Firefiles Demo: My Process and Decisions

Hey, I'm gonna share here what was my build process?

we're super quick and my decision so basically I started talking to cloud chat about the project and brainstorming it.

I, I spent like a couple of one and a half, one hour and a half two hours preparing doing code research, task preparation and preparation preparing all the field And after it I finish with a project here.

So I have a quick summary that I could like Send to the AI etc.

To have context and a couple of issues.

So all these issues I went one by one until finish it.

I just like finishing this last one that makes sure the rhythm is good for you and All these tasks for example have a great description with my skill that I use to generate tasks with a why a context, what to build acceptance criteria out of scope and how to verify so AI can work really fast and well with a description like this.

My key decisions was for foundations of the project like I use my own stack foundations and stack right I use my own full stack code base so I have like a open source project called Create my next idea that already has a lot of things included for example Chatcn UI my own design system and app shell Drizzen configure it with postgres and other things that are out of scope here like authentication etc.

For database I choose Neon postgres with drizzle because I think it's fast to iterate it's easier to use with typescript and I can connect NEON oversell in one click for the transcription storage I just create a column meetings table and I have more explanation here but it's basically because the meeting output data is not super big in most of the cases for LLMs I'm using Cloud Haiku Viantropic SDK and whisper from OpenAI but via replicate the background AI processing I'm just using wait until from Vercel I didn't I put like out of scope create some type of job queue or some trigger.dev on this case but because like it's a simple procedure for AI output quality validation I'm also using the cloud haiku to validate if the transcription summary is really a meeting or not or if something that is not a real meeting and here I have all my code patterns it's well explained and my decisions for automated test basically the front end I'm testing unity and component integration for the back end.

just like integration against real DB and mocking the third party APIs when necessary.

All the libraries are used but basically next JS TypeScript Postgres, with Drizzle, ChatCN with Till it's at CS React Query, React Hook Form, ZOD and VTest with RTL for tests out of scope plus authentication and multi tenant job queues and retry orchestration.

So all of this was my decision and I really spent a time like researching product and I could easily create the whole process, the whole code here using cloud code.

So yeah, that's a really nice product.

Thanks for this opportunity.

I really enjoy to create this.

And let me know if you have more questions.

Let me know if you need to have more clarification.

More than happy to share.

So see you soon.

Bye.