Internal documents in S3 will be used by Amazon Bedrock Knowledge Bases for retrieval-augmented generation.
Split long documents into searchable units and ingest them with metadata.
Which processing approach is most appropriate?
Chunk the documents, generate embeddings and metadata, and synchronize them into the knowledge base vector store.
Detailed explanation
A single huge chunk gives coarse retrieval and can exceed useful context limits.
A single huge chunk gives coarse retrieval and can exceed useful context limits.
Chunking creates retrieval units while embeddings and metadata support semantic and attribute-based search.
Chunking creates retrieval units while embeddings and metadata support semantic and attribute-based search.
Audit events are separate data and add unrelated context to business documents.
Audit events are separate data and add unrelated context to business documents.
ACLs describe access control and do not represent the meaning of document content.
ACLs describe access control and do not represent the meaning of document content.
Check official documentation
Compare the following points with the official documentation.
Amazon Bedrock Knowledge Basesのデータソース、チャンク、埋め込み、同期の公式説明を確認する。
What to confirm
文書変換とアクセス制御・監査イベントの役割を区別できる。
Key points
- Bedrock Knowledge Bases
- Chunking
- Embedding
- Metadata
Notes
- Environment: AWS公式DEA-C01試験ガイドと各サービスの公式ドキュメントの確認
- Confirm current AWS service behavior in the official DEA-C01 documentation.
Foundation review
Retrieval quality
Chunk size and overlap affect relevance, context limits, latency, and cost.
Metadata filter
Attributes can restrict retrieval to documents or chunks meeting a condition.