I have a question about the store parameter in the API usage.
I noticed that when I set store=True, the conversation logs appear in the OpenAI Playground under my project history. I’m currently building a GPT-powered chatbot for an academic study involving hundreds of participants who will chat with the model simultaneously.
I already store all chat logs in my SQL database, but I was wondering whether enabling store=True could be useful as an additional backup — in case something goes wrong with my database. Therefore, I’d like to know:
(1) Is it appropriate or advisable to use store=True for an app with many concurrent participants?
(2) Does the Playground (or OpenAI’s system) have any storage or scalability limits for stored conversations?
(3) Are there any best practices for using store in production or research settings?
The “store” option stores a response ID object from making one API call.
There is no API endpoint for listing response IDs
The response ID corresponds to one particular API run, not a conversation
The documentation only commits OpenAI to storing for 30 days
The data is fragmented between an input list and an output generation, requiring two different API calls, returning different formats.
Contrary to documentation, there no automatic deleting currently, and no delete button.
Thus, the data is just as nearly lost unless you have a record of every recorded response ID anyway…in your database.
The best practice would be not to commit to ANY storage of user data on a platform that cannot keep their promises of limited persistence and offers no method to purge said data. Ensure that the store parameter is deliberately set to false in all API calls unless you have a particular use for which server-side persistence is required, that you understand.
Excellent - thanks very much for the clear explanation! I shall set store=False in my backend API calls going forward. Really appreciate your guidance.