Make ML Skills and Datasets public via URL + API Key - Preview

We are pleased to announce the launch of very demanded feature:

Make ML Skills and Datasets public via URL + API Key

What does this enable?

  • Deploying ML Skills and continuously improving them on AI Center running in the Automation Cloud without needing to migrate from your existing on-premises Orchestrator.
  • Deploying ML Skills and continuously improving them outside of Robots (e.g. add them to your UiPath Apps or even use them within a 3rd party application).

How does this work

The first step is to expose your entities (Datasets and ML Skills) as an endpoint to be able to call them from your on-premises (or not connected) infrastructure.

  • Make Dataset public

To make your dataset your dataset public simple select the option while creating or updating your dataset:

Then in dataset details you will see the endpoint and corresponding API Key, you have the possibility to change the API Key but note that old one will stop working and old processes may break.

  • Make ML Skills public

To make your ML Skills public, first deploy your ML Skill as usual and then change deployment to make it public:

updateSkill

Again this will expose and endpoint and corresponding API Key to consume this endpoint.
Note: Document Understanding Skills won’t expose an API Key as you can simply use your Document Understanding API Key for all of them. This means that API Key needs to be the one associated with the account where your Skills are deployed.

  • Call them from Studio

Download and install new ML Services Activities package (v1.1.6-preview) from preview feed. Both ML Skill and upload file activity have a new field call Connection Mode. Select endpoint and complete the details for accessing your entity (endpoint and API Key) as String.
For ML Skill, you also need to manually set the type of data that you want to send (Remember that JSON is JSON as a string here).

You are all set, you can now start building your intelligent automation process and complete the feedback loops from your on-premises existing infrastructure to AI Center on UiPath Cloud.

20 Likes

@loginerror @Jeremy_Tederry
what are the different Check the functions, methods and libraries used to work with outlook , rest api and pdf files

@loginerror @Jeremy_Tederry
How to mark the voting buttons of outlook emails without UI interface ?
there are 8 customized options marked on different conditions .
can we do with Exchanged Web services or graph api or any other method not using screen scrapping but through backend ?

Hello @Jeremy_Tederry , I did try to make one skill public, but it failed complaining that I need a necessary license. I’m on an Enterprise plan as part of the MVP program. Am I missing something? The Skill I’m trying to make public is an Untrainable one.

Thanks

Hmmm… I seem to in the AI Center Twilight zone these past few days. I enabled the Auto update option and gave it a shot, and it worked!

Does that not make Auto update a required feature?

:thinking: :confused:

Is there any documentation on accessing the public AI skill from an external client such as a Web page?

Do you mean by skipping the activities and performing the API call to the service on your own? I don’t think this is officially supported (@Jeremy_Tederry will correct me on that, I’m sure).

Hi @loginerror . I’m confused. The original post by @Jeremy_Tederry states the following:

Can you kindly clarify ?

Thanks

1 Like

So true :sweat_smile:

I fiddlered around and found out this:

curl --location --request POST 'yourURL' \
--header 'X-UIPATH-License: yourAPIKey' \
--header 'Content-Type: application/json' \
--data-raw '{"data":"test"}'
2 Likes

And I post manned it :laughing: And Success! First response from publicly accessible AI Skill!

This is great!

Thanks for the help @loginerror :+1:

1 Like

And to round this off, made a quick video on the subject.

3 Likes

Hi @loginerror, do you have API documentations for this ? I have cases to utilize this public API, especially to upload files into The Datasets from external system.

I got results ‘400 Bad Request’ when trying to access the ML skill from postman.

FYI, I’m using DocumentUnderstanding IDCards packages, so the data is image base64.

does anyone else know how to fix it?

any update on this?

hi @loginerror , Would you be able to provide the same for datasets URL and API, It seems to not work with this format.

Any work around to fix this issue even I am getting the Predication failed error with status code 400 Bad Request

Hello there,

Managed to make this work for me.
After training the OOB image classifier from AI Center and making it public, I called the engine from Postman using the base64 image as payload.
I simply converted the image using this online tool and added it in Postman like this:


Content-Type header is set to “application/json”.
Make sure to add the encoded string, not the image source (the one with data:image/img extension;base64) as the API call responds with “Prediction Failed” otherwise.

2 Likes

Managed to upload files in Dataset using Postman, using a REST API call as opposed to the Upload File activity. For some use cases, it might be better to be able to add the files directly from the API call as opposed to involving additional UiPath Processes