Tonight I’ve burned 20k a.i. tokens trying to get the bot to fix its own mess [Would you believe that when a user edits a record it’s expected that said record be saved??]. I let it keep trying its own fixes without reading much, but realized I was stuck in a loop. Now I’m trying to coax the bot into deleting just the messed up section and starting fresh.
How does one manage such a situation without burning through a stack of tokens trying to address a problem?
Based on your other posts, it sounds like you are using an external database. One thing that can help conserve tokens is to use discuss mode to talk with the AI and figure out what information it might be missing. For example, it could be that the code doesn’t match the column names in your database. This actually just happened to me where the AI was missing the context of the column headers and wasn’t inputting the data into the table correctly.
I could probably give you better tips but I would need more context on what you are trying to use and build.
Then entire reason I landed on Thinkable is that I do not want to use an external database! I want all data to be stored locally. That said, there is a lot of data that needs to be read in at some point.
The bot said I could simply paste all 5200+ rows of 70+ columns right into the dialog box. But the bot, despite all its general knowledge, was not given the parameters of the box it’s stuck in. So Thunkable stopped that with a limit error (without informing me what the limit is either).
edit: The bot also suggested I could put the data table into the project “assets”, but we don’t have that sort of functionality that I can see.
The bot explained how I could use Google sheets to bring in the data, so I think that’s been done. But I suspect it burned a lot of tokens for that one operation. I don’t have a good way to know, except to watch the one coarse total in the account info. Is there a more granular way to watch usage that I haven’t found?