“The AI response exceeded the maximum length. Please try simplifying your request or breaking it into smaller parts.”
Now what? I’ve asked the bot three times ‘how can we do this better’?
I’m asking the bot to display re-scaled values in a record after the record’s relative unit (i.e. number of grams in a serving) is edited. That’s all. I’m not even asking it to update on the fly - the user is told they must close and re-open the item to see changes.
In related news, I’m also shopping for a GPU to run a local LLM for coding. It’s getting silly how hard (and expensive) it is to make the bot do the simplest explicit changes.
Hi @sdmahaney_sla
We can understand that it is frustrating.
I recommend using discuss mode and asking the AI to provide more information about the specific area in your app before sending the next prompt.
I did exactly that, four times.
The last several prompts:
“In the edit food modal nutrient values are not being shown according to the updated alternate unit after save. Fix this.”
“The last prompt failed. Do you know why?”
“The last prompt failed. Do you know why? Can you adjust for successful completion?”
“The last prompt failed. How can we so this differently?”
“We keep getting “AI response exceeded the maximum length”. Can you structure a response to be more compact?”
The response to all was:*
The AI response exceeded the maximum length. Please try simplifying your request or breaking it
into smaller parts.*
So I’ve moved on to editing other parts of the app. But this problem still needs to be resolved. Am I sinking time into a dead end?
You can’t rely on Thunkable’s AI as of right now. You’re better off asking for ideas from Gemini or Claude. Last time I used Thunkable was around a year and a half ago and the AI went horribly wrong. Use the AI for help when you need it, but avoid asking it to create code for Thunkable blocks. Sorry!
I gave up on AI for Thunkable block generation months ago.
I tried the direct coding AI model and after a little positive experience pre-paid for a year.
It’s a complete disaster. The bot doesn’t know the limits of its implementation and the limits are such that anything more than a trivial demo app can’t be generated. Even if those limits were raised, the monthly credits would get burned up in the first week every month.
All I want to do is make a logging app. Most data is user-generated, the volume is trivial. There are four record types, with fixed master and user-generated versions of each. The workflow lets the user copy an item from a master to a user table, edit that record, and save it. Another record might be a list of items from the first table, with quantities and date and time. Then the user should be able to edit that record, including date and time.
This AI has been completely incapable of enabling that workflow.
[This was absurdly complicated in the manual Thunkable block version, but I got it working. The visual UI side in Thunkable blocks was just hopeless to deal with.]
I have a bunch of GPUs on an ebay watchlist now. From what I’m reading local LLMs, and even the best commercial models, still suck for 100% AI generated code, but with a local model I can keep poking at it at my leisure.