Alibaba Cloud Elasticsearch Agent Engine Edition is a cloud-native search engine for retrieval-augmented generation (RAG), code search, agent memory, and multi-tenant search. It hosts knowledge bases, code repositories, and agent context retrieval on horizontally scalable Collections and Slices, with pay-as-you-go storage and read and write resources that you configure on demand. For the complete product positioning, performance reference, and version information, see What is Agent Engine Edition.
Core concepts
Agent Engine Edition organizes retrieval workloads around the following concepts:
Collection — The unified access point through which applications connect to Agent Engine Edition.
Slice — A unit that organizes a knowledge base, code repository, agent, or tenant within a Collection, without creating a separate index for each one.
Backing — A resource that you pre-create before you provision tenants or import data in bulk.
Architecture
Agent Engine Edition is built on the Elasticsearch 9 series and decouples persistent storage from indexing and query workloads:
Storage — Index data and persistent state reside in object storage.
Index nodes — Handle writes and index building.
Search nodes — Handle queries.
DiskBBQ — A disk-native vector index.
Product capabilities
The following topics describe the capabilities of Agent Engine Edition:
Get started
If you are new to Agent Engine Edition, complete the following steps. Before you begin, prepare the network and account for the instance. If you plan to provision tenants or import data in bulk, pre-create Backings and register Slices first.
Create and connect to an instance: verify the connection, and complete the create, write, and query operations in the examples that retrieve by tenant or by knowledge base.
When you develop applications, use API reference to look up requests, parameters, responses, and error handling. Before you provision tenants or import data in bulk, follow Manage Collections and Slices to pre-create Backings and register Slices.
Before you go live, review Limits, read Instance Operations, and use your business data to verify capacity, query quality, and latency.