We’re implementing an enterprise AI knowledge base using Vertex AI and have discovered a significant discrepancy between the documentation and actual UI capabilities regarding grounding with multiple data sources. We need clarification on the correct architectural approach.
Our Use Case:
-
Internal knowledge base for Marketing and Development teams
-
Multiple data sources: Google Drive, Slack, Jira, Confluence, GitLab
-
Need unified AI-powered search and question answering across all sources
Documentation vs Reality Gap: The Vertex AI Search Grounding documentation states: “Grounding to your data supports a maximum of 10 Vertex AI Search data sources and can be combined with Grounding with Google Search.”
However, in Vertex AI Studio, the grounding configuration UI only allows specifying a single data store path, not multiple sources. This creates a fundamental architectural limitation.