OpenSearch:Select an OpenSearch edition

Edition

Scenario

Benefit

LLM-Based Conversational Search Edition

Applicable to conversational search scenarios such as intelligent customer service and enterprise knowledge base. Applicable to multiple industries such as e-commerce, content community operations, and education.

• Provides dedicated conversational search service for enterprises. LLM-Based Conversational Search Edition generates conversational search results based on business data, which is more secure and stable.

• Provides built-in large language models (LLMs) to help build an end-to-end conversational search service. You do not need to access additional LLMs.

• Provides abundant built-in natural language processing (NLP) models and algorithm models such as text slicing, text vectorization, and image vectorization models.

• Provides search results in multiple formats such as text, links, and images. This meets personalized conversational search requirements.

• Provides a one-stop process that allows you to build an enterprise-specific conversational search service by importing only business data.

• Provides a fully managed service without the need for deployment or O&M. This reduces development and O&M costs.

Industry Algorithm Edition

Applicable to enterprises and developers who require a one-stop solution for quickly constructing intelligent industry-specific search systems, industry best practices, industry query analysis algorithms, industry sort models, customized tuning models, and reliable search performance. Key industries include e-commerce, gaming, content community operations, and education.

• Provides intelligent search algorithm models and best practices for multiple industries such as e-commerce, content community operations, and gaming.

• Allows you to customize exclusive word segmentation and NLP algorithm models based on your business requirements.

• Offers search guide features such as drop-down suggestions, top searches, and hints, and supports manual intervention.

• Serves as an out-of-the-box service edition which frees you from O&M and allows you to complete the end-to-end configuration of intelligent industry-specific search with a few clicks.

High-performance Search Edition

Applicable to enterprises and developers who require high performance in big data search scenarios, such as searches for orders, coupons, logistics, and insurance policies. High-performance Search Edition provides a one-stop solution for the quick construction of search systems in a cost-efficient manner.

• Allows you to retrieve hundreds of billions of data within milliseconds and provides search solutions for scenarios such as orders, logistics, and insurance policies.

• Supports connection to Alibaba Cloud services with a few clicks, such as ApsaraDB RDS, MaxCompute, and PolarDB. You can connect to self-managed databases by calling API operations or using SDKs.

• Provides multiple search modes such as combined search, powerful full-text search, and fuzzy search with a prefix, a suffix, or an infix.

• Supports scaling with a few clicks, which frees you from O&M on underlying clusters and enables you to flexibly cope with different business scenarios.

Vector Search Edition

Applicable to enterprises and developers who have online vector retrieval requirements for scenarios such as graph or text searches, audio or video searches, NLP vector searches, Q&A, recommendation, and risk monitoring. Vector Search Edition is also suitable for those who are concerned about the performance, effectiveness, and stability of the search service.

• Performs deep optimization for vector retrieval scenarios, so that multiple scenarios such as graph searches, audio and video searches, personalized recommendations, and Q&A are supported.

• Supports 1 billion 128-dimensional vectors or thousands of queries per second (QPS) within milliseconds even when the service is of the minimum specifications.

• Provides multiple built-in vector retrieval algorithm models to support searches based on primary keys, vectors, and namespaces.

• Adopts highly available engineering architecture, supports horizontal scaling of data, ensures data timeliness, and safeguards system reliability for highly concurrent retrieval.

Retrieval Engine Edition

Applicable to enterprises and developers who have large-scale text retrieval scenarios and require a high-performance, cost-effective, and customized search service.

• Provides multiple built-in flexible indexing and search policies to support high-performance retrieval of a large amount of big data.

• Allows you to retrieve hundreds of billions of data within milliseconds and synchronize the updated data within milliseconds.

• Provides multiple built-in vector retrieval algorithm models to support combined search modes such as tag and text inverted indexes and vector indexes.

• Supports a wide range of custom SQL functions, engine plug-ins, and algorithm models, so that you can flexibly customize your business processing logic.

• Allows you to use standardized SQL programming languages and integrated development environments (IDEs) to reduce the cost of engine learning and development.

Feature

Industry Algorithm Edition

High-performance Search Edition

Retrieval Engine Edition

Vector Search Edition

LLM-Based Conversational Search Edition

Data source

ApsaraDB RDS, MaxCompute, PolarDB, and API

ApsaraDB RDS, MaxCompute, PolarDB, and API

Object Storage Service (OSS), MaxCompute, and API

OSS, MaxCompute, and API

MaxCompute and API

Table schema

Offline LEFT JOIN operation on multiple tables

Online JOIN operation on multiple tables, which is supported only for storage-optimized exclusive clusters

Online JOIN operation on multiple tables

Independent tables

Offline JOIN operation on built-in fixed table schema and custom table schema

Data update

Data updates by calling API operations or using the SDK, and automatic data synchronization by using change tracking Data Transmission Service (DTS) instances

Data updates by calling API operations or using the SDK, or automatic data synchronization by using change tracking DTS instances

Data updates by calling API operations or using the SDK

Data updates by calling API operations or using the SDK

Data updates by calling API operations or using the SDK

Data backup

If an application has no data source, the application can inherit data from the online application.

Not supported

API data source, which supports data restoration from historical indexes.

API data source, which supports data restoration from historical indexes.

Not supported

Index type

Text index and vector index

Text index

Text index and vector index

Vector index

Text index and vector index

Analyzer

Industry-specific analyzer, multi-language analyzer, general-purpose analyzer, and lightweight custom analyzer

Multi-language analyzer and general-purpose analyzer

Industry-specific analyzer, multi-language analyzer, and general-purpose analyzer

Primary key and vector indexes

Text and vector analyzers

Intervention in text analysis

Supported

Supported

Supported

Not supported

Not supported

Semantic understanding

Stop word filtering, spelling correction, term weight analysis, synonym configuration, named entity recognition (NER), and custom query analysis

Not supported

Not supported

Not supported

Not supported

Sort method

Sort expressions, sort plug-ins (Cava scripts), and sort algorithms (CTR models and category prediction models)

Sort expressions and sort plug-ins (Cava scripts)

Sort expressions

Sort by vector proximity

Sort expressions

Search guidance: drop-down suggestions and hints

Supported

Not supported

Not supported

Not supported

Not supported

SQL syntax

Not supported

Supported only for storage-optimized exclusive clusters

Supported

Not supported

Not supported

A/B test

Supported

Not supported

Not supported

Not supported

Not supported

Business report

Supported

Supported

Not supported

Not supported

Not supported

Monitoring and alerting

Supported

Supported

Supported

Supported

Not supported

Slow query optimization

Supported

Supported

Not supported

Not supported

Not supported

Automation level

O&M-free

O&M-free

Self-service O&M based on the provided monitoring metrics

Self-service O&M based on the provided monitoring metrics

O&M-free

Region

Industry Algorithm Edition

High-performance Search Edition

Vector Search Edition

Retrieval Engine Edition

LLM-Based Conversational Search Edition

China (Hangzhou)

✓

✓

✓

✓

China (Shanghai)

✓

✓

✓

✓

✓

China (Qingdao)

✓

✓

China (Beijing)

✓

✓

✓

✓

China (Zhangjiakou)

✓

✓

✓

✓

China (Shenzhen)

✓

✓

✓

✓

China (Hong Kong)

✓

Singapore

✓

✓

✓

✓

✓

Germany (Frankfurt)

✓

✓

✓

US (Virginia)

✓

✓