Why am I doing this?
Now, more than ever, anyone who says, "I can build something better," had better put their money where their mouth is. But, that's not my motivation. True, the first AI-powered workout app I used had a limitation that frustrated me, but the real reason for this endeavor is more about trying out tech than filling a void in the app stores.
CloudFlare hosting
The top of my list to explore is hosting on CloudFlare. I have played around with CF workers a bit in the past, but I have yet to build a full app. We use Next.js at work, so I was particularly interested to see if I could get that running well in this environment. I have implemented OpenNext to get the job done. Looking forward to a CloudFlare implementation of the new Next.js adapter API, but OpenNext will suffice in the mean time.
NeonAuth BetterAuth
I have played with Neon databases, as well, and I think the idea of Neon Auth is fantastic. And, it probably is. With Neon databases, you get db branching, which is amazing for development. Add that to not having to roll my own auth also seems like a big win. But, on CF Workers, it's a world of workarounds. Neon Auth wants to use proxy.ts. Proxy.ts does not play nicely with OpenNext and CF Workers since it requires a Node runtime.
Between that and the fact that Neon Auth is a paid wrapper of BetterAuth, I thought I would just go to the source. I still get db branching since I'm using Neon.
Neon
Said what I need to say about that. DB branching is awesome.
Tessl
I have been getting whiplash trying to keep track of what I am and am not supposed to do in terms of putting up guardrails for AI. Tessl promises to fix that, so I'm giving it a try. So far, I'm happy with the results. I built a couple of Tessl tiles and used the feedback from Tessl's optimization algo to dial them in to reduce context expense.
What It Does
This is pretty simple. The goal of the app is to start with an interview of the user to see understand goals, like/dislikes, no-go exercises, available equipment, etc. From there, it makes a daily plan for the user. The user gives feedback on how a given day has gone. Future workouts take the feedback into consideration for future planning.