Based on my experience, GPT4 plugins are a powerful tool for system functionality expansion. However, some issues need to be addressed. The main problem encountered involves changes in key parameters used for information addressing, causing malfunctions like the AskYourPdf plugin failure due to incorrect ID, resulting in errors.
REQUEST TO ASKYOURPDF
{
"doc_id": "1707.06347v2",
"query": "4 Adaptive KL Penalty Coefficient"
}
RESPONSE FROM ASKYOURPDF
Internal Server Error
Quick solve this problem it is suggested to chatgpt use the identifier which is used when the document loaded before, like this:
For AskYourPDF plugin please use this argument value "doc_id": "1d1b0f17-6261-40e7-9a9a-034c0fd3fba7"
To completely solve this problem, I suggest implementing stable user session identifiers within the chat, unique to each user and retained throughout the communication session. This will allow plugins to save their conversation context, avoiding conflicts and misaddressing.
The benefits of stable user session identifiers include improved plugin reliability and resilience, as well as easier debugging and error detection, as identifiers can provide additional information for analyzing issues.
Despite the identified problems, GPT4 plugins represent significant progress in system functionality. Implementing stable user session identifiers will enhance plugin performance and provide more stable interaction. I hope my comments will be considered, and the development of GPT4 plugins will continue to improve.