/ /

$vectorSearch（聚合阶段）

定义

$vectorSearch

$vectorSearch 对Atlas 集群中的数据执行语义搜索。

If you store vector embeddings on your Atlas cluster, you can seamlessly index the vector data along with your other collection data. You can use the $vectorSearch stage to pre-filter your data and perform semantic search against the indexed fields.

矢量数据的宽度必须小于或等于 8192 维度。

兼容性

$vectorSearch聚合管道阶段在以下环境中可用：

MongoDB Atlas clusters running version 6.0.11 or later
MongoDB Enterprise 部署运行版本为 8.2 或更高版本的 Kubernetes 操作符
MongoDB Community 运行版本8.2或更高的部署

To learn more, see MongoDB Vector Search.

行为

从MongoDB8.0 开始，您可以在阶段中使用$vectorSearch $unionWith阶段。

限制

您不能在$vectorSearch $facet阶段或$lookup 阶段中使用阶段。

了解详情

To learn more about creating MongoDB Vector Search indexes, see Index Vector Embeddings.
To learn more about $vectorSearch pipeline stage syntax and usage, see Vector Search Queries.