Hey man, I downloaded my code and just deleted my account, honestly your system keeps trying to charge me $1600 now and that’s just not going to fly. I appreciate you trying to help but we realized this software is just not ready for our company to build anything useful with. The sad thing is that I got really far and we were handling diamond search and querying very well, hundreds of thousands of them and last month we did not have these issues, we had some overages but only about $200 which was cool. This has 100% something to do with Agent 2. When you guys have sorted it out maybe we can give it another try but I can’t keep pouring time and money into a black hole right now, the time is the worst of it to be fair. I opened another account in case I want to give it another wack (that’s another $40, which is fine) but I won’t hold my breath.
Yikes. Godspeed my friend. Sorry to see you go.
I’d like to see the video where Matt Palmer said it’s for building disposable toy software, btw.
To clarify, I said its my opinion it is only for making toy software, I linked to where he wants to be able to create disposable software, which I agree with, its fine for small tasks, stuff that will not have repeated mission critical use.
Hey Hypler - Kody is an amazing person, and I trust him completely. I can relate to the frustration you’re feeling and truly hope this gets resolved for you.
I can concur from my experience that Replit is good for prototyping but Agent/Assistant creates more problems once what you are building goes beyond simple stuff. I dont have the confidence to build anything MVP worthy on it yet - cost, quality, maintainability to MVP and beyond being main factors. Easier to switch once comfortable with prototyping concept or user validation.
I think it all comes down to workflow. I think the majority of users just hope for the best, and the best is never what you’re going to get. It takes patience, and it takes lots of backups/snapshots to survive the frequent breakdowns by Agent.
But, I have produced some solid product, multiple things in production with live traffic. To say that it can’t be done is simply not true. But its an enormous learning curve, and Agent still needs a LOT of work to stay on task and not break existing functionality. Seems like a better system prompt would solve most of this, but idk.
They are still trying to charge my card so I canceled it. Their billing system isn’t even working because they canceled the invoice and now it came back higher. Good luck guys I hope this doesn’t happen to you, my warning to you guys, if you have any real amount of data, it will wreck you.
Hey @hypler , sorry about the slow updates here like I said I’ve been travelling this week and didn’t follow up with support after I’d sent over everything from you. Support cancelled the initial invoices but it retried again because your Repl that was incurring the charges still had that history which hadn’t been flushed out, we added the new feature to limit it but it didn’t apply to the existing one. Now that you’ve deleted the repl and the DB it should not happen anymore, I asked support to cancel the new attempts and eng to confirm there’s nothing incurring any more charges.
Was there ever a solution to understanding how to prevent this from happening? I am not sure I understood this.
Yes, we shipped that you can configure how long the historical database logs are retained for. Default was 7 days, you can drop it down now to a couple hours. If you’re doing something that continuously modifies the database and the history isn’t as relevant, drop it down extremely low so your overall storage doesn’t blow out.
I saw the email and came back. I did appreciate the response, I got pretty discouraged and didn’t know how to stop the charges and didn’t want to cancel my card since that is a whole mess on its own. I deleted the account and unfortunately that didn’t stop the charge so I had to cancel the card anyhow. I wish I didn’t do that because even though I downloaded the code it will be sort of a mess to get back where I was. I asked Kody if there is a way to restore the account,I am willing to keep working with Replit if there is a better way to control this issue which it looks like there is, now that I am aware of it.
This is the last version of the diamond app, I also had a analytics platform I made to collect data and parse it for us and I made a Wordpress plugin to go with, I didn’t record any video of that working but there are some screenshots somewhere on this forum. Thanks for everyone who weighed in, I’m thinking about how to move forward at this point and will keep you posted. Thanks for Kody following up.
I’ve said it before on other forums. DO NOT use Replit DB - create your own in GCP or AWS. On those platforms you get the data retention built into the price. One issue with this is replit will always try and change the enviroment variable for the DB - to fix this make a new env varialbe for example AWS_DB_URL - and then update the application just to use that.
Where can I set a max budget to stop all charges and use in the platform? I have the usage one set, but if the above example ignores that maximum isn’t that a concern?
Truth is, it should not be this big of an issue. If AWS works so easily, Replit should keep pace - or use AWS as the backbone.
EDIT: Bad scoop on my first post of this, the usage limit WAS set on the teams account. When we looked at it to confirm it looked like there wasn’t a usage limit set but that’s because the Org had been deleted, not because it wasn’t set. It was set per the screenshot hypler posted. The cause was the below because neon does periodic updates to the usage not continuous like AI:
The usage had 2x giant spikes each over a different 1 day period: 1 from 1GB to 128 GB, which sent a first invoice but allowed a grace period for payment, then during that grace period it spiked from 128GB to over 500 GB, which triggered the lockout and the big invoice. The user was locked out but it continued growing, so when we unlocked the retries were for the final growth amount.
We’re an end to end google cloud shop, not AWS, and on AWS if you did historical retention for your DB you’d have exactly the same issue. They by default don’t do that, vs we do so that the Agent can do rollbacks.
If you know what you’re doing we support using a more tailored option or tool than the default neon integration, as multiple people on this thread have reported doing. The vast majority of Replit’s users aren’t software developers and aren’t able to safely version code with the database manually which is what would be required to get a similar experience to rollbacks.
Wait, the screenshot above is of my Teams account and it clearly shows the usage limit, am I missing something on that? I get that Neon was creating snapshots of the database when I was trying the diamond import which was 800mb, and when we deleted it was deleting that much. From the look of it the fix is to just remove any database retention or lower it to 1 hour or something, BUT to clarify isn’t the screenshot I posted above from my Teams account? It shows $400 limit.
Sorry Hypler, that was my bad, edited the post above with explanation why it looked like it wasn’t set when we were checking, it was just the periodic neon updates as explained. The usage was set on the Teams.
Understood. It happens. Thanks for helping me out with all this.