I have cut quite a bit from andurel, no more DI (dependency injection) mode, no more vanilla css vs tailwind options, no multiple database options
I thought I knew how to keep products/tools simply but this process has thought me to always:
Cut cut cut
introduce later if needed
went through the basecamp cli to get some inspo for making andurel's cli more agent friendly
made a ton of changes that will be available in nightly tomorrow
think release candidate next week is realistic
I want non-developers to confidently be able to say:
> spin up a live version of this PR on this URL with this data
or
> add servers to regions where usage is high
to their llm of choice and have deploy crate provide the APIs to do that in a secure way without breaking the bank..
New pitch/vision for DeployCrate:
> The future of infrastructure isn’t more dashboards, it’s delegation. DeployCrate is the infrastructure platform built for agents, giving developers the control of self-hosting with an interface designed for AI to operate safely and predictably.
was kind of frustrated with my testing of glm 5.2 and kimi 2.7
would just blow through tokens doing the same reads over and orver
then I switched to the pi coding agent instead and both models stopped going on and on in loops
the harness does seem to matter
putting deploy crate on 'standby' for a little bit so I can focus on finishing andurel
having built multiple projects simultaneously for a long time has not been a good strategy 😂
will get it to a release candidate and then pick up work for deploy crate again
@rei_hoshiko7 Nice to hear that react+inertia resonates. I was a bit skepitcal about adding inertia due to having written a ton of react before llms and seeing the mess it would often become
But having explored it with an llm i'm actually quite excited
andurel is _very_ close to having the final structure for v1 release
i'm finishing up some things, adding support for react in inertia and then re-building a project with it
if nothing comes up, v1 will finally be out
almost 1 year in the making
I'm back to building SPAs but this time with AI and Dom is having doubts using StaticBackend for his SaaS
Give it a listen!
We also have some very interesting episodes coming out soon with the creator of Wails @leaanthonycymru as well as @thorstenball, the co-founder of AMP
@yatish_me I'm using gonertia (https://t.co/8jzR5YNnLF) so more interested in the andurel specific API than other starter kits
I'm adding SSR support now and also adding inertia run cmd to shadowfax so everything works together
Considering openapi style client generation for fe/be comms
kimi 2.7 code on @FireworksAI_HQ is crazy fast
might be realistic to have a similar spend as the 200$ plans using these professionally mon-fri
obviously not the same amount tokens but could be done
@yatish_me Thanks @yatish_me! I haven't used inertia before (building a project with it right now) so I'd love feedback/help with the API and structure if you've the time 🙏
codex to "make modern dashboard, use shadcn, make no mistakes" is a real time save
all public or user facing pages are still done by hand using templ+datastar
but i got to admit, llms are great at writing javascript
testing out the latest version of andurel by building out my consultancy site
having recently added inertia support it's a good candidate for testing out that feature since I don't really care about how the dashboard is made, just need to be there
adding shadcn and telling...