Flow and architecture of plugins

I’m learning about plugins and their flow before I build one related to the travel industry.

Questions I have are:

Can I use plugins within a chat bot, like whats app?
Can I use chatgpt embedded in my website and have access to the plugin?
Flow of events within the plugin. How does it know what to invoke and when(flow)?
Can I async talk back to chatgpt, after approval (manual process)?

Are there any documents that talk about the flow and architecture?

Thanks

I don’t know about using chatgpt plugins directly in anything outside of ChatGPT.

I have a theory that OpenAI will try to make an “everything app” where plugins allow you to do everything straight from the ChatGPT app. Normally what happens though with these types of apps is that they try to limit activity outside of the app itself. Every social media website down ranks content that moves the user off the platform, and I suspect ChatGPT efforts will revolve around the same strategy.

You can however use toolkits like langchain and semantic-kernel to build out your own pipelines. And those can definitely be used in other apps if they have an accessible API.

2 Likes

Thanks for the reply.

I’m looking for a flow of events when a plugin is initialized into the chatgpt UI.

I’m looking for a diagram of the flow and which direction the data flows.

I will continue researching to see what I find.

Thanks