Sep 19, 2023
Motive
The main objective of OpenCopilot is to replace the current complex web UIs with a superior interface - one that is intuitive for humans. This improved interface is none other than a human language interface. Instead of navigating through a many buttons to accomplish tasks, users should have the ability to simply articulate or describe their intentions using natural language. This is precisely what we are building: a text input system that can be seamlessly integrated into any product. With this interface, users can write their desires to the product, and it will respond accordingly.
Just last month, we launched OpenCopilot V0.0.1-beta, an open-source tool that allow you to craft an AI copilot tailored to your specific product. This copilot has the capability to interact with your underlying APIs, offering your customers an interface that surpasses traditional buttons and mouse clicks. All that's required from you is to provide it with a Swagger file detailing your APIs, and it will handle communication with your backend APIs as needed.
As a result, your users will have the freedom to engage with your product through conversation rather than relying on button clicks.
The pets store case
We used to have a basic dashboard for managing pets, allowing users to create and delete pets, and view sales stats. It was a typical CRUD product. However, when we integrated OpenCopilot, it underwent a significant transformation.
Now, users can ask more in-depth questions about the data. Instead of showing stats in a table, users can inquire about specific details, and OpenCopilot will fetch the information from the backend and provide it.
All what is needed is to plug 5 lines of JS code on the pets store, in this beautiful chat bubble will show up

It's a simple chat box, where your users can ask questions, for example, you can as it to create a pet for you, and it will do, based on your description.

But this is pretty basic stuff, you can ask it a more challenging question, something like "Analyze the decline in sales in July", which is something the standard app can't do by itself, and now it's just native

Ok, maybe a bit more complex question?

This really awesome thing, a simple CRUD app is now much more powerful, insightful, and with 0 learning curve.
The problem
OpenCopilot encounters challenges when dealing with multiple API calls, making it difficult to handle complex user requests. Recognizing the need to make OpenCopilot accessible to a wide range of users, we sought to provide a solution that allows product owners to define their system flows to assist the copilot.
For instance, if your product includes a complex process like an "add to cart" functionality, it can be challenging for the copilot to determine which APIs to call without some guidance.
Introducing OpenCopilot flows
This is precisely why we introduced this feature. It empowers users to define intricate system flows, such as "add to cart" or any other custom processes they require. The process is straightforward and user-friendly. Once you've uploaded your Swagger file, you can effortlessly specify these flows, which will then be seamlessly integrated into the copilot's functionality.
You will start with an empty flows area, just like this one

And once you provide your Swagger file, you will be able to see all of your endpoints

Then you will be able to create your own flow:

and then you can put your API flows:

And that is it, now your users can ask your copilot questions like "Please order me pets with tag pending".
What is next?
You can get staerted with flows using:
We are very excited about the world of products copilots, and our goal is to build a tool that allow any SaaS product to have such copilot that can talk with their backend system with out the need to hire 10 AI engineers to do so. it just the right way to communicate with products, the native way.
We are open source, and we are very committed to this product help us by:




contribute on GitHub
© 2023 OpenCopilot