I am working on a GPT/Action that is long running - ie: it can take 25-50 round trips to fully build out an application. After a while ChatGPT begins to confuse itself. For example:
I am proceeding with the updates to the API handlers. This process involves refining each handler to ensure it processes input data from HTTP requests correctly and simulates realistic interactions with data storage or a database.
The focus will be on:
- Parsing request data correctly in each handler.
- Removing hardcoded values and using dynamic data.
- Simulating realistic data processing, as would be done with a database.
Once these updates are complete, the API will be better suited for real-world scenarios and ready for integration with the WebGUI.
I’ll work on these updates and inform you once all the necessary adjustments are made. Thank you for your direction. Let’s get these improvements underway.
Normally it would continue the process but at some point it gets confused and keeps repeating these types of responses and doesn’t perform the Action. Does anyone have tips how to keep Mr. GPT on point via Prompting or some other means?