j0d1

The AI spits out _some_ code => the developer needs to review & cleanup the mess, or accept that it worsen the system for the sake of delivering more rapidly.

That's the dynamic with AI and a lot of developers hate working like that, yet there is pressure to work this way.

That's really cool! I am working on a diagramming editor and I wondered if I could expand the use cases to include knowledge organization. The challenge is mainly to display a lot of data on a canvas and don't kill the CPU/GPU in the process. How do you handle displaying that much information at once?

The self-hosting option is nice but, from what I understand, I would not even need to do that if you provided end-to-end encryption with PGP. i.e. the form is encrypted with my public key, then sent to your server, then sent to my e-mail, then decrypted with my private key.

With such option, the self-hosting is still a nice to have, if you cease to provide the service at formbee.dev.

This is really neat! I would love to see something similar for laptops. I bought a used Lenovo T80s (8th generation i5 CPU / 8GB of RAM / 256GB SSD) for 150$cad on eBay to work on my product (web app) and it is working flawlessly with Debian.

I fixed the issue, you can reload the page!

Fixed in production!

Thank you for your comment!

I decided to work on gg because I want my colleagues to grok our complex software architecture, and I don't feel like I am able to achieve that with textual documentation + static diagrams alone.

I have been in a lot of brainstorming sessions drawing boxes and arrows on a whiteboard. I have produced a lot of diagrams with mermaid, draw.io, miro, etc. I produced a lot of documentation to explain how the software is built. Those are all good tools to get everyone on the same page, and yet, I feel like there is a missing piece of the puzzle to explain a software architecture simply and concisely that even the new junior developer recruit will understand.

For those reading this, do not click on that link above. It executes a script that renders the app unusable until you clear your local storage.

Thanks for the report, I'll fix that.

Yes, me too :) I am working on a viewer / presentation mode that will do just that.

I think it is a great idea!

There is no plan yet to support importers directly in gg but the graphs produced with gg have a YAML/JSON representation under the hood so producing a "draw.io to gg" importer is feasible.

Thank you for your comment :)

I'm working on supporting webm first, as I am close to have a working feature using the MediaRecorder API of the browser. Exporting to gif will come after that, as it can be done by third-party tools as a workaround in the meantime (i.e. webm to gif).

Someone submitted an issue on Github to add labels on links (i.e. lines) and I think it's a great idea for a v1. In my mind, the animations won't get in the way of reading the labels, but implementation will tell! I will also probably implement them like I implemented titles on systems, which is to avoid routing links over them.

> "Animated diagrams" are great for generating engagement on social media, but in practice an animated line doesn't tell you anything an arrow wouldn't.

Instead of showing a little green box moving from system A to system B, I could have shown static arrows alongside the link between system A to system B, or something similar.

I chose the moving green box as it represents the data flowing in the system, in a tangible / concrete manner.

It communicates:

- A data is created/produced/etc. at system A.

- The data is transferred/copied/moved/etc. to the system B.

- The data is received/consumed/etc. by the system B.

I'll create an issue to support exporting the animation in v1. Any particular reason for APNG vs GIF?

The animation in the README.md was captured with the MediaRecorder API. It produced a webm that I "cropped -> clipped -> to gif" manually. I am not that far to produce a looping webm that is perfectly cropped & clipped. I may release that first, and then tackle gif / apng / ??? later.

My algorithm to detect / resolve collisions has a bug in it. In some cases, a system will be displaced on the right, then on the left, then on the right... until it reaches a maximum number of allowed displacements. When it happens, moving a system feels very slow.

I intend to fix that bug (eh!) and I'm also thinking about moving this costly algorithm in web workers (+ parallelize it).