Also I think the DX of lazygit is impressive

I use lazygit inside nvim. It feels really nice. I use Snacks plugin for lazygit
github.com/folke/snacks...

Snacks is freaking amazing. I use a lot of things from it

ok 😂, but no seriously what's that? a static site in Liveview?

What I'm watching?

I lost a nice package name while i was working on the thing. I prefer to own the name before releasing

This is the present if you want typed Phoenix routes in your Typescript code #elixirlang
hex.pm/packages/way...

Holy shit, saving jobs with this PR

Do you have some implantation inside ash code on how to make ecto a replaceable piece? I'm not promising nothing but sounds an interesting thing to check

It's ok, for now is just a learning tool for me. I built typescript generation porting Wayfinder from Laravel to Phoneix and you shipped rpc typescript xD
All are exercises. If later are helpful for someone is a nice byproduct

Also related to this @jakeprem.com commented that maybe how Rails handle polymorphism is not the most idiomatic way in Ecto and pointed me to this doc

bsky.app/profile/jake...

What's your input on this? really interesting in hearing it

I was thinking on relying on Ecto as a hard requirement. But totally open to listen to ideas.
What other storage would make sense for handling file storage. Sorry, too many storages

...commands I'm trying `Makefile`. Don't know if nothing of this works but it looks professional 😂

...individual package development. I'm probably wrong but is my feeling.

So now that I choose (3) I have the problem of reusing common parts like tooling for format, credo dyalizer and alias.

This is done in a helpers/mix_helperx.ex that I use in each package.

Regarding on how to run dev

... development of the packages

1. Different repos for each package
2. Monorepo with an Umbrella project
3. "packages" folder with different mix projects

I'm going with (3) because I feel I dont want to fight Umbrella projects. I think is better for apps development more than for...

(1) is simpler, but final users pay the price of installing, let's say, Azure and Google dependencies when they just want to

(2) split in different hex package and avoid installing unnecessary dependencies on final apps.

I go with (2) but here I have three more choices on how to handle ...

"...to the attachments ledger table. Then all attachment records point to the ledger table, using it as a sort of constrained join table."

This is in your README. How this work? how you achieve DB consistency?

Super nice documentation! Thanks

"To that end it (currently) uses a Ledger approach, where polymorphism is achieved by adding a foreign key from the main table (e.g. Post) ..."

I understand your point but i think a full implantation of active storage adds a lot of value in a lot of common cases. I'll try to build a prototype and report how it goes