AI 客栈
返回首页

AWS 发布 Strategy Mosaic 语义层教程：MCP 接入与 RLS/OLS 贯穿 Agent 查询

  • 核心内容：同系列另一篇教程使用 Strategy Mosaic 的维度语义层，连接 Amazon Aurora PostgreSQL 与 Amazon Redshift，通过两个 linked models 在 Mosaic 引擎内合并中间结果；AgentCore 托管 Strands agent，经 Mosaic MCP endpoint 调用 get_projects、get_models、get_semantics 与 query 四个工具。
  • 身份与权限：前端使用 OAuth 2.0 Authorization Code flow，refresh token 由 AWS Secrets Manager 保存；Mosaic 在查询前校验 Use MCP Server 权限、certified model 门槛，并对调用者执行 RLS 与 OLS。
  • 多接口复用：同一套 Mosaic 模型除 MCP 外，还可通过 JDBC/ODBC、XMLA 与 REST 服务 BI 工具、SQL 客户端和自定义应用；教程同时说明 live pushdown 与 in-memory 两种执行模式的取舍，以及 Apache Ossie 格式的导入导出。
  • 行业意义：教程展示了“一次定义、多接口复用”的语义层路径：Agent 先读取指标定义、层级与关系再生成查询，BI 与 SQL 客户端复用同一套治理规则；对多工具并存的企业分析栈，这减少了指标定义在 Agent 与 BI 之间漂移的空间。 🔗 AWS Machine Learning Blog
Reference architecture. Inbound authentication path and outbound query path between the client, Amazon Bedrock AgentCore components, the Strategy Mosaic MCP endpoint, the Mosaic engine, and Amazon Redshift plus Amazon Aurora.Customer attribute definition in Design Center. cid is the ID form used to join on the underlying key. Full Name, First Name, and Last Name are display forms.Geography hierarchy in Design Center. Attributes chain from State through City, ZIP Code, and Location to Customer, with Rewards and phone attributes hanging off the customer node.Purchase Amount metric authored in Design Center. The formula is Price * Quantity. The description tells the agent what the metric means in business terms.Advanced-metric authoring in Design Center, with the base formula and filter condition shown on the Breakdown tab.Identity propagation chain from the human user through the enterprise IdP, Amazon Bedrock AgentCore, AWS Secrets Manager, and the Mosaic MCP endpoint. The front-door OAuth handshake runs once per user (steps 1-4). The per-request path runs on every agent tool call (steps 5-10).