AWS 展示 SageMaker AI 与 Bedrock AgentCore 构建多智能体工作流
原标题：Building agentic workflows with SageMaker AI and Bedrock AgentCore
AI 摘要
AWS 展示了如何将 SageMaker AI 上的 OpenAI 兼容端点与 Bedrock AgentCore 运行时结合，构建多智能体工作流。该架构通过 Bedrock AgentCore 容器连接编排器（Claude Haiku 4.5）、预算智能体（Claude Sonnet 4.6）和财务分析智能体（Qwen 3.5 9B），实现成本优化、数据驻留和模型灵活性。文章重点介绍了集成机制，包括如何通过自定义 OpenTelemetry 跨度获取 SageMaker 端点的令牌级可观测性，以及 vLLM 流式响应中 stream_options 参数的必要性。
正文节选
Building agentic workflows with SageMaker AI and Bedrock AgentCore A common challenge in building agentic workflows is mixing managed foundation models (FMs) with your own cost-optimized or domain-specific models, without rewriting your agent framework to do it. In this post, we show you how to combine OpenAI-compatible endpoints on Amazon SageMaker AI with Amazon Bedrock AgentCore runtime, a capability of Amazon Bedrock AgentCore, and its managed deployment. Specialized agents can collaborate o