I am using the ChatGPT API (model 4o) with the Code Interpreter feature in C#. I have provided a JSON file to the assistant as part of the system instructions and have configured the platform by logging into the OpenAI website. Before making a request, I create a thread for each user, and I provide the initial prompt for each thread. My temperature setting is configured at 0.1.
However, I am encountering an issue where two separate threads are being created for two users, and while both users are asking the same question, the responses are inconsistent. Specifically, one user receives the correct data as expected, while the other receives responses such as “No data found” or “I can check again if you want.”
What could be causing this inconsistency in the responses?