Goal is to simply connect Make to an OpenAI Assistant and ACCESS uploaded Assistant knowledge files"
Open AI Assistant does not access KNOWLEDGEBASE, if I don’t have any selection in the module or if I select to use the knowledge in the module, either way it does not work.
I CANNOT GET the ASSISTANT to READ the KNOWLEDGE when using the OpenAI Module.
I have RESOLVED THE ISSUE, but I am leaving this post up in case it can help anyone in the future.
MODULE: OpenAI (ChatGPT, Whisper, DALL-E)
THREAD ID: NONE
MODEL: NONE
FILE ID: Make sure to select the proper file. I uploaded/updated the “Knowledge” file multiple times while building the Assistant. BUT I only have ONE FILE in the knowledge for the assistant, but something weird happened in make where it listed the knowledge filename multiple times, so if you are experiencing issues make sure to try testing all filenames with the same filename if you have multiple in the FILE ID Dropdown.
TOOLS:NONE (Here I did NOT select my file)
INSTRUCTIONS: NONE (This will override your Assistant instructions, you want this blank)
Yes, you only want to map fields here if you want to override the default assistant settings.
For thread ID, if you want to reuse the same thread at a later date, you have to store & retrieve the thread ID. This can be useful if you have multiple “chats” by different users using your assistants. More info about it is here Telegram chatbot (incomplete) - #3 by samliew
I marked this resolved before but it is UNRESOLVED OpenAI Assistants DO NOT WORK CONSISTENTLY ON THIS PLATFORM.
It RANDOMLY accesses the knowledge files.
I tried a million things to figure this out including a timer to see if knowledge files take time to populate in OpenAI.
I created a random number generator to use in knowledge filename so when uploaded to the Assistant there would be no potential issues during different scenario runs. Ther is NO SOLUTION.
And I cannot COMPREHEND HOW I AM THE ONLY PERSON WHO REALIZES THIS ON THE ENTIRE PLATFORM.
UPDATE: Make Support checked out a video I sent, they went through everything in my series of Scenarios and Data store and they think it’s an OpenAI issue