LIVE

Amazon Bedrock Adds Tabular Search to Managed Knowledge Bases for Multi-Step Data Queries

🏢 Big Tech 2 min read 👁 2

Amazon Web Services has introduced tabular search for Amazon Bedrock Managed Knowledge Base, enabling agents to execute calculations and lookups across CSV and Excel spreadsheets alongside unstructured documents.

Amazon Bedrock Adds Tabular Search to Managed Knowledge Bases for Multi-Step Data Queries🏢Big Tech

Photo: AWS Machine Learning Blog

Key points

  • Amazon Bedrock Managed Knowledge Base now supports tabular search for CSV and Excel files.
  • The system runs code in an isolated sandbox to execute calculations across entire sheets alongside document retrieval.
  • Integration works natively with existing agentic retrieval flows across sources like Amazon S3 and SharePoint.

Amazon Web Services (AWS) has updated its Amazon Bedrock Managed Knowledge Base service to include tabular search capabilities. The feature is designed to help artificial intelligence agents answer complex analytical queries by pulling data from structured table files such as CSV, XLSX, and XLS, while simultaneously pulling facts from related documents like PDFs and Word files.

Traditional retrieval-augmented generation often struggles with tables because splitting rows into standard passages can strip away context and fail to match business terms to raw numeric data. The new capability addresses this by using retrieval to identify the correct file, after which an isolated sandbox environment executes code across the entire table or sheet. This approach allows the system to perform calculations and lookups over every relevant row.

According to AWS, the functionality integrates directly into existing agentic retrieval workflows via the AgenticRetrieveStream request. It supports various data connectors, including Amazon Simple Storage Service (Amazon S3), Microsoft SharePoint, and Atlassian Confluence. The system returns answers that combine tabular calculations with document facts, complete with source citations and execution traces.

What to watch next: Monitor how enterprise adoption scales for hybrid document-and-spreadsheet workflows, and observe potential latency impacts as multi-step agentic execution adds extra turns to query response times.

Source AWS Machine Learning Blog — aws.amazon.com/blogs/machine-learning/get-answers-from-tables-with-tabu… ↗

Contact the newsroom

Got a tip, a proposal or a partnership question? Write to us and we will reply shortly.