Plan and manage an Azure AI solution 1.1 Choose the appropriate Foundry services for generative AI and agents
You're building a Foundry IQ knowledge base.
Study guide skill: Choose an appropriate method for retrieval and indexing
Question B4-003
You're building a Foundry IQ knowledge base. Content comes from (1) an existing, tuned Azure AI Search index that you want to reuse as is, and (2) a SharePoint site whose content must not be copied into Azure AI Search and should be retrieved at query time. Which knowledge sources should you create?
Select one answer.
Answer: B
A search index knowledge source wraps an existing index, so the tuned index is reused as is. A remote knowledge source never ingests content into Azure AI Search; it retrieves at query time through the platform's APIs, which is what remote SharePoint does. Indexed SharePoint and Azure blob knowledge sources generate an indexer pipeline that ingests content, and a file knowledge source uploads files into Azure AI Search.
Deep dive
How it works
A Foundry IQ knowledge base draws on knowledge sources, which are independent objects you create separately and then attach. Each one is either indexed or remote.
Indexed knowledge sources have content in a search index before query time, through one of three paths:
- Bring your own index: a search index knowledge source wraps an existing index on your search service. Nothing is rebuilt, so your tuning is kept.
- Direct file upload: a file knowledge source (preview) uploads files directly to Azure AI Search, which stores the extracted content in a generated index.
- Auto-generated indexer pipeline: Azure blob, Azure SQL, OneLake, indexed SharePoint and others make Azure AI Search create a full pipeline (data source, skillset, indexer and a populated, chunked index).
Remote knowledge sources connect directly to an external platform. Content is never ingested into Azure AI Search; it's retrieved at query time through that platform's native APIs. Remote SharePoint (preview) is one of them. Because there's no backing index, remote sources don't return citation URLs.
So requirement 1 (reuse the tuned index as is) is a search index knowledge source, and requirement 2 (don't copy SharePoint, retrieve at query time) is remote SharePoint. Results from both still flow through the same ranking pipeline and are merged and reranked.
The other options, and when you'd use them
- File knowledge source for SharePoint + blob for the index: A file source uploads documents into Azure AI Search, which is the copying you must avoid. A blob source builds an indexer pipeline from a blob container; it doesn't wrap an existing index.
- Blob for the index + remote SharePoint: The SharePoint half is right, but a blob knowledge source generates a new pipeline from a container instead of reusing your index.
- Search index + indexed SharePoint: The index half is right, but indexed SharePoint generates an indexer pipeline that ingests the site's content into Azure AI Search.
Exam tip
"Reuse my index" = search index knowledge source. "Don't copy, query live" = a remote knowledge source. Anything that generates a pipeline or uploads files is indexed, so it copies content.
References
Last reviewed: