Oriel — AI & technology worth noticing, every day.

Get answers from tables with tabular search in Amazon Bedrock Managed Knowledge Base

Web page · https://aws.amazon.com/blogs/machine-learning/get-answers-from-tables-with-tabular-search-in-amazon-bedrock-managed-knowledge-base

Amazon Bedrock Managed Knowledge Base now supports tabular search, so agentic retrieval can answer questions over CSV, XLSX, and XLS files. It looks up values and runs calculations across every relevant row, then combines the results with facts from related documents to answer multi-step analytical questions.

First seen by Oriel on .

AI summary

This AWS post introduces tabular search in Amazon Bedrock Managed Knowledge Base, which combines calculations over CSV or Excel tables with facts from related documents through agentic retrieval.

  • Tabular search runs model-written code over complete CSV, XLSX, or XLS files in an isolated, network-free sandbox, avoiding incomplete results from passage retrieval.
  • Agentic retrieval retrieves relevant passages, plans table operations, stages original files, computes results, and combines printed outputs with document passages and citations.
  • The finance example calculates revised variance percentages across every department and identifies Sales, Engineering, and Operations using policy, memo, and workbook data.
  • Tabular search adds at least one agent turn and is slower than text-only answers; governed metrics, dashboards, warehouse-scale joins, and reporting controls remain suited to BI tools.
  • The walkthrough uses Amazon S3, a managed knowledge base, Python 3.10 or later, Boto3 1.43.123 or later, and permissions including bedrock:GetDocumentContent.

#amazon bedrock #tabular search #agentic retrieval #rag

Generated by gpt-5.6-sol. Check the original for details.

Aliases

SourceExternal IDURL
AWS Machine Learning Blog3813dbf1ca3824362df03525d76f7db4f5be7fd5https://aws.amazon.com/blogs/machine-learning/get-answers-from-tables-with-tabular-search-in-amazon-bedrock-managed-knowledge-base/ Read discussion