatGPT understood the API and could use it to provide meaningful answers.” – Bob Coret, Developer of Open Archives Plugin

“As an API-first company, we already had specifications for all of the endpoints we planned to make available, which made the development process significantly easier. However, I found that I needed to tune some of our descriptions and examples to make them easier for ChatGPT to understand. Having a local environment where I could make changes to our descriptions and quickly validate them was crucial. I used the same series of prompts each time I made a change that I could granularly understand what improved performance and what didn't.” – Katie Adams, Chief Product Officer at Polygon.io

“The development of the initial version of our plugin didn't take much time since we already had functioning API methods that we required. We integrated our API into a ChatGPT plugin, following the guidelines. Subsequently, we successfully completed the review process, and our plugin became available in the store.” – Mariia Yehorenko, Project Manager at Serpstat, Developer of Serpstat SEO Tool

Biggest Challenges Faced During The Development Process

As we delve deeper into the world of ChatGPT plugin development, it’s crucial to acknowledge that the journey isn’t always smooth sailing. In this section, we turn our attention to the challenges that developers often encounter along the way. From dealing with slow response speeds to managing the complexities of natural language interpretation and API management, we’ll discuss these hurdles in detail.

Our aim is not to discourage, but to prepare you for the realities of the field. After all, every challenge is an opportunity for learning and growth. So, let’s explore these obstacles, understand their implications, and discuss strategies to overcome them. Here are a few responses from developers on the biggest challenges they faced during their development processes:

“During the development process, I faced a few challenges that were quite daunting. One of the biggest challenges I encountered was trying to create a seamless user experience that was both intuitive and easy to use. I overcame this challenge by conducting extensive research on best practices and user behavior and then implementing these findings into the design of the product. Another challenge I faced was trying to ensure the safety and security of the product. This was particularly challenging because there are always new threats and vulnerabilities that arise, and it's important to stay ahead of the curve. To overcome this challenge, I worked closely with security experts and implemented rigorous testing and monitoring procedures to ensure that the product remained secure at all times.” – Illya Liudogovskyi, Developer of UKR School Books Plugin

“Since ChatGPT executes the API freely, there were many cases where it did not execute as intended; I rewrote the APi documentation and rewrote it to work without entering the required fields as APi, so that ChatGPT's behavior would work no matter what.” – Takagi Yuusuke, Founder of Automation co, Ltd.

“The hardest part was OpenAI's service reliability — the plugin store and set up often goes down, and is quite finnicky.” – Benjamin Hendricks, Developer of Ndricks Sports

---

“The biggest challenges that I faced throughout the development process were simply figuring out new ways to deal with issues that very few people had dealt with before. When I published the highPerplexity ChatGPT plugin, there were only about 100 other plugins in the plugin store and users couldn't even search the store to find the plugins that they wanted!”

— Nathan Clay, Founder and CEO of the HighPerplexity Prompt Engineering Platform

---

Essential Knowledge And Skills For Developing Plugins

As with most web development, there are some key skills that you need in order to develop your plugin for the ChatGPT Plugin Store. We asked expert developers what those skills are, and here’s how a few of them responded: 

“You need to know Python and Javascript and their frameworks for Web Development and server side. And you need to know about RestAPI and OAuth as well which i was expert in already so never faced any issue.” – Haseeb Mir, Developer of Code Runner Plugin

“You should also be extremely, ludicrously, super familiar with utilizing APIs (particularly APIs that enable interaction with LLMs).” – Nathan Clay, Lead Frontend Engineer at Trinity One

“Understanding how the language model interacts with your API spec and making sure the spec is written out very descriptive.” – Adnan Ahmad, Developer, Shuto.io

---

“Knowledge of python, in my case Django (but it could be Flask or FastAPI), and some web hosting basics (DNS, SSL, certificates)”

— John Capobianco, Developer Advocate at Cisco Training Bootcamps

---

Most Rewarding Aspects of the Plugin Development Process

Now, we’ll focus on the lighter side of the plugin development process – the most rewarding parts. In this section, developers will say which parts of the process brought them joy and made them realize that the entire process was worth it. Here are some of their responses:

“Solving a problem or filling a gap in the market is one of the most valuable aspects when developing SignalPlus. By providing a powerful and easy-to-use tool, SignalPlus can help users solve some of the problems and challenges encountered in crypto market transactions, and fill in the gaps or deficiencies in existing software or applications.” – Kai, Developer of SignalPlus Plugin

“Seeing it out there on the Plugin store and having people get benefit from it.” – Ruze Richards, Developer of Diagram Plugin

“Aside from learning about LLMs, the most rewarding aspect is seeing so many users trying out the plugin and being able to learn from their queries. It was also a great moment when I was first able to point the plugin back to its own source code, which allowed GPT-4 to find the root cause of a problem I had been unable to resolve for 2 hours.” – Viktor Ferenczi, Developer of AskYourCode Plugin

---

“Receiving feedback from users or interested parties that they found me through my plugin. That means it did it's job!”

— Benjamin Hendricks, Developer of Ndricks Sports Plugin“

---

Advice for Beginners and Mistakes to Avoid

As we continue our exploration of ChatGPT plugin development, it’s time to address an equally important aspect – the common pitfalls and mistakes to avoid. This section is dedicated to helping you navigate around these potential stumbling blocks, ensuring a smoother and more efficient development process. Here’s some advice that successful developers have for those who are just starting:

“If you already have an API, described by an OpenAPI description, then testing this as a plugin is easy. The OpenAPI description of your API is loaded each time ChatGPT is make a request to your API, so changes (in description or parameters) are picked up immediately.” – Bob Coret, Developer of Open Archives Plugin

“Try to solve a real problem that you are passionate about. I think there’s tons of plug-ins that are fun, but aren’t actually improving processes or solving problems for people. The road may be tough but if you love the issue you are trying to solve, it makes things so much easier!” – Landon Garrison, Developer of Tutory Plugin

---

“Start with a clear vision of what you want your plugin to achieve. Break down your project into manageable tasks, and don't be afraid to ask for help or look for resources online or ChatGPT when you're stuck. Always test your plugin extensively.”

— Edmund Cinco, Developer of CSV Export Plugin

---

Now, our interviewees will go over some common mistakes that are important to avoid during the development process:

“Avoid unnecessary development and strive to simplify things. Always attempt to optimize descriptions so they're understandable to ChatGPT. The response should be crafted in a way that ChatGPT can interpret it, as it will then convert it into something human-understandable.” – Dewmal Handapangoda, Senior Software Engineer at E-Green

“New developers should come up with a simple idea and just focus on solving one immediate problem at a time until things start coming together. Pasting code snippets and documentation into ChatGPT and asking for advice can be extremely helpful for this.” – Nathan Clay, Lead Frontend Engineer at Trinity One

---

““Don't rush! OpenAI is slow to communicate back many times. Don't make the plugin descriptions hard to understand, the endpoints confusing, etc. Don't be cute, don't be complicated. No one is going to see your code except for a chatbot so let the chatbot show off what you provide instead.”

— Greg Howe, Developer of Gimmee Air Quality, Chat Tool Finder Plugins

---

Methodology of the Survey

As we began to explore the rich terrain of ChatGPT plugin development, we knew that the best way to do so would be to glean insights from those who have firsthand experience in this domain. Thus, we decided to conduct a survey, reaching out to a large cohort of seasoned plugin developers who have successfully launched their plugins on the ChatGPT Plugin Store. The details of our approach are outlined below.

Selection of Participants

We sent out survey invitations to a group of 600 developers, all of whom had a plugin live on the ChatGPT Plugin Store at the time. This criterion ensured that our participants not only had the theoretical knowledge of plugin development, but also the practical experience of seeing their plugins through the process of review and approval.

Survey Design

The survey was designed with an aim to draw out the participants' experiences, challenges, and advice related to developing plugins for ChatGPT. We ensured that our questions covered a broad spectrum of aspects – from understanding the workings of GPT-3 and GPT-4 APIs, to conceptualizing and implementing a plugin, to the best practices they've adopted.

Response Collection and Analysis

Responses to the survey were collected over a period of two weeks, after which we undertook an in-depth analysis of the feedback. Each response was carefully reviewed to extract meaningful insights and compile the diverse array of tips, tricks, and advice shared by the developers. 

Integrity of Results

We believe that our results present an accurate and valuable overview of the experience of ChatGPT plugin developers, as well as the practices they adopt for successful plugin development.

Each participant in the survey was selected based on their demonstrated experience, and their responses were analyzed in a consistent and unbiased manner. 

Please note that while these insights are instructive and enlightening, they represent the opinions and experiences of a specific group of developers. As the field of ChatGPT plugin development continues to evolve, new strategies and approaches may emerge. Therefore, it's always crucial to keep learning and stay updated with the latest developments in the field.

This study's methodology and its commitment to integrity are aimed at ensuring that the insights we provide are credible and beneficial for both budding and experienced ChatGPT plugin developers alike. We hope this will serve as a valuable resource to help you navigate your journey in this exciting field of technology.

Conclusion

We hope you found this post informative and helpful. By interviewing over 40 different successful plugin developers from many different countries, we did our best to offer as many unique perspectives as possible on the process of creating a plugin for ChatGPT. Whether you’re just starting out with your development journey, or are a seasoned developer, we wish you the best of luck and good fortune. Stay tuned to our blog as our interviews with plugin developers will continue. We have many more developers who have shared their insights, and we can’t wait for you to hear from them. Happy developing!