- Now it waits till the client finishes work
- It tracks the program with System.out.println instructions explaining what is happening in crucial moments
- I put all urls in properties so that they are in one place and easily accessible to make sure all of them are legit and there is no shady stuff going on in this project
- I also put user prompt and assistant instructions in properties for easy customization
- I changed default model to gpt-3.5-turbo because it is 15x cheaper, and the user will most likely do some test runs at first, so its gonna save him a lot of money
- I made it simple to revert to gpt-4 model and put its name in GPTModel enum, also added a comment about the current pricing and limitations
- Also some minor adjustments like removing some unnecessary comment and adjusting variable names to make the code more understandable
- I added println that will show you all the responses so that you see what you are billed for
Not the nicest code I ever written but its decent and gets the job done. I have massive concerns with Assistant API from OpenAI, not sure if I want to contribute and use it. Basically, a single and simple instruction and message that should be answered with a few words might result in tens of responses and there is no way to control that. Will report these concerns as Issue.

Assistant API might get very expensive in case of any issue during processing the prompt, therefore I added appropriate section in README.md. These issues forced me to stop of using Assistant API, and I think users should be aware of risks associated with using Assistant API.