ChatGPT in XoJo and accounts

General
1

I haven’t tried the example yet. My question is, if we must setup an account and provide payment info? What happens when we distribute the app? Can we have the code ask the user to put their info in? If so, what about users that don’t want to provide that info?

2

I have been using the example and works very well.

Yes you would want your app to allow users to enter their own key and those that do not want to pay, well no AI for them :grin:

You could mention that OpenAI still give a credit for new account but please double check, I know I did when I signed up.

3

That’s what I figured. Thanks for you response. :slight_smile:

4

Typically the way you’d approach this is by selling a subscription to your app. Your app then communicates to your own webservice, which acts as a proxy to openai’s services (this way you don’t expose your own api key to the user).

Your user’s subscription cost should cover the gpt credits you need to pay for, and you can limit them or upsell them to higher subscription tiers based on the number of api calls they instigate.

Or, of course, you could allow them to add their own credentials.

2 Likes
5

Yes indeed that is a great way
Just depends what the OP has in mind for his app :grin:

6

In enterprise applications, the most interesting thing will be when we can associate our data with gpt. For example, in a crm app, if the ai can access a database, understand/analyze it, and the customer can ask questions like “how are my prospects evolving at the moment?” (I’m deliberately asking a non-specific question, since it’s the point here of an ai to provide us with an opinion that we couldn’t really provide via a simple algo).

Apparently we can already send documents, so maybe structured data, but I haven’t looked yet.

If anyone is exploring the topic, please share your findings!

7

If allowed by the service agreement, of course.

1 Like
8

Hi, yes it is possible and works excellent for us. Basically you have two options to achieve what you want:

  1. You create a preconfigured “agent” where in the SYSTEM prompt you tell him what he needs to do (“analyze the provided data and respond to the user how are his prospects evolving at the moment”), then in the ASSISTANT or even USER prompt you include the data you want to be analyzed (json, csv, xml). For this type of “agent” you don’t need to allow the user to ask him questions, because it shouldn’t be a conversational agent, it just needs the data to be analyzed and report back the findings to the user.

  2. a conversational “agent” where you need to learn on how to use the tools function of openai. basically the user will be able to chat normally with the agent and maybe at some point in the conversation the user will ask the “agent” hey, remind me how are my prospects evolving at the moment . the “agent” will see what tools has access to and use the tools to retrieve the needed data from a database/webpage/document/file etc. this is a bit more advanced where you also need to learn the notion of Context-Tuning in AI.

so yeah, it is very much possible and I am already using implementation of both methods described above in our CRM, that uses our clients data.

1 Like