Why A is the answer

Synchronization of the knowledge base data source

Keep the index in step with the source data, so that retrieval stays fresh.

In detail

• Sync the data source, so the index follows the documents.
• Alert when a sync fails.

See it in different ways

mindmap
  root((RAG))
    Ingest
      Sync data source
      Detect personal data
    Chunk
      By structure
      Hierarchical
    Search
      Vector
      Keyword
    Rank
      Reranker

stateDiagram-v2
  state "Source changes" as s0
  state "Sync runs" as s1
  state "Index current" as s2
  [*] --> s0
  s0 --> s1
  s1 --> s2
  s2 --> s0

Why the other options are wrong

B. A larger chunk overlap in the existing index

Overlap does not add new documents.

C. A manual upload of each changed document by the support team

Manual uploads are error prone.

D. A full rebuild of the model each night, using the latest documents

A model rebuild is far too much.

Related questions

• Question 1You need managed RAG over documents in S3, with the least custom code. What is the best fit?
• Question 2Retrieved chunks are relevant but often cut a clause in half, so answers miss conditions. What do you change first?
• Question 12Teams should switch models without code changes. What do you design?
• Question 16Retrieval misses exact product codes, while meaning-based queries work well. What helps?

Learn more

• Our lesson: Models, data and retrieval
• Bedrock Knowledge Bases
• Bedrock Prompt Management
• Generative AI Lens