tunde825Q: Greta Limits?
Greta refuses to execute my prompt and yet consume credit. How can I force Greta to output the result of my prompt? It seems that it freezes after a while and doesn't react to further prompting.
I got different results from executing the same prompt. Basically, some aspects of the result I was expecting to see manifested in different instances (including a test account), but now I can't have the result replicated in the main account. How can I download / combine the results from other instances/accounts, which combined together resembles the result I want?
How can I import/merge code from one Greta instance to the other?
How can I edit the dummy or sample text in the application deployed?
How can I get specific buttons and links to work?
Suggestion: Add import, export, clone, and save features.
Deduct credits only after deploying.
Shubham_Questera
Jul 2, 2025A: Quick answers:
Freezing & Credit Issues: Usually happens if your project is large or if the prompt is too complex—try breaking it down into smaller requests. If it doesnt result anything it reverts the credits as well
Combining Results/Merging Code: Right now, you'll need to export manually, there is no merging code from different instances.
Editing: Just clearly prompt Greta what changes you need (text, buttons, links). In future we will have visual editor to do those.
For Suggestions: Not sure I understand properly. Import, Export is already there. What is clone and save? This is between chats?
AI prompting gets easier as you do it. You need to be clear and add as much detail as possible because it will "fill in the blanks" if you do not.
Answers to Questions:
> Currently there is no merge option but there is Github
> You ask Greta to make the changes, screenshots help show what you want and can highlight where to make changes.
> Tell Greta what you want those buttons to do
@JakeVan - I appreciate the suggestion, but there are some practical issues with the GitHub workaround:
The GitHub approach has limitations:
•
Agents can't actually execute git push/pull commands or directly access GitHub repositories
•
The agent would need you to manually copy/paste all files from GitHub back into the chat
•
This defeats the purpose since you'd still be manually transferring...
files
•
You'd lose all the conversation context and agent memory about your project
The real issue seems to be:
1.
**Agent state corruption** - Something causes the agent to become unresponsive mid-project
2.
**Context limitations** - Agents hit undisclosed token/context limits
3.
**Memory degradation** - Agent "forgets" project details over time
4.
**System constraints** - Hidden limitations...
that aren't communicated to users
Hmmm, maybe. I definitely meant to push the project to Github and start a new chat/project by repulling that repo back in. This "in theory" should reset the agent while keeping the most up to date repo.
The big limiter for this workaround is the lack of branch control.
I haven't hit the "dumb agent" issues yet but I can imagine that is extremely annoying.
Is the issue of the agent giving up in every project? or just the ones you work on the most? and around what number of requests do you start having issues?
I have used a couple thousand word prompts that it handled better than expected so maybe its only certain tasks. Could the giving up(assuming it was in multiple projects) be with a similar prompt or request?(Supabase, Stripe, form create, esc)
I have the same problem. I tried a few things.
1.specific prompt. Even used another AI to enhance the prompt. But no response from greta. However if I try the exact prompt in a new chat new project it works.
2. after greta gets stuck,if i select one of the suggested components, e.g feedback. It works, after I can run my command it refused before. But it stops again after awhile. Frustrating.
That's definitely not supposed to happen. Is it always with the same app?
Shoot support an email detailing the issue, what app its happening with(especially if its the same one), and any other information they might need to debug this.
In the meantime can you push to Github and pull that project into a fresh chat and get it to work? If this works, it will be a workaround in the meantime.
@lavoism - Your experience matches mine exactly:
•
Commands work in fresh chats but fail in ongoing projects
•
Selecting suggested components temporarily "unsticks" the agent
•
The cycle repeats, making long-form development nearly impossible
What we really need:
•
**Transparent limitations** - Clear documentation of context limits
•
**Better error messages** - Instead of silence, tell us what's
tell us what's wrong
•
**Fair billing** - Don't charge credits when the system fails
•
**Reliable persistence** - Projects should maintain state consistently
The GitHub workaround, while creative, doesn't solve the underlying system issues that make long-form development frustrating and expensive.
Hey @JakeVan, yeah I tried Github as well pulling code into a brand new chat and command still do not work. Only running a component unstucks sometine for awhile. I sent support email. They asked for time to resolve issue. Which i am ok with. But if it is a known limitation please disclose.
Yeah that sounds like you all just found a bug they weren't aware of.
But yeah I agree, if it is a known issue it should have been disclosed.
I talked to Shubham on a video call earlier and I do not think there are intentional limitations set directly by him or his team.
He has lofty goals and Greta working is required to pull it off to support the other massive projects on the horizon.
I have high hopes for Questera AI, so this is a good first test to see how they handle knocking out big bug fixes when it matters.
Just give them as much detail as you can and we will see the results, they did not expect Greta to be as big of a hit as it is.
Shubham might have to expand the team sooner than expected. lol
Yeah, no worries that is why they come to appsumo for us to user test and give feedback for improvement. The bugs are expected. By no means implying Shubham knew before launch. Just saying as they discover or many user report, share the plans to address. Which they have done. A roadmap would be nice. I believe in the team and in the vision. Great things take time....
Appreciate Pandazis.com and lavoism. You guys said it even much better. I laid out my frustrations and went to bed LOL! Greta is indeed a brilliant idea, and I just upgraded from Tier 3 to 4 believing all these bugs will be fixed. @JakeVan, thank you for the valuable inputs to the Greta discuss. @Shubham, export, and no import, is a missed opportunity. Pls look into the merge possibility. Thanks!