Vertex AI Search Application for Structured data returning generic error "Problem loading results"

Google Cloud Build with AI Custom ML & MLOps
1

Hello,

I have a persistent issue for Vertex AI app. In the preview section I constantly get the ‘Problem loading results, please try again.’ error and nothing seems to resolve it. Here are all the steps I have tried and confirmed so far.

  1. My data is structured and comes from a BigQuery table. It is formatted and ingested correctly, one document per row, title is a unique ID and the data is contained in a key-value pairs json format. I also tried ingesting through a file from Cloud Storage
  2. I select Structured data as my Data Source and all the fields are the correct data types. There are 2 string fields and 5 number fields
  3. Both of the service accounts have the needed access to BigQuery and both BQ and AI Apps are in the same project
  4. There is nothing helpful in the logging - no errors, just a few notice events related to the ingestion
  5. The example outlined here throws the same generic error when I follow it step by step: https://cloud.google.com/generative-ai-app-builder/docs/try-enterprise-search
  6. I tried with generic words and different data sizes. I also tried waiting, this is the 3rd day in a row I am researching this.
  7. When I tested the Vertex feature 7 months ago, it was returning results.

I am based in Europe and it says so in the Documents tab of the Data Store. It probably has nothing to do with the issue but is it possible the feature is somehow restricted here?

2

Hi @stan-hennessey,

Welcome to the Google Cloud Community!

It looks like you are encountering an issue caused by a mismatch between your structured JSON data and your Data Store schema configuration in Vertex AI Search, which may be preventing your app preview from loading results correctly.

Here are the potential ways that might help with your use case:

  • Update Schema Definitions: You may want to begin by testing your simplified BigQuery table that includes only a title and content text columns; if you’re working with JSON blobs, ensure your Data Store schema defines them as an Object type with each subfield including two strings and five numbers accurately mapped, typed, and marked as displayable.
  • Configure Search Application Display Fields: In your Search Application’s Configuration tab, make sure you’ve selected the correct Title and Content fields from your indexed schema, and that both fields are marked as displayable in your Data Store schema.
  • Validate Runtime Permissions: Make sure your Compute Engine default service account (PROJECT_NUMBER-compute@developer.gserviceaccount.com) has the necessary roles, such as Vertex AI User, to access indexed data during runtime.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.