I just wanted to say hi as I have been using FS for about a week now to try to put together an AI powered portfolio manager which includes a bunch of AI agents to run portfolio simulations and backtesting for assets. A few screenshots are attached.
Overall I’ve been pretty impressed with Firebase Studio, though at one point it went round in a circular loop trying to fix a JS bug and basically got stuck. So I rolled back to an earlier commit and I also needed to get it to revert back to Next.js 15.3.3 because for some reason it downgraded it while trying to fix the bug. Anyway since then it’s progressing pretty well
I would recommend publishing your MVP as soon as possible, there are a lot of type checking and build errors that can come up that can go unnoticed during development, causing a delay when you are ready to publish.
true story. the initial setup to publish is quite involved if you’ve never done it before. i built an app in 1 day. publishing took 3 because of my inexperience of publishing on google cloud. also some features broke during publishing which seemed fine in preview. cause the publishing to fail. troubleshooting took a long time. next time will be quicker i am sure.
@bigUXenergy - sorry to hear about this! If you have any specific feedback around how we can make the publishing experience better, we’d love to hear it and address them!
A simple green light / red light while in the prototyper which can run basic compile time checks like tsc and show when pre-commit hooks fail would be a good first step. For me, I notice when I don’t see a git hash in the prototyper file changes I know I have errors to fix in code view because of the pre-commit hooks failing. A button in the prototyper that could run the build command and say if it passed or failed and then feedback loop on the output would be convenient.
Thanks for the feedback. However I’ve had to put a hold on this, because it was hallucinating the figures when the AI was running backtesting and also the output was not deterministic. After 20-30 attempts it was just going round and round in a loop not getting everything right after I kept asking it to check the figures by reviewing the trading log that it was generating.
When I come back to this I’ll need to be more specific about what I want to do, though this time I did supply a PRD as best I could.
It’s very important to use Git so you can also store a stable version of your software. I built a portfolio using Firebase Studio and also utilized GitHub Actions to ensure the code is clean. I then deployed it to Netlify very quickly. I can highly recommend this approach!
Your project looks really great. Actually I am working on an AI driven Portfolio Tracker as well. So far I have been focusing on the data structures in the DB, fetching data from a data provider and enabling the logging of buy/sell transactions and calculating metrics such as average buy price etc. I have implemented a conversational AI and tuned it to understand and respond related to the data in the database.
I guess I started with the basics whereas it looks like you have gone for very sophisticated look and feel and AI functionality.
I would be interested if you have done anything related to data structure and data feeds or if you are just testing the AI with test data hardcoded in for the moment?