Creating visual schemas by AI or assistance

API
1

Hello everyone, I want to ask the team about something. There’s a program that works with diagrams (in the engineering field). The diagrams are in an open format, JSON. How could we make it so that AI learns to write them correctly? Right off the bat, it writes some nonsense, with incorrect fields in the file. Can this be improved somehow by invoking some web methods? Maybe the team can suggest how web methods could help AI improve diagram generation?

2

Can you give a slightly more concrete example/sample? I’ve found I have to in my instructions describe the format of the JSON I want to get it close. I still then have to run regex on it to delete the ```json it starts the message with followed by doing it again at the end to get rid of the same thing at the end of the message.

3 
{ "class": "go.GraphLinksModel",
  "copiesArrays": true,
  "copiesArrayObjects": true,
  "linkFromPortIdProperty": "fromPort",
  "linkToPortIdProperty": "toPort",
  "nodeDataArray": [
{"key":1, "name":"Unit One", "loc":"101 204",
 "leftArray":[ {"portColor":"#fae3d7", "portId":"left0"} ],
 "topArray":[ {"portColor":"#d6effc", "portId":"top0"} ],
 "bottomArray":[ {"portColor":"#ebe3fc", "portId":"bottom0"} ],
 "rightArray":[ {"portColor":"#eaeef8", "portId":"right0"},{"portColor":"#fadfe5", "portId":"right1"} ] },
{"key":2, "name":"Unit Two", "loc":"320 152",
 "leftArray":[ {"portColor":"#6cafdb", "portId":"left0"},{"portColor":"#66d6d1", "portId":"left1"},{"portColor":"#fae3d7", "portId":"left2"} ],
 "topArray":[ {"portColor":"#d6effc", "portId":"top0"} ],
 "bottomArray":[ {"portColor":"#eaeef8", "portId":"bottom0"},{"portColor":"#eaeef8", "portId":"bottom1"},{"portColor":"#6cafdb", "portId":"bottom2"} ],
 "rightArray":[  ] },
{"key":3, "name":"Unit Three", "loc":"384 319",
 "leftArray":[ {"portColor":"#66d6d1", "portId":"left0"},{"portColor":"#fadfe5", "portId":"left1"},{"portColor":"#6cafdb", "portId":"left2"} ],
 "topArray":[ {"portColor":"#66d6d1", "portId":"top0"} ],
 "bottomArray":[ {"portColor":"#6cafdb", "portId":"bottom0"} ],
 "rightArray":[  ] },
{"key":4, "name":"Unit Four", "loc":"138 351",
 "leftArray":[ {"portColor":"#fae3d7", "portId":"left0"} ],
 "topArray":[ {"portColor":"#6cafdb", "portId":"top0"} ],
 "bottomArray":[ {"portColor":"#6cafdb", "portId":"bottom0"} ],
 "rightArray":[ {"portColor":"#6cafdb", "portId":"right0"},{"portColor":"#66d6d1", "portId":"right1"} ] }
 ],
  "linkDataArray": [
{"from":4, "to":2, "fromPort":"top0", "toPort":"bottom0"},
{"from":4, "to":2, "fromPort":"top0", "toPort":"bottom0"},
{"from":3, "to":2, "fromPort":"top0", "toPort":"bottom1"},
{"from":4, "to":3, "fromPort":"right0", "toPort":"left0"},
{"from":4, "to":3, "fromPort":"right1", "toPort":"left2"},
{"from":1, "to":2, "fromPort":"right0", "toPort":"left1"},
{"from":1, "to":2, "fromPort":"right1", "toPort":"left2"}
 ]}

The format is quite simple, and AI even sometimes writes it correctly. However, the problem is that even a minor error can lead to complete distortion. And it’s not even clear how to train for this.

4

Ok, so I may not have the best answer here but now having seen your JSON what is it you want the ai to fill in? Also, how does the AI get this information?