Chatbot powered by GPT-3.5-turbo api

[Ts01]

Hi all,

I am not sure if i am writing in the right forum, so if an admin thinks (s)he should move my topic, please do.

I want to create and distribute via a Teams channel to a lot of users a chatbot that is powered by GPT-3.5-turbo api. I have created a prototype powerautomate flow that does this via an HTTP request and a power app that uses a custom connector. 

In both cases i have hit a dead end. 

1) With power automate i cannot find a way to distribute the flow to the users. I need each user to interact with the bot independently. Also even if i could solve this, my approach does not maintain context so every question the user might ask is smth new for GPT.

2) With power apps apart from some technicalities which I think i will manage to address, the main issue is that using a custom connector requires a per month license for each end user, which adds up to a large amount of money.

I am also considering going the powerVA way.

Anyone have any insight, recommendations, ideas, opinions to share regarding this? How would you tackle this? What would your approach be?

Every contribution will be deeply appreciated.

Thank you very much in advance.

[4-more]

Any news on this?

[ferzool]

Firstly, if you want to create a chatbot that maintains context between user interactions, you might consider using a dialogue management framework or building one yourself. A dialogue management framework would enable your chatbot to keep track of the conversation history, user preferences, and other contextual information. This will help your chatbot to provide more personalized and relevant responses to users.

Regarding your specific concerns with Power Automate and Power Apps, there are a few things to consider.

With Power Automate, you could create a custom connector that interacts with the GPT-3.5-turbo API and use it to build a chatbot. However, as you noted, it might be challenging to distribute the flow to individual users and maintain context between user interactions. One option could be to deploy your chatbot to a web server and provide users with a link to access it. This way, you could maintain context by storing user data in a database and use it to personalize subsequent interactions.

Regarding Power Apps, if you are using a custom connector that requires a per-month license for each end user, it could be costly, as you mentioned. In this case, you might consider using a different platform that offers more affordable licensing options or building a custom chatbot using a programming language and hosting it on a cloud platform such as Azure or AWS.

Finally, regarding Power Virtual Agents (PowerVA), it could be a good option as it is a no-code platform that enables you to build chatbots without writing code. It also provides built-in dialogue management features that enable your chatbot to maintain context between user interactions. However, note that it might not be as flexible as using a custom connector to access the GPT-3.5-turbo API. You could evaluate if the features provided by PowerVA meet your requirements for your chatbot.

Overall, the choice of platform depends on your specific use case and requirements. I hope this helps!

[Ts01]

Hi, 

Thanks for getting back. I actually chose the Power App way. So far using a custom connector but i have seen that MS actually offers (currently it is in preview stage) it’s own OpenAI connector, therefore the license cost for the app users will be eliminated in the future.