@Joshua_Allan_0 unfortunately (and fortunately!) this is a known issue currently in progress with the Studio Team and if all goes well, to be fixed in 23.10 
As a workaround, would you mind mind reading the activity every time you want to modify the prompts, or using Studio Web? We know this is not a nice experience yet, but again, we’re working towards fixing it!
I have tried extraction and classification and it is really game changer.
Observed below issues in activities hope this will be resolved in upcoming upgrade
*** Incase i am changing the prompt then getting error in output variable of activity where we need to delete the variable and recreate it which resolve the issue however ideally this should not happen as a person could change/add the prompt multiple times to get better result and this could be hectic since we have to delete and recreate variable every single time when we change the prompt.**
*** Also, sometime the same activity doesn’t extract the data however when we delete the activity and same activity reused in project then it extracts the data which i observed when i did multiple invoice testing.**
This could be replacement for orthodox Document understanding.
Great work by UiPath
hello @prem.SR - thanks a lot for your feedback, I’m very happy to hear you are enjoying our generative capabilities and our new set of activities!
with regards to your 2 points:
- Incase i am changing the prompt then getting error in output variable of activity where we need to delete the variable and recreate it which resolve the issue however ideally this should not happen as a person could change/add the prompt multiple times to get better result and this could be hectic since we have to delete and recreate variable every single time when we change the prompt.
I assume this is the experience of Studio desktop - correct me if I’m wrong. If so, know that this is a known issue of ours: Extract Document Data generates the output variables - anytime you change a config in it be it either another model selected or modified prompt inputs - however, Studio desktop cannot currently generate variables - this is in progress with the team and we’re hoping to have a fix for it soon, planned for 23.10
- Also, sometime the same activity doesn’t extract the data however when we delete the activity and same activity reused in project then it extracts the data which i observed when i did multiple invoice testing.
This indeed seems very strange - can you give us some steps/documents to reproduce the issue?
1 Like
can we connect to showcase the issue
send me an invite on monica.secelean@uipath.com 
hi @Monica_Secelean in the future would you consider increasing the max character limit for the prompt values, and probably the option to select other LLM like claude-2 to enjoy a much larger token limit (up to 100k), so we can provide a much larger instruction?
@Monica_Secelean is this a known issue? I’m not doing anything special, but I’m using local studio (Windows compatibility not cross platform)
I am facing the same issue in studio. What do you mean by “reading” each time I modify the prompts.
Error: System.InvalidCastException: Unable to cast object of type ‘UiPath.IntelligentOCR.StudioWeb.Activities.DataExtraction.DocumentData1[UiPath.IntelligentOCR.StudioWeb.Activities.SWEntities.CustomGptDocumentType5Bf9D38142214Db7B3F5701284B5Db76.Bundle.CustomGptDocumentType5Bf9D38142214Db7B3F5701284B5Db76]' to type 'UiPath.IntelligentOCR.StudioWeb.Activities.DataExtraction.IDocumentData1[UiPath.IntelligentOCR.StudioWeb.Activities.SWEntities.CustomGptDocumentType1F254C847D904Ded9E3BB5C1Af772447.Bundle.CustomGptDocumentType1F254C847D904Ded9E3BB5C1Af772447]’. at Namespace_1923.Main_Expressions.Main_Expressions_TypedDataContext3.__Expr5Set(IDocumentData1 value) at Namespace_1923.Main_Expressions.Main_Expressions_TypedDataContext3.ValueType___Expr5Set(IDocumentData1 value)
at System.Activities.XamlIntegration.CompiledLocation1.set_Value(T value) at System.Activities.Location1.ReferenceLocation.set_Value(T value)
at System.Activities.ActivityContext.SetValueCore[T](LocationReference locationReference, T value)
at System.Activities.ActivityContext.SetValue[T](OutArgument1 argument, T value) at System.Activities.OutArgument1.Set(ActivityContext context, T value)
at UiPath.IntelligentOCR.StudioWeb.Activities.ExtractDocumentDataWithDocumentData`1.<>c__DisplayClass30_1.b__1(AsyncCodeActivityContext endContext)
at UiPath.Shared.Activities.AsyncTaskCodeActivityImplementation.EndExecute(AsyncCodeActivityContext context, IAsyncResult result)
at UiPath.IntelligentOCR.StudioWeb.Activities.DataExtraction.ExtractorAsyncTaskActivity.EndExecute(AsyncCodeActivityContext context, IAsyncResult result)
at System.Activities.AsyncCodeActivity.System.Activities.IAsyncCodeActivity.FinishExecution(AsyncCodeActivityContext context, IAsyncResult result)
at System.Activities.AsyncCodeActivity.CompleteAsyncCodeActivityData.CompleteAsyncCodeActivityWorkItem.Execute(ActivityExecutor executor, BookmarkManager bookmarkManager)
We will update the prompt limit to 500 - doesn’t this suffice ? If not, you may want to consider a specialized extraction model maybe 
not sure on your use case though!
@connor.orzen I meant that, every time you change the prompt, the output object gets regenerated - as this doesn’t happen (because of our known issue), you face the error 
sorry for that, the fix for it is planned for this month! for the moment, we recommend you remove the EDD activity and add it again - sorry for the inconvenience!
Thanks so much for the quick response! I found that rather than delete and re-configure the activity, the actual fix was to delete the output variable and re-initialize it after any prompt changes.
Hope this helps anyone else effected until the patch goes live.
One last question, what licensing does the generative preview require/what will it once it goes live? Currently I’m not seeing it pull any AI units licensing and am wondering if it will when out of “preview” stage. Anyway, absolutely love the new range of possibilities and ease of use enabled by generative capabilities.
2 Likes
Hi @Monica_Secelean thanks for the response. one of the area a longer prompt character limit will be useful is for us is (1) to provide more specific and detailed instructions about the output format or (2) as workaround for tabular or dictionary type of outputs, combining description of multiple fields and merging them in one json formatted field
@connor.orzen sorry for my late reply - thanks a lot for your tips for other users encountering the same issue! In the meantime, the issue should be fixed - but let us know if you still face it Wrt licensing & charging: find here details - let me know if it does not help, happy to take other questions 
@zell12 hey Russel! we’ve increased the prompt size to 500 chars - is this sufficient? let me know otherwise
Are these activities going to be free or any plans to charge in the future?
@AI_GPT find here details about our charging logic: Document Understanding - Metering and Charging Logic - we do however provide some free usage to our community users hope it helps, let me know otherwise!
1 Like
Thanks for sharing the details. how does metering work as we are not passing the AI units API key for these activities?
Hello,
I’m trying to use the classify and extract document data activities, with the generative classifier/extractor however both encounter this error:
I have setup a prompt like so:
However, I notice in the locals panel, the value being passed into Prompt is incomplete. Any idea why this might be?
Edit: I solved this by upgrading studio to 23.10 (was previously on 22.10.10).
@AI_GPT we charge the tenant you run the automation in
Ok, Thanks for clarification. You mean the Org level, as the allocation of AI units on Tenant level is not possible. May I know why the license consumption model has been updated? Is there any way to limit the license consumption through a user role, As I do not want every one connected to tenant can consume the AI units.