AI Model Predict action started failing with The API operation 'PredictV2' requires the property 'body/responsev2/predictionOutput/Prediction' to be of type 'Integer' but is of type 'Float'.'

[RPA_trainer]

My flow with custom AI model was running fine since few days and it started failing suddenly with error The API 'commondataserviceforapps' returned an invalid response for workflow operation 'Predict' of type 'OpenApiConnection'. Error details: 'The API operation 'PredictV2' requires the property 'body/responsev2/prediction Output/Prediction' to be of type 'Integer' but is of type 'Float'.'  
I have checked on the input data and it appears all columns are integer but still it fails with same error.  Even the action was successful in previous runs then output was in decimal, not sure what is affecting this.

Can anyone experiencing the same.

[CedrickB]

Can you share a screenshot of your flow so we can get more context to understand the possible cause please?

[RPA_trainer]

I am trying with all columns as 1 just to rule out any data issue.

[CedrickB]

Apparently you are passing a value to one of the fields of your prediction which is a float (Something like 41.23) instead of an expected integer  (something like 41).

Check your input parameters, what can do is using a conversion to int for all your parameters.

See Solved: Convert a float variable to INT - Power Platform Community (microsoft.com)

[RPA_trainer]

to avoid this I tried to pass 1 to every column and it did worked couple of times but after one day it started to fail again.

[CedrickB]

Can you send an email to aihelpen@microsoft.com, it looks like we will need a sharing session to investigate.

[MelMacLondon]

Good morning, I have a similar problem.. sometimes there will be text in the item but sometimes float.

'The API operation 'PredictV2' requires the property........... ...........  to be of type 'String' but is of type 'Float'.'

I am happy if the AI detects it as a string (to cater for when someone has entered "n/a" as an example)

Is there any work around for this?

[Antrod]

Hi @MelMacLondon ,

Unfortunately, you can't provide a string if it requires a float. If that makes some sense in your business case, you can perhaps turn 'n/a' to a specific float value before calling the predict.

[MelMacLondon]

Unfortunately, it might not just be N/A - could be a variety of hand written notes, "not Applicable", "couldn't find it", "general comment" .... 

[Antrod]

Perhaps you can test if the input is of type float and if not, apply the same float value for every string?

[MelMacLondon]

The error is occurring on the actual AI Builder model... I do not get any results out to do any such test. I would be happy with just string values for all

[BlvckAI]

i Have the same error message and even if i try to convert the passed value it still gives me this exact error.

what i try to do:

i tried: INT(value) which should make any gives value an integer. how do i solve this

test run with error: as you can see G/L account is just a 3 so i assume its a float. even if its not i tried to convert it with INT(). yet is still given the same error.

[BlvckAI]

This doesnt work. its still gives the same error.