You need managed RAG over documents in S3, with the least custom code. What is the best fit?
Why C is the answer
Amazon Bedrock Knowledge Bases
Knowledge Bases handles ingestion, chunking, embeddings and retrieval as a managed service.
In detail
- Knowledge Bases gives managed RAG over S3 with little code.
- Build your own pipeline only when you need control the service lacks.
Use case
An illustrative scenario, not a real incident.
A team wants its S3 manuals searchable by an assistant this week. Knowledge Bases handles chunking, embeddings and retrieval, so it writes little code.
See it in different ways
mindmap
root((RAG))
Ingest
Sync data source
Detect personal data
Chunk
By structure
Hierarchical
Search
Vector
Keyword
Rank
Rerankerflowchart LR S0["S3 documents"] --> S1["Ingestion and chunking"] --> S2["Embeddings"] --> S3["Retrieval"]
Key point: Knowledge Bases is managed RAG.
Why the other options are wrong
- A. A SageMaker training job
- A training job does not serve retrieval.
- B. Amazon Athena queries over the PDFs
- Athena queries tables, not PDFs.
- D. A Lambda function that greps S3 objects
- grep is not semantic retrieval.