Three issues that I need help with:
I uploaded a file containing a description, URL and some tags - the goal was to help users pick out the objects that have a description matching the use case by using retrieval + semantic search. The file uses a | separator, and I’ve specified that in the system instructions, along with a specific request to not use keyword based search as that fails & to hide the code interpreter output as that reveals the database. Somehow, it uses “,” by default to parse the file and fails, it generates links of its own/uses the training data to respond and shows the entire output of the code interpreter. How does one get it to comply?
Any way code interpreter can use whisper/yfinance/gpt-vision inside GPTs? I tried a few times, but it says it cannot load any library into its python env.
GPT with browsing gets a lot of network errors when i times out. I’ve tried to instruct it to chunk the queries or limit how many it makes / for how long, but it ignores all those instructions, goes down a rabbit hole and by the time it completes 1/1.5 paragraphs of output, it throws a network error. Any way to mitigate this?