Error "LLM configuration is invalid: connected model cannot be reached" when configuring Google Vertex AI Agent

[rpa.eie.spain]

Hi Community,

I am encountering a blocking issue when trying to configure a Conversational Agent in UiPath Studio Web using the Google Vertex AI connection.

Current Scenario (What works): I have successfully set up the connection in the Integration Service (Status is “Connected”). To verify the credentials and permissions, I performed a test in UiPath Studio (Desktop):

• I used the activity “Generate Text Completion using Gemini”.
• I selected the model gemini-2.0-flash.
• Result: The activity executes successfully and generates a response. This confirms that the Service Account has the necessary Vertex AI User role and access to the GCP project.

The Issue (What fails): The problem occurs when I try to use this same connection to build an AI Agent in Studio Web:

1. In the Agent configuration, I select the Vertex AI connection.
2. I choose the “Custom Model” option (as the model is not in the dropdown).
3. I manually input the model name (I have tried gemini-2.0-flash and gemini-1.5-flash).
4. When attempting to validate or test, I get the following error:

LLM configuration is invalid: connected model cannot be reached

¿Has anyone experienced this discrepancy where the connection works in standard activities but fails in the Agent Builder? Does the “Custom Model” field require a specific syntax (e.g., the full path publishers/google/models/...) or are there stricter IAM requirements (like Vertex AI Viewer) for the Agent validation process compared to standard runtime execution?

Any guidance would be appreciated.

[Anil_G (Anil Gorthi)]

@rpa.eie.spain

For now can you create a connection in any non shared folder and try to connect from that connection please

cheers

[rpa.eie.spain]

I just successfully created a new Google Vertex connection in the Integration Service within ‘My Workspace’.

However, when I try to configure the custom model, I can only select connections from ‘Shared Folders’; I cannot select the one from ‘My Workspace’.

¿Any ideas?

[Anil_G (Anil Gorthi)]

@rpa.eie.spain

can you try creating a standalone folder and not under shared

go to tenant → folders → create folder

create connection in that and then try with the same

personal workspace might not be exposed

cheers

[rpa.eie.spain]

Same error… I created a new project in Studio Web, and it doesn’t work either.

I don’t understand why it’s not working… it’s frustrating.

[Anil_G (Anil Gorthi)]

@rpa.eie.spain

Please right click inspect and check for errors in calls …That might show what error we are seeing

cheers

[rpa.eie.spain]

This is my Request:

And this is the Response:

{
  "agents": {
    "isAvailable": false,
    "isCompatible": false,
    "isModelNameSimilar": true,
    "unavailabilityReason": "VendorInvalidRequest",
    "vendorRawResponse": {
      "error": {
        "code": 400,
        "message": "Invalid JSON payload received. Unknown name \"if\" at 'tools[0].function_declarations[0].parameters': Cannot find field.\nInvalid JSON payload received. Unknown name \"then\" at 'tools[0].function_declarations[0].parameters': Cannot find field.\nInvalid JSON payload received. Unknown name \"else\" at 'tools[0].function_declarations[0].parameters': Cannot find field.",
        "status": "INVALID_ARGUMENT",
        "details": [
          {
            "@type": "type.googleapis.com/google.rpc.BadRequest",
            "fieldViolations": [
              {
                "field": "tools[0].function_declarations[0].parameters",
                "description": "Invalid JSON payload received. Unknown name \"if\" at 'tools[0].function_declarations[0].parameters': Cannot find field."
              },
              {
                "field": "tools[0].function_declarations[0].parameters",
                "description": "Invalid JSON payload received. Unknown name \"then\" at 'tools[0].function_declarations[0].parameters': Cannot find field."
              },
              {
                "field": "tools[0].function_declarations[0].parameters",
                "description": "Invalid JSON payload received. Unknown name \"else\" at 'tools[0].function_declarations[0].parameters': Cannot find field."
              }
            ]
          }
        ]
      }
    }
  }
}

I tried changing the model to several others, but the result is the same.

[Anil_G (Anil Gorthi)]

@rpa.eie.spain

looks like using Open AI V1 connector might help here

Introducing: Bring Your Own Model to UiPath | UiPath

cheers

[rpa.eie.spain]

I have generated a new OpenAI V1 Compliant LLM connector and tried to use it with a “Custom model” configuration within an Agent in Studio Web. However, I am now encountering the following error:

I’ve attached the Response here in case it helps.

{
    "agents": {
        "isAvailable": true,
        "isCompatible": false,
        "isModelNameSimilar": true,
        "incompatibilityReason": "MULTIPLE_TOOL_CALLS_WHEN_PARALLEL_TOOL_CALLS_NOT_ALLOWED"
    }
}

With the current OpenAI V1 Compliant configuration, the “Model” section now shows many “pre-loaded” models (which didn’t happen before), but even so, it is still not working.

Any ideas?

[Anil_G (Anil Gorthi)]

@rpa.eie.spain

Looks like many things are down

@loginerror @Samrat_Dutta - Any help here please

cheers