Alibaba Cloud launches Elasticsearch Agent Engine for multi-tenant AI search
Alibaba Cloud introduced the Elasticsearch Agent Engine at the 2026 Yunqi Conference as search infrastructure for Agent contexts, alongside OpenSearch-AgenticSearch. The engine uses object storage, separate read and write layers, and Collection/Slice abstractions for multi-tenant retrieval; Alibaba Cloud also reported that DiskBBQ, FalconSeek 2.0, and elastic scaling were part of the broader AI search product line. Alibaba Cloud stated that, under a reference workload of 10 billion documents and more than 10,000 knowledge bases, hot-query P99 was about 60 milliseconds, vector recall exceeded 95%, and overall cost was about 70% lower than self-built Elasticsearch.
The release positions search and context retrieval as core infrastructure for enterprise Agents, RAG, AI coding, and Agent memory, with Alibaba Cloud targeting large-scale multi-tenant workloads.
作者：阿里云计算平台AI搜索产品负责人 汤祯捷
2026 年云栖大会，成为过去十年最为火爆的科技盛会。而在 AI 搜索分论坛上，一个共识被反复提及：AI大模型解决了"理解与生成"，但没有解决"在正确的时刻搜索到正确的信息"。
当 Agent 从单轮问答走向长任务执行，搜索的角色发生了根本变化——它不再是一个供人浏览的结果页，而是 Agent 上下文工程的供给系统。上下文是否准确、完整、可溯源，直接决定 Agent 后续的推理与行动质量。
围绕这一变化，阿里云 AI Search 产品线在本届云栖大会集中发布两条主线产品能力：
| 主线 | 产品 | 一句话定位 |
|---|---|---|
| 新引擎 | 阿里云 Elasticsearch | 成为 AI 时代 Agent 上下文的搜索基础设施 |
| 新范式 | 阿里云 OpenSearch-AgenticSearch | 定义 AI 搜索时代的信息检索新范式 |
#Alibaba Cloud#Elasticsearch Agent Engine#OpenSearch-AgenticSearch#OpenStore#FalconSeek 2.0#DiskBBQ#Alibaba Cloud Elasticsearch 9.4#OpenSearch