I've been struggling with this for some time without resolution and have 48 hours to make it happen or bale. I am looking for an example of a use case where a virtual agent conversations compiles a query and send it to gpt4.0, receiving a response that can be inserted into the VA conversation. It seems very straight forward. I can do it programmatically inside a custom app.
The current automate Create Text with GPT using Prompt that I keep getting pointed to is build with only 3.5, which is not valuable here. 4.0 is perfect as I have validated that outside directly with chatgpt using my 4.0 version.
In the Microsoft arena, I have my OpenAI set to 4.0. My playground works great! The problem is getting that result into the VA/Automate flow.
I see 2 approaches, neither seems to work properly. 1) I can call a direct HTTP request from VA. I have played around with the headers as best I can but it still fails with various issues depending on the variations I use. I can get it to validate with the correct end point and API key, but then it fails. This approach is fine if I can figure it out. An example would be awesome. 2) Use another automate action that connects to 4.0 (external chatgpt is best but happy to connect to my azure openai 4.0 as well).
I found that ai builder only points to 3.5 while Microsoft looks to security issues etc. Not sure if that is true but very unfortunate as 3.5 is not even close to a solution for our application.
Thx in advance!