Kind of a shame this didn't get more traction, because this has a lot of potential.
I ran a few generations with this using Anthropic models and tested it against a data layer (to test UML), a systemprompt (mind map) and various bits of code (flowcharts etc.)
It's kind of fascinating, the initial diagram is alright, you can see where it might be going, but by the time selft-critique round 4 rolls around, the diagram is actually useful.
Worth playing with it. I tried abither approach where I feed a codebase into Claude, let it document the flows and feed the result into diagen for visualization, also works well as an alternative to letting it roam over a code file directly.
Kind of a shame this didn't get more traction, because this has a lot of potential.
I ran a few generations with this using Anthropic models and tested it against a data layer (to test UML), a systemprompt (mind map) and various bits of code (flowcharts etc.)
It's kind of fascinating, the initial diagram is alright, you can see where it might be going, but by the time selft-critique round 4 rolls around, the diagram is actually useful.
Worth playing with it. I tried abither approach where I feed a codebase into Claude, let it document the flows and feed the result into diagen for visualization, also works well as an alternative to letting it roam over a code file directly.
Very cool stuff.
I'm not sure I would call their sample diagram in the README beautiful or reflective.