Release notes

Release date

Feature

Type

Description

2025-01-02

Feature addition, update, or optimization

New

1. The CAGRA algorithm is introduced to support GPU acceleration, and GPU-accelerated nodes can be directly purchased. This can significantly improve performance in high QPS scenarios.

2. Hybrid searches based on text and vectors are supported. Inverted indexes can be created based on text fields, and text vectors can be added to combine full-text searches and vector searches. This improves the search effect in RAG scenarios and semantic search scenarios.

3. Range indexes are supported to enable queries of documents in a specific range. When a range index is used to replace range filtering specified in a filter clause, the query performance is significantly improved.

4. Whether to create an index with the same name as a field can be specified based on your business requirements.

5. Statistical syntax is supported.

Release date

Feature

Type

Description

2024-12-05

Feature addition, update, or optimization

New

1. The QGraph algorithm is supported. This algorithm is developed based on the Hierarchical Navigable Small World (HNSW) algorithm. During index creation, QGraph automatically quantizes raw data. It can reduce the index size to 1/8 of the size of raw data and improve performance by several times compared with HNSW.

2. JSON files can be used as full Object Storage Service (OSS) data sources.

Release date

Feature

Type

Description

2024-09-27

Feature addition, update, or optimization

New

1. AI Search Open Platform is integrated with OpenSearch Vector Search Edition. During field configuration in the OpenSearch Vector Search Edition console, model services such as embedding and optical character recognition (OCR) can be directly called in the OpenSearch Vector Search Edition console.

2. Expired data can be automatically deleted based on specific time to live (TTL).

3. Policies used to load resources for tables can be customized.

2024-09-10

Feature addition, update, or optimization

New

1. Data Lake Formation (DLF) data sources can be used as full data sources.

Release date

Feature

Type

Description

2024-07-01

Feature addition, update, or optimization

New

1. Query logs and write logs can be viewed on the Log Query page.

2. Aliases can be created for tables.

3. Default values of fields can be customized.

Release date

Feature

Type

Description

2024-06-04

Feature optimization

Optimized

1. The performance of table creation is optimized.

2. The number of Query Result Searcher (QRS) worker replicas can be modified.

3. API performance and parameters are optimized.

Release date

Feature

Type

Description

2024-03-26

Feature addition, update, or optimization

New

1. Vectors can be managed in the OpenSearch Vector Search Edition console, such as adding, deleting, and previewing a vector.

2. Monitoring metrics are updated.

3. The finite-state machine (FSM) process on the Change History page is optimized.

4. The number of table shards is decoupled from the number of Searcher workers. The number of shards is decoupled between tables.

5. Comments can be added for fields.

6. Schemas can be exported or imported.

Release date

Feature

Type

Description

2024-01-17

Feature addition, update, or optimization

New

1. Primary key-based queries, queries based on uploaded images or text, and hybrid queries are supported in the console.

2. Multi-subject recognition in image search scenarios is supported.

2024-01-08

Feature addition, update, or optimization

New

1. Hybrid retrieval is supported.

2. OpenSearch Vector Search Edition allows you to configure multiple vectors for a single document.

3. A text-to-vector conversion model is added. The model can automatically convert Chinese or English text that consists of 128, 512, or 1,024 tokens to vector data.

Release date

Feature

Type

Description

2023-12-26

Instance specification adjustment

New

Single-worker instances can be purchased.

Release date

Feature

Type

Description

2023-10-12

Instance specification adjustment

New

1. QRS workers and Searcher workers with smaller specifications are available for purchase. The minimum specifications of a QRS worker are 2 vCPUs and 8 GB of memory. The minimum specifications of a searcher worker are 2 vCPUs and 16 GB of memory.

2. QRS workers and searcher workers support cloud disks and SSDs.

3. A free quota of 100 GB of storage space is available for indexing in a single OpenSearch instance.

4. The instance rental fee is adjusted.

2023-10-10

Feature addition, update, or optimization

New

1. An enhanced image-to-vector conversion model is added for e-commerce.

2. Resources can be updated based on the data changes of the current full version.

Release date

Feature

Type

Description

2023-09-07

Service upgrade

New

1. The data query feature is optimized.

2. OpenSearch Vector Search Edition SDKs are updated.

   1. For more information about OpenSearch Vector Search Edition SDKs for the new engine version, see Developer guide (vector_service_1.0).

   2. For more information about OpenSearch Vector Search Edition SDKs for the old engine version, see Developer guide (ha3_3.10.0).

Release date

Feature

Type

Description

2023-08-31

Feature addition, update, or optimization

New

1. OSS data sources are supported.

2. The fields of the TEXT type are canceled.

Release date

Feature

Type

Description

2023-07-21

Service upgrade

New

The new version of OpenSearch Vector Search Edition is officially released. The new version is easy to use.

Release date

Feature

Type

Description

2023-06-07

Feature addition, update, or optimization

New

1. The ignore_invalid_doc and enable_rt_build parameters are added for vector indexes.

2. Instance alerting can be configured based on accounts.

3. The Pause Switch feature is added.

4. Dictionaries can be configured by uploading files.

5. A template of word embedding is added.

Release date

Feature

Type

Description

2023-03-30

Feature addition, update, or optimization

New

1. New features: The image search capability is added. OpenSearch Vector Search Edition supports image embedding and implements features such as searches for images based on specific images or text.

2. New feature: Data compression is supported, including the compression of fields, attributes, and indexes.

3. New feature: Access over the Internet is supported. IP addresses whitelists are supported.

4. New feature: Data can be queried by using RESTful APIs.

5. Optimized feature: The Vector: Image Search template is added.

Release date

Feature

Type

Description

2023-01-09

Feature addition, update, or optimization

New

1. New features: A variety of built-in analyzers and their intervention dictionaries are added. The built-in analyzers include the entertainment content analyzer, general analyzer for e-commerce for English, IT content analyzer, e-commerce analyzer, person name analyzer, and geographic location analyzer.

2. New feature: Custom multi-value delimiters are supported.

3. Optimized feature: Advanced parameters that are seldom configured for vector indexes are hidden by default. This helps simplify the configuration and O&M.

4. New feature: More monitoring metrics are supported. You can receive an alert notification if the index change rate exceeds a specific threshold or the full or incremental index data is not updated for a specific period of time. The threshold or period of time can be customized.

5. Optimized features: Metrics about offline services during reindexing are displayed in the change history. The index building progress can be viewed.

Release date

Feature

Type

Description

2022-09-21

Feature addition, update, or optimization

New

1. Optimized features: DUP requirements on vector fields are lifted. Limits on the field order are added for vector indexes. More field types are supported by vector indexes. Default values are set for advanced settings of vector indexes to simplify the configuration.

2. Optimized feature: Timestamps in the format of the wall-clock time are supported.

3. Optimized features: Index tables can be switched on the Deployment Management page. All data sources, index tables, and clusters can be displayed on the Deployment Management page.

Release date

Feature

Type

Description

2022-08-16

Feature addition, update, or optimization

New

1. New features: Index data recovery and quick scaling of shards are supported.

2. Optimized feature: The default alerting configuration is added.

Release date

Feature

Type

Description

2022-05-20

Feature addition, update, or optimization

New

1. Optimized features: Quick scale-out of clusters, index schema configuration by clicks, online query configurations, and scenario-based O&M of instances are supported.

2. New features: Vectors can be used as plug-ins for the sorting phases. Multi-value indexes and RANGE indexes are supported.

Release date

Feature

Type

Description

2021-12-31

Official release of OpenSearch Vector Search Edition

New

After years of development, OpenSearch Vector Search Edition can meet the business requirements for high availability, high timeliness, and low costs. OpenSearch Vector Search Edition also provides an automated O&M system on which you can build your own search service based on your business requirements.