I’d like to use the API to retrieve a ChatGPT session using the /c/id URI. But I don’t see this in the documentation.
Would this be a thread retrieval?
Is the ChatGPT URI the same as a run_id or thread_id used in queries?
Is a ChatGPT session in a chat.completion object that can be read with proper authorizations?
Applications include:
Basic search utility for a user to search through their own messages and find text, perhaps filtering on messages from user or assistant, date, or within a specific context.
Ability to export a session to markdown or another format.
Ability for students to submit entire sessions to a teacher in a digital format, other than a link, perhaps for digital processing.
Ability for a user to get a summary or comparison of many conversations.
Ability for a user to categorize their own collection of discussions.
I know a user can manually export their entire history, and all data is in the resulting zip file. This is a manual process, difficult, and not efficient. I’m looking for direct - authorized! - access to a session via the API.
Thanks for your reply, Paul! I tend to opine similarly.
ChatGPT is a separate application. Unfortunately the application (all implementations) are frustratingly limited for the masses. It follows the same “single textbox and submit button” pattern that simple people have enjoyed with Twitter, Facebook, and all other social media. And it hasn’t been developed as a full application with it’s own API, so the limitations there are the limitations we need to accept until they generously allocate resources to make changes to that application. I’m frequently amazed that one of the most technically advanced companies in the world with such high valuation continues to publish these brain-dead apps for the masses.
I tend to prefer a “lead, follow, or get out of the way” approach. If the ChatGPT app isn’t going to support various features, allow others to do so.
Of course, I know we can create our own completely separate chat application that does whatever we want, avoiding the basic implementation from this fine provider. But that fragments the ecosystem. I think there would be value to OpenAI publishing a ChatGPT API (kinda like the Custom GPTs) that will help to endear even more people to the core app(s). Extensibility is the model of the millenium. This “we don’t do stuff and we won’t allow others to do it either” is very 1990’s.
I would appreciate other suggestions but I think we have the real answer already.
Thanks @jochenschultz but while that option certainly brings with it new benefits, it also brings new challenges/differences/issues.
For example, I’m already paying for ChatGPT Plus features, and not interested in changing this toolset. I’m fine with using the API and paying for access to do special things. That is, I’ll ask OpenAI for some improvements to ChatGPT, but for others I’ll offer to meet them half-way and code into the platform - rather than leaving it. For my current inquiry, I’m hoping (but not really expecting) that they do have interest in meeting us half-way on stuff like this.
Also, switching to another tool doesn’t help the mass of people who are using ChatGPT and who might benefit from tools that integrate with this platform.
So I’d like to start with the tier-1 provider here first and see what they do.
I didn’t say you got to use it for the chats… I meant you use chatgpt and then export it and do what you want with the chat session. It was only meant as a temporarily work around.
However you may also build a custom gpt and let it store your prompt and also show the answers and then prompt it to save the answer again to an endpoint where you can store the chat session this way…
ChatGPT is a consumer product, and its disclosed features are what are offered.
It uses decidedly different mechanisms than API and is not programmatic for a user.
It communicates with your web browser behind the scenes to relay information in ways compatible with the client scripting and authentication within, and with a completely different dependency tree of messages and sessions than anything documented, but doing anything other than saying “hi robot” (or wondering why your full name, address, phone number, and billing is being transmitted over the wire just upon a page load) runs afoul of the way it is intended to be used.
Attempt to or assist anyone to reverse engineer, decompile or discover the source code or underlying components of our Services, including our models, algorithms, or systems (except to the extent this restriction is prohibited by applicable law).
Automatically or programmatically extract data or Output (defined below)…
A “copy chat” button is a feature request. And it would be a larger feature than you might expect when the page content is loaded dynamically.
Well, respectfully, we don’t actually know the full roadmap for the ChatGPT UI because it hasn’t been published. That’s why I ask. And I express interest so that they know that there is such interest in integrations with that product.
As to reverse engineering, there hasn’t been a hint of any such desire or intent in this thread, so mentioning that is somewhat OT.
As to a “copy chat” button, I posted a separate request for that UI feature in the Discord #suggestions channel. That’s where such things belong and all we can do is to follow the rules as they’ve been defined.
I’m content to simply make the case for support of a ChatGPT API and to see where it goes … or not.
