Overview
OpenSearch is an all-in-one platform for developing commercial intelligent search services, built on a large-scale distributed search engine developed by Alibaba. It currently provides mid-end services for the core search business of Alibaba Group, including Taobao, Tmall, and Cainiao. Drawing on years of industry search experience and the traffic peaks of Double 11, OpenSearch has refined a full suite of high-performance, high-timeliness, highly available, and highly stable search services. It includes six editions to meet the search needs of a wide range of industries: , LLM-based Conversational Search Edition, Industry Algorithm Edition, High-performance Retrieval Edition, Vector Retrieval Edition, and Retrieval Engine Edition. As a platform-based service, OpenSearch makes professional search technology simple, low-barrier, and cost-effective, so that search is no longer a business bottleneck. You can implement product search features at low cost and iterate quickly.
To build a search service with OpenSearch, you only need to:
Create an OpenSearch instance.
Edit and configure the instance.
Synchronize or upload documents.
Submit search requests from your app or web client, access OpenSearch via API, and get search results.
Product architecture
LLM-based Conversational Search Edition
A SaaS platform where you can interact with the system through the console or API.
Built on LLM models and industry-leading NLP models, it lets you quickly build an enterprise-specific large model for Q&A search by simply importing your business data.
For Q&A search, it intelligently generates answers based on the business data returned by the search, ensuring that the results are secure, stable, high-quality, and reliable.
Industry Algorithm Edition
A SaaS platform where you can interact with the system through the console or API.
Scenario-based configuration: You only need to create an application instance, configure a data source, and configure the column structure and search attributes. After the index is rebuilt, you can run search tests through the SDK or console.
Industry templates: To help you quickly build industry-specific search services, the Industry Algorithm Edition offers many industry algorithm packages for industries such as e-commerce, content communities, gaming, and education. After you select an industry vertical, you can create search-related features and services in one stop, including query semantic understanding, ranking, and search guidance. This enables faster integration while ensuring intelligent search effectiveness for your industry.
High-performance Retrieval Edition
A SaaS platform where you can interact with the system through the console or API.
Scenario-based configuration: You only need to create an application instance, configure a data source, and configure the column structure and search attributes. After the index is rebuilt, you can run search tests through the SDK or console.
Big data retrieval: Compared with the Industry Algorithm Edition, the High-performance Retrieval Edition removes the heavy industry algorithm capabilities. Based on general search capabilities (analyzers and ranking), it focuses on the throughput of business queries and writes, providing you with second-level response and real-time query capabilities for big dataset retrieval scenarios.
Vector Retrieval Edition
A PaaS platform where you can operate and maintain Vector Retrieval Edition instances through a visual management console.
Scenario-based configuration: You only need to create an application instance, configure a cluster, configure a data source, and set up index tables. After the index is rebuilt, you can run search tests through the SDK or console.
Vector retrieval: The Vector Retrieval Edition provides rich query parameter controls and a hybrid search method that combines inverted indexes and vectors. It offers highly reliable, high-performance support for scenarios such as image search, audio/video search, recommendations, and Q&A.
Retrieval Engine Edition
A PaaS platform where you can operate and maintain Retrieval Engine Edition instances through a visual management console.
Scenario-based configuration: You only need to create an application instance, configure a cluster, configure a data source, and set up index tables. After the index is rebuilt, you can run search tests through the SDK or console.
SQL query: To reduce your learning curve, the Retrieval Engine Edition supports SQL query syntax.
Low cost: To reduce business costs, the Retrieval Engine Edition includes multiple built-in index compression strategies, meeting your query needs at a lower cost.