Build an AI Research Assistant with OpenAI, Bubble, and LLM Toolkit - Live Build p2 - Threads/Msgs

[@NickaGillis]

regex for the annotations if you just want to hide them. We were able to get it to print the files in a list that it used to source. 【.*:.*†.*】

[@LaunchableAI]

yeah regex is a good idea. In the next video I will probably use "schedule API workflow on a list" with a multi-select to get bubble-native references to the PDFs.

[@nickgillis1482]

​​@@LaunchableAIGiving the assistant instructions with an example of the reference format it will oblige and give a list, combine with the regex, much cleaner. However, that's just for documents uploaded to the vector, and I'm assuming the thread too. I'm curious, we appended a document code to each file #t_code. Haven't tested it yet but we're creating a document reference list in bubble that doesn't actually hold the file but some information on it. Ideally that file is hosted elsewhere. But the doc code can be printed or omitted based on the instructions/examples provided. Spent 2 months testing instructions and formatting. Not sure this helps but floating some ideas.