Run-Time variable Key JSON Parsing

[breiter56]

Will this solution work? Parse JSON to Cluster - NI Community
It seems like it's possible but it would be creating a VI for each result?
Do you think I could I manage some raw string parsing to return the predictions value as an array instead of objects?

PS, I do know how to handle the JSON string into a cluster, if it were a constant structure.

I'm trying to handle the results of an AI inference in LandingAI's Landing edge. The inference result payload is a JSON formatted string. (either by using their web API and HTTP or by reading a .json file it generates).  The "predictions" object is variable. Each time it runs it returns different keys, so I'm lost, how to handle it. 
Payload 1:
{"type":"ObjectDetectionPrediction","predictions":{"4adc9111-c13e-4ee0-9e44-ad945447d94f":{"score":0.7076729,"labelIndex":1,"defectId":208450,"labelName":"Ding","coordinates":{"xmin":1111,"xmax":1144,"ymin":596,"ymax":621}},"c1e50fcb-89e4-4673-a7d9-32a3fbd18ecb":{"score":0.3431902,"labelIndex":1,"defectId":208450,"labelName":"Ding","coordinates":{"xmin":1325,"xmax":1353,"ymin":550,"ymax":574}}},"metadata":{"image_id":"","inspection_station_id":"","location_id":"","capture_timestamp":"2024-09-21T07:17:45.7689464-05:00"},"model_id":"354f3887-f9eb-413b-aaee-97d80a51c068","latency":{"decoding_s":0.0094819,"preprocess_s":0.0018961,"waiting_s":8E-07,"infer_s":0.6588475,"postprocess_s":2.38E-05}}

Payload 2:
{"type":"ObjectDetectionPrediction","predictions":{"834c9ac4-fef0-4c51-ad36-208c73794270":{"score":0.7076729,"labelIndex":1,"defectId":208450,"labelName":"Ding","coordinates":{"xmin":1111,"xmax":1144,"ymin":596,"ymax":621}},"8b89faf1-6067-49e1-8f81-3c0cd8319c09":{"score":0.3431902,"labelIndex":1,"defectId":208450,"labelName":"Ding","coordinates":{"xmin":1325,"xmax":1353,"ymin":550,"ymax":574}}},"metadata":{"image_id":"","inspection_station_id":"","location_id":"","capture_timestamp":"2024-09-21T07:17:56.7923955-05:00"},"model_id":"354f3887-f9eb-413b-aaee-97d80a51c068","latency":{"decoding_s":0.0080896,"preprocess_s":0.0024012,"waiting_s":9E-07,"infer_s":0.6673403,"postprocess_s":3.43E-05}}

If no object was detected, predictions are an empty object:
{"type":"ObjectDetectionPrediction","predictions":{},"metadata":{"image_id":"","inspection_station_id":"","location_id":"","capture_timestamp":"2024-09-21T07:32:28.3631828-05:00"},"model_id":"354f3887-f9eb-413b-aaee-97d80a51c068","latency":{"decoding_s":0.0063953,"preprocess_s":0.0021194,"waiting_s":1E-06,"infer_s":0.6943869,"postprocess_s":1.08E-05}}

[drjdpowell]

Try this library: https://www.vipm.io/package/jdp_science_jsontext/

[cordm]

You can do it in one go with JSONText if you define the "predictions" as a map.

[breiter56]

Thank you. I am exploring the add-in and learning the Path notation. I have no experience with the Map datatype. I assume if would figure out how to use it and the 'Find Item (as LVType)"?
In the mean-time, it looks like I can find predictions by Name and just use the array length to determine accept reject. I'll expect I'll need to work towards checking a size threshold against the coordinates as well.