Create AI Service Connection File - Open AI | Error when create *.ais file

[IgnacioYañez]

I try to Create AI Service Connection File (Data Management) and use into Classify Objects Using Deep Learning (Image Analyst) using the Vision Language Context-Based Classification Deep Learning Package.

When i try to create the OpenAI connection, the Use the model document says:

To connect directly to OpenAI, set the service_provider to "OpenAI" and provide your OpenAI API key using the api_keyfield. You must also specify the model to use by setting the deployment_name field (e.g., "gpt-4o" or "gpt-4"). The remaining parameters can be left blank when connecting directly to OpenAI, as they are only required when using OpenAI services hosted on Azure.

But, when i try to configure the deployment_name field, appears an error

Error 004120: Invalid connection parameter. The correct connection parameter for the <value> service provider is <value>".

ArcGIS Pro 3.5

Description

An incorrect connection parameter was provided for a specific service provider.

Solution

Verify the correct connection parameter for the specified service provider and update it accordingly.

Wht'a the correct way to create an OpenAI Connection?

Thanks! 

[Muhammad_Elbagoury]

Hi @IgnacioYañez ,

I just tested the Create AI Service Connection File tool with OpenAI, and it worked correctly when I set the connection parameter as follows:

Key: Model
Value: gpt-4o

After creating the connection file this way, it was successfully recognized by the Vision Language Context-Based Classification package in ArcGIS Pro.

Hope this helps!

[IgnacioYañez]

Thanks for your help Muhammad. I already have the connection file!

Saddly the geoprocessing don't work for me 😪

Can you share me how you can use the model?

[Muhammad_Elbagoury]

Your are welcome.
I’m actually facing this strange scenario too — the tool runs, but the output shows ClassLabel and Reason fields with unknown values.
I’m still investigating why this happens and will update once I figure out the cause.

[Muhammad_Elbagoury]

Hello again,
I just want to follow up if you were able to have your results after the new update released.
I resolved it but just checking that with you.

Regards

[AduriSreenivas]

Architecture Discussion — External Workflow Orchestration with ArcGIS Enterprise Services

Hi all,

Following recent guidance around third-party AI integration in ArcGIS Pro (e.g., .ais connections and DLPKs), I’m looking at an architecture where an external orchestration layer coordinates multi-step workflows, while ArcGIS Enterprise remains responsible for all spatial processing and data management.

The intent is not to move analysis outside ArcGIS, but to manage workflow sequencing, monitoring, and retries externally when required.

Concept (high level)

External orchestration layer

• Controls workflow execution and sequencing
• Calls ArcGIS REST services

ArcGIS Enterprise

• Feature Services → data access
• GP Services → spatial analysis
• Hosted Feature Layers → persisted outputs

All geometry processing remains server-side in ArcGIS.

Example workflow

Use case: parcel suitability analysis.

Typical flow:

• FeatureServer/query — retrieve inputs
• GPServer/submitJob — run analysis
• Poll async job status
• Write results via applyEdits or GP output layer

The orchestration layer acts only as a controller.

Questions

1. Is external orchestration of FeatureServer + GPServer workflows a common / recommended enterprise pattern?
2. Key production considerations (job polling, token lifecycle, service limits, GP timeouts, handling intermediate outputs)?
3. Any real-world examples using a microservice / workflow engine / scheduler driving ArcGIS mainly via REST?

Interested to hear how others approach orchestration while keeping ArcGIS as the core spatial execution platform.

Thanks for any insights.