# AWS 博客介绍零件替代多智能体助手：规则评分，决定权留给工程师 > AWS 机器学习博客2026年11月2日发布技术文章，介绍用 Amazon Nova Multimodal Embeddings、Amazon Bedrock AgentCore 和开源 Strands Agents SDK 构建多智能体工程助手，用于工程零件替代评估中的证据汇编、确定性规格评分和评估文档草拟；文章称工程师保留决定权，系统不自行批准替代。 - 栏目：云计算与工程软件 - 本站发布：2026-11-05 17:49（北京时间） - 原始消息发布：2026-11-03 01:01（北京时间） - 消息来源：AWS 机器学习博客 (https://aws.amazon.com/blogs/machine-learning/agentic-engineering-intelligence-with-amazon-nova-multimodal-embeddings-and-amazon-bedrock-agentcore/) - 本页网址：https://jecy.cloud/news/articles/2026-11-02-1a2ef60cffc0.html ## 文章展示的助手与设计规则 AWS 机器学习博客2026年11月2日发布技术文章，介绍一套用于工程零件替代评估的多智能体助手。文章称，这套系统把此前的多模态检索能力接入多智能体流程，用于汇编替代证据，工作从检索资料延伸到草拟评估。文章将一条设计规则贯穿全文：工程师做每一个决定，系统只负责汇编证据和草拟文档，不会自行批准替代。因此，文章展示的是评估前的证据准备和文档草拟，而非自动审批；系统把检索结果和后续评估材料组织成证据包。[^1][^2] ## 用公差规则代码做规格适配 规格适配由Python公差规则代码完成：每个候选零件与参考零件以及工程师提出的要求比对打分，文章称这一层为“确定性规格适配评估”。相同输入会得到相同的通过或标记结论，因此结果可进入正式工程评审。每次评估还会生成一条审计记录，文章称该记录与AS9100第8.4和8.5.6节对齐。规则代码给出规格适配的通过或标记结论，最终决定仍由工程师做出。[^3][^4][^5] ## 图像查询走单独索引路径 文章还说明一条实现限制：工程师上传图表图片作为查询、用来识别零件时，代理直接调用Amazon S3 Vectors索引，因为托管知识库的检索API不接受图像输入。图像查询因此没有走托管知识库的统一检索接口，而需要单独的向量索引路径。文章以技术实现说明为主，没有给出该系统在真实制造流程中的验证结果，也没有说明它已投入生产使用；同样没有报告准确率、处理时间或替代评估成功率等指标。[^6] --- ## 原文摘录 [^1]: "This post picks up where that retrieval work left off, putting those capabilities inside a multi-agent system that assembles substitution evidence, from retrieval to a drafted assessment." — AWS 机器学习博客 (https://aws.amazon.com/blogs/machine-learning/agentic-engineering-intelligence-with-amazon-nova-multimodal-embeddings-and-amazon-bedrock-agentcore/) [^2]: "The engineer remains the decision-maker throughout, because the system assembles evidence and drafts documents without ever approving a substitution on its own." — AWS 机器学习博客 (https://aws.amazon.com/blogs/machine-learning/agentic-engineering-intelligence-with-amazon-nova-multimodal-embeddings-and-amazon-bedrock-agentcore/) [^3]: "Three pillars define the architecture: Deterministic spec-fit evaluation – Tolerance-rule code written in Python scores each candidate part against the reference part and the engineer’s stated requirements." — AWS 机器学习博客 (https://aws.amazon.com/blogs/machine-learning/agentic-engineering-intelligence-with-amazon-nova-multimodal-embeddings-and-amazon-bedrock-agentcore/) [^4]: "Because the same inputs consistently produce the same pass/flag verdicts, the results can go into a formal engineering review." — AWS 机器学习博客 (https://aws.amazon.com/blogs/machine-learning/agentic-engineering-intelligence-with-amazon-nova-multimodal-embeddings-and-amazon-bedrock-agentcore/) [^5]: "Every evaluation also writes an audit trail entry aligned with AS9100 sections 8.4 and 8.5.6." — AWS 机器学习博客 (https://aws.amazon.com/blogs/machine-learning/agentic-engineering-intelligence-with-amazon-nova-multimodal-embeddings-and-amazon-bedrock-agentcore/) [^6]: "For image-as-query (uploading a chart to identify a part), the agent calls a native Amazon S3 Vectors index directly, because the managed knowledge base retrieval API doesn’t accept image input." — AWS 机器学习博客 (https://aws.amazon.com/blogs/machine-learning/agentic-engineering-intelligence-with-amazon-nova-multimodal-embeddings-and-amazon-bedrock-agentcore/) --- 说明：本文由 AI 依据信源整理，并经事实核对；文中数字与成绩均为来源方公布的口径，未经本站独立验证，原文更新后以原文为准。欢迎 AI 与读者引用，请注明来源并链接回本页。