Release notes for OpenSearch Vector Search Edition-OpenSearch(Open Search)-阿里云帮助中心

OpenSearch Vector Search Edition feature releases and updates.

2026

May

Date	Feature	Type	Description
2026-05	Product feature updates	New	Introduces a vector caching feature that uses a `cache table` to automatically compare field content during data ingestion. The model recomputes vectors only for changed data, significantly reducing model API calls and computational costs. Introduces the `object field type` and `nested query`, allowing you to store multiple fields in a nested structure and retrieve data from it to support complex query models.

April

Date	Feature	Type	Description
2026-04	Product feature updates	New	Adds a vector fusion feature that lets you use a `multimodal model` to encode content from multiple fields, such as text and image, into a single vector. This enables cross-modal retrieval.

March

Date	Feature	Type	Description
2026-03	Product feature updates	New	Adds support for vector models from `Model Studio`, enabling vectorization of multimodal data such as text and images during preprocessing for index building and queries. Supported models include multimodal and text vector models.

2025

January

Date	Feature	Type	Description
2025-01-02	Product feature updates	New	The CAGRA algorithm now supports GPU acceleration. You can purchase GPU-based nodes to achieve significant performance improvements in high-QPS scenarios. Supports `hybrid search`. Build an `inverted index` on text fields and combine `full-text search` with vector search to improve search quality in RAG and semantic search scenarios. Adds support for range indexes. A range index queries documents within a specific range and significantly outperforms a range filter in a filter clause. You can specify whether to create an index that has the same name as a field. Supports statistical syntax.

2024

December

Date	Feature	Type	Description
2024-12-05	Product feature updates	New	Introduces the `QGraph algorithm`, an improvement on HNSW, which automatically quantizes raw data during index building. This reduces index size to 1/8 of the original and delivers performance several times that of HNSW. Supports the JSON format for files from a `full data source` in `Object Storage Service (OSS)`.

September

Date	Feature	Type	Description
2024-09-27	Product feature updates	New	Integrates with `AI Search Open Platform`, allowing you to call platform `model services` such as embedding and `OCR` directly from the `console` when configuring fields. Supports automatic data cleanup based on `time to live (TTL)`. Supports custom resource policies for table loading.
2024-09-10	Product feature updates	New	Supports `Data Lake Formation (DLF)` as a `full data source`.

July

Date	Feature	Type	Description
2024-07-01	Product feature updates	New	Adds a log query feature for viewing `query and write logs`. Supports table aliases. Supports custom default values for fields.

June

Date	Feature	Type	Description
2024-06-04	Feature optimization	Optimized	Improved table creation time. Supports adjusting the number of `QRS worker replicas`. Improved the API.

March

Date

Feature

Type

Description

2024-03-26

Product feature updates

New

The console now supports vector data management, including adding, deleting, and previewing data.
Updated monitoring metrics.
Improved the Change History FSM.
Shard decoupling: The number of table shards is now decoupled from the number of Searcher workers, and shard counts are also decoupled between tables.
Supports adding remarks to fields.
Supports schema import and export.

January

Date	Feature	Type	Description
2024-01-17	Product feature updates	New	The `console` supports primary key queries, image or text uploads for search, and hybrid queries. Supports multi-subject recognition for image search.
2024-01-08	Product feature updates	New	Supports `hybrid retrieval`. Supports configuring multiple vectors for a single `document`. Adds a new text-to-vector model that supports automatic vectorization for Chinese and English text with token lengths of 128, 512, and 1024.

2023

December

Date	Feature	Type	Description
2023-12-26	Instance specification adjustment	New	Supports purchasing a `single Searcher worker`.

October

Date	Feature	Type	Description
2023-10-12	Instance specification adjustment	New	New, smaller node specifications are available: `QRS worker`s start at 2 vCPUs and 8 GB of memory, and `Searcher worker`s start at 2 vCPUs and 16 GB of memory. `QRS worker`s and `Searcher worker`s support both cloud disk and `SSD` types. Each `single instance` now includes 100 GB of free index storage. Adjusted the `instance rental fee`.
2023-10-10	Product feature updates	New	Adds an enhanced e-commerce image-to-vector model. Supports resource updates based on data changes from the current `full version`.

September

Date	Feature	Type	Description
2023-09-07	Product redesign	New	Improved query methods. Upgraded `SDK` versions. `SDK` for the `new engine` (Developer Guide) `SDK` for the old engine ( Developer Guide (ha3_3.10.0))

August

Date	Feature	Type	Description
2023-08-31	Product feature updates	New	Table configuration now supports `Object Storage Service (OSS)` data sources. Removes the `TEXT field type`.

July

Date	Feature	Type	Description
2023-07-21	Product redesign	New	OpenSearch Vector Search Edition is now generally available.

June

Date	Feature	Type	Description
2023-06-07	Product feature updates	New	Adds the `ignore_invalid_doc` and `enable_rt_build` configurations for vector indexes. Supports configuring an `instance alert` at the account level. Adds the `Pause Switch` feature. Supports dictionary content upload from a file. Adds an incremental `text vector template`.

March

Date	Feature	Type	Description
2023-03-30	Product feature updates	New	Introduces end-to-end image search. This includes support for image embedding to enable `text-to-image search` and `image-to-image search`. Supports `data compression` for fields, attributes, and indexes. Adds `internet access` and `IP whitelist` features. Adds support for data retrieval by using a `RESTful API`. Adds the Vector: Image Search template.

January

Date	Feature	Type	Description
2023-01-09	Product feature updates	New	Adds multiple `built-in analyzer`s, including an entertainment content analyzer, a general English e-commerce analyzer, an IT content analyzer, an e-commerce analyzer, a person name analyzer, and a geographic location analyzer. All `built-in analyzer`s support dictionary intervention. Supports user-defined `multi-value delimiter`s. Infrequently used advanced configuration parameters for vector indexes are collapsed by default to simplify configuration and O&M. Adds alert monitoring for when the index change rate exceeds a specified threshold or when full and incremental updates are not completed within a specified period. You can customize alert thresholds. The Change History page now displays offline full-build metrics, which allows you to monitor the build process.

2022

September

Date	Feature	Type	Description
2022-09-21	Product feature updates	New	Removes the DUP requirement for vector fields, adds field order restrictions for vector indexes, expands the field types supported by vector indexes, and adds default values for advanced vector index configurations to simplify setup. Supports setting a `timestamp` by using natural time. On the Deployment Management page, you can switch index tables to go live. All data sources, index tables, and clusters are displayed on the Deployment Management page.

August

Date

Feature

Type

Description

2022-08-16

Product feature updates

New

1. Adds index data recovery and quick shard scaling.

2. Improves alerting and adds a default alert for the Vector Search Edition.

May

Date

Feature

Type

Description

2022-05-20

Product feature updates

New

1. Adds quick cluster scale-out, click-based index structure configuration, query configuration in online configurations, and scenario-based instance O&M.

2. Vectors can be used as plug-ins in the sorting phase. Supports multi-value indexes and range indexes.

2021

December

Date	Feature	Type	Description
2021-12-31	Official release of OpenSearch Vector Search Edition	New	OpenSearch Vector Search Edition is officially released, providing a highly available, low-latency, and cost-effective solution with automated O&M for building custom search services.