So, like others my prototyper history/context got wiped out. Fair game (early stage platform). But how do I go from here? Is there a way to resume working with prototyper on my project?
Or am I resigned to working only on the code/IDE side of things?
TL/DR: any help get an existing project that lost prototyper context/history back on track with Prototyper?
The long of it (for those who care):
I have now committed the code to Github and started a new project from that codebase, but the outcome is that project does not even show prototyper. Sucks to have to work with IDE only since the experience and ability seems very different (this is a much more coder focused approach). Whereas, call me lazy, I had figured how to get Prototyper to get deep into my app just with prompts and some finagling to get it to do exactly what I need. It was astonishing how far it got until… the new update wiped out the history/context of Prototyper and now it’s stuck thinking I am starting a project even though I have the codebase there, etc.
This significantly sucks because the speed at which I was making progress was much faster than that of the IDE since the IDE seems to be fairly unknowledgeable about the app context - especially since I had worked mostly on prototyper. I had gotten the app up to the level of creating users, roles, different functionalities committing to Firebase, etc. In short, as to the point I had stopped it was all working quite well. Now (in IDE) it is a sluggish process (comparatively). So ideally:
I can get the context/history back
I can get Prototyper from being stuck in “it’s a new app”
I can get prototyper to work on a new codebase (maybe by having it look through the code and perhaps ask questions to build it’s context?)
Not that coding (IDE) is bad, but the experience and progress is significantly better in Prototyper (at least as it was from starting from scratch and keep going). The REAL bummer would be for me to have to diss all the work done and start from scratch to re-build the Prototyper’s context. I spend close to two weeks on this and arguably second time around would be faster, but really sucks to think I can only “keep the thread” going with Prototyper if stuff like context is not wiped out - would mean every time this or something like it happens I would have to restart (from scratch, sort of speak).
Lastly, maybe this time I have to bite the bullet and live without it, but a good feature would be to (if possible) commit a prototyper log of sorts to the source control so it can “revive” in case it goes kaput like it did recently. Not sure about the intricacies of Gemini and how it flows through (with) Prototyper but seems like this persistence feature would be a must.
Got room for another in that boat @Jose_daVeiga ? well said and same issue. Glad to see I am not the only one with this issue, however, sad to see there is no solution just yet.
I agree with the revive concept you mentioned as the when the prototyper came back from the first death, it sold me the trust it wouldn’t be an issue. Then, when that one bit the bullet as well, the next clean slate just wanted to “prototype this app” into space. Also, what ever code I throw at it just keeps spamming that “prototype this app”. Is that what you are getting?
Not a big deal just carry on the ai will pick up what it has done. In my experiance it gets slow with a long context window. I happend to me a couple of times before but did not really stop or hinder my progress.
I have stopped trying anything on that project while I wait for the Google team to figure out what happened (in hopes I can resume).
But I took all the code to Github and then started a new project with the GitHub codebase. Unfortunately this seems to be such that prototyper is not available for such projects (from existing code).
As I suggested before, great use case would be to have a way/process/method for the user to guide the AI around the app (say start with the AI trying to undestand the codebase, ask questions, etc - maybe the user has to upload a sort of project description or a spec doc of sorts that explains what is the project and where the codebase is at, etc).
Way I see it, a Product Manager with tech chops at the architecture and even code level (but like me is not day in day out in the code anymore) could develop a full complex app this way, even start from an existing app and modify/extend/enhance/etc. This is actually how far I was with the project in Prototyper. It took me a bit of bumping my head into the “circular spiral of death” (when the AI just goes in circles) to understand how this works: basically one needs to have a concept of the app to be built, how it is built baby step by baby step at a time, and be very disciplined as to what is a “pause gate” in the development and only then move on to the next step. The rolling back history feature was awesome when this happened.
Unfortunately even in the same project it seems that Gemini in IDE mode has no clue what happened in Prototype mode (although I was not 100% clear if the reverse is true - seemed like when I sporadically changed some stuff in the codebase the Prototype Gemini didn’t get hindered or confused by it, nor did it seem to remove or change my edits or additions). Overall in the new project with the same codebase the progress is much slower and much more painstaking. Gemini wavers between telling me it can’t edit the code and editing it. It gives me super detailed explanations of what to do but can’t seem to do it (although Prototype mode did about 99% all by itself with no issues). Seems to me like Gemini in IDE mode assumes that the developer is the driver and is only asking for help with conceptual stuff and as such it’s much more hands off and much less encompassing as far as looking to understand the whole codebase/app/etc.
Not sure which is which for their target customer base. I think right now AI tools are divided generally speaking between those for the complete non-coder and for the super coder who will benefit and be super productive (but does not need high-level help). The fascinating thing for me about prototype was the fact I haven’t gotten down and dirty with code for a while but I still understand most of the Architecture/frameworks/etc and as such it was amazing how I was blowing through certain parts (most of the challenges were with firebase and user creation and the respective rules) of the app. I was literally doing a 2 day per scope advances which before would likely take me a week in my best days of coding. And with the advantage I was addressing little details (usually a nag for coders but a must for a PM and users) and the app was getting quite “final” at each step. Also really good was the sprint style of approach I was taking which usually requires meetings and pre-planning or pre-design work that I alone was able to perform on the fly.
All in all, bummer the history blew and I can’t get it back to continue. My app is not extreme but complex enough to note it’s not just a website and some forms. In fact, as things were going before it all blew up, I was estimating I would have an app fully baked within 2 weeks. At the pace I was going, if the AI didn’t get overwhelmed at some point (this has happened to me with almost all and any other tool I have used expect for VS studio code helper or such other tools - those for coders) it was looking like this might be the best fit for someone like me.
Can you explain how you “carry on”? I tried it and, like Jiyn mentioned, prototyper seems stuck in thinking I am starting a new app (even though in IDE the codebase is all there). What did you do to get it unstuck from it thinking it’s starting from scratch? (see attached image - I always get the annoying “App Blueprint” answer no matter what I ask or tell it in prototyper since the history was obliterated…
@Jose_daVeiga I was found that in the coding section, menu to the left, there is a “Source Control” icon (Located beneath the search icon) that housed my conversations with prototyper. Something called the “Source Control Graph”. Not sure exactly what it is, but I was able to view the chats and corrections made at some point. Not sure if that helps with anything but in my instance, gives me hope on the developer roll out that not all is lost with the app I am working on. I hope on the roll out within the next few days there is something more client side that houses our conversations with the prototyper so that future AI Engines have something to directly point to in these instances.
Thanks. Ya, each “step” in your chat seems to be committed into a Git (source control) repo. It’s a great idea. Also, I am guessing the context history should also suffer such a fait (if it does not then I’m sure this is a high item on the dev team’s backlog). It’s exciting. I would love to develop an app that my company can work with, but given it is not critical path (basically my app would replace a paper/email process and introduce a tracking on the manufacturing floor that is in place for 35 years) I am just loving this process. I have tried a lot of the tools out there and honestly I still think the hype is big but, like with any product the progress is slow… but lot’s of fun. Makes me wish I was more in the day to day weeds of this. Kind of my biggest pain is the wait to see what comes next.
It has happened to me many times on the same project, at least with me if I ask the prototyper if he remembers about “this and that”, he goes and replies with full details about the thing I was asking, so even if the chat is lost, it does have the full context, at least in my project… BUT I have the opposite problem I want the prototyper to forget the context, why ? Because he keeps undoing and re-doing previous changes to the code, which I’m pretty sure is a context management problem… it isn’t looking at the code base as it progresses with time, so he keeps reverting old stuff, then later on fixing it… maybe deleting all past commits will do the trick ?
** A few hours later **… yepz, got to do a reset again and he forgot everything this time and wants to prototype the app again, no matter I say he won’t look into the code base… any workround ? I think the main problem is this, the prototyper should be able to realize there’s a project already and go trough it to get the necessary context instead of trying to start over
** A few minutes later ** I GOT THE WORKAROUND FOLKS! here is what I did:
although the old prototyper memory was gone, the project was still fine on the backend <>
first I linked my github account to the source control and .
pushed the project to github (it then created a new repository named “studio”)
accessed the studio repo and downloaded an offline version, to be safe…
cloned the repo to my local machine (something good to have, I’ll be able to work locally in vscode)
Ok, until now I was just backing up the project, now comes the fun part:
With your project open in the prototyper, ask for anything, something very simple.
prototyper will plan, and you can ask him to prototype de app
he’ll proceed to prototype and delete all your original project files… DON’T FREAK OUT
when he is done, go to the backend <> and head to the Source Control panel once again
hit the … from the source control dropdown
if there are any changes to be committed first go to Changes → Discard All Changes … then confirm
after that go to Commit → Undo Last Commit … then confirm to undo commit
… menu again go to Changes → Discard All Changes … then confirm
if some files are still left, waiting to be Staged:
… menu again go to Changes → Discard All Changes … then confirm
if not you successfully reverted the project the new prototyper messed up.
Delete any temp/cache folder you have like .next
In my project I also had to delete node_modules and package-lock.json
reset the project VM
npm install
NOW FOR THE LAST PART, let’s talk to the new prototyper:
Back to the prototyper interface (the magic wand)
Prompt him: “take a look into the codebase, I’ve made substantial changes to the app we’re working on. let me know what you learned, don’t code anything”
he’ll then proceed and get to know the project, YES!!! He’ll do it!
then prompt again to be safe: “ok, for now one we’ll work with this app you just looked at. You’ll for now on disregard the original app idea and code you prototyped. sounds fair ? can you do that ? don’t code, just explain”
he’ll agree with it, and this will be probably enough to keep his context in check not trying to return to the prototype he first coded
if he does messes up, revert using the prototyper version control and be sure to prompt him to disregard the original prototype app and work exclusive with the one you ask him to read before sending new task, with time the new context should get stronger in memory.
That’s it! You got your very good and excellent prototyper friend back ready to work on your project again, and in my case, free from previous context, which does more good than bad… just make sure to keep a notepad updated with important information you talk with him, like planning, next features, desired future changes, etc.
NOTE: This will most certainly enable a local uploaded project to also use the prototyper, just follow the same steps.
@DaC Cool. I didn’t try it on the original project (have hopes that the fix they promised by EOW will fix it). But just for fun, and because I finally had time, I gave this a shot with a new project. And IT WORKED! Thanks!
Basically I created a new prototyper type project. Then just got it to create the code as per your example. But then I just set a new remote on the git and deleted the studio directory and got the whole codebase in there. Then deleted the temps and reset the VM and went into Prototyper and used your prompts. And Tadaaa! Works Still having some stuff missing (like the firebase link at the top of the firebase studio UI is not showing like it does in the original project) but pretty much back to working and iterating. Right on!
Still having some stuff missing (like the firebase link at the top of the firebase studio UI is not showing like it does in the original project) but pretty much back to working and iterating. Right on!