跳到正文

Amazon Bedrock Managed Knowledge Base 新增表格搜索功能

AWS Machine Learning Blog当事人· Josh DeMuth·· 8 小时前精选Get answers from tables with tabular search in Amazon Bedrock Managed Knowledge Base
打开原文
原文
图片来自 AWS Machine Learning Blog
谁
AWS
引文
In this post, you learn how tabular search answers questions that need both table data and facts from documents, and when another tool is a better fit.

发生了什么

Amazon Bedrock Managed Knowledge Base 现已通过智能体检索(agentic retrieval)支持表格搜索。对于需要跨 CSV、XLSX 或 XLS 文件进行计算或查询的问题,它会在隔离沙盒中对原始表格运行代码,并将结果与相关文档段落相结合。该示例演示了一个涉及工作簿、政策文档和结账备忘录的财务问题;表格分析会增加一轮智能体交互,因此耗时比纯文本回答更长。

推荐理由

团队可以对完整的 CSV 和 Excel 表格进行计算,同时将结果与相关文档中的规则相结合,从而让多步骤业务问题在单次智能体检索请求中同时利用这两种数据源。

接下来读什么

查看这件事的完整进展AWS Adds Tabular Search to Amazon Bedrock Managed Knowledge Base

这件事到目前为止

AWS added tabular search to Amazon Bedrock Managed Knowledge Base through agentic retrieval, allowing it to answer questions that combine table data with document passages. When retrieved passages come from a CSV, XLSX, or XLS file and a question requires calculations across rows or lookups of specific values, the system runs code over the original tables in an isolated sandbox and merges the results with related text. AWS said the feature adds at least one extra agent turn, making answers that use table data take longer than text-only responses. As of the announcement, tabular search can run code on CSV files up to 500 MB, matching the ingestion limit, and on Excel sheets up to 200 MB. AWS also noted that governed metrics, recurring dashboards, warehouse-scale joins, and reporting controls still belong in business intelligence tools.

这条收录于 11月17日 AI 日报看这一期日报的全部内容