What Is KooSearch?

Overview

KooSearch offers an enterprise-grade, out-of-the-box retrieval augmented generation (RAG) service. By integrating enterprise knowledge bases consisting of both structure and unstructured data, an industry-leading search model, a high-performance CSS vector database, and a variety of large language models (LLMs), KooSearch allows you to quickly build your own AI search and document Q&A applications. As a promising way to mitigate LLM hallucination, RAG helps LLMs generate more accurate, reliable, and secure answers.

An Elasticsearch-based vector database hosted by Cloud Search Service (CSS) powers a vector search engine. CSS is a fully managed, distributed search service based on open-source Elasticsearch and OpenSearch. You can use it for structured and unstructured data search, and enable vector-based composite search, statistical analysis, and reporting. It provides vector search capabilities for KooSearch.

KooSearch is available only in CN-Hong Kong and AP-Singapore. KooSearch is in the open beta test (OBT) phase. To trial-use it, submit a service ticket.

Highlights

• Out-of-the-box Q&A service

  KooSearch provides an enterprise-grade RAG service that:

  • Supports a variety of document formats, including .doc, .docx, .pdf, .pptx, .ppt, .xlsx, .xls, .csv, .wps, .png, .jpg, .jpeg, .bmp, .gif, .tiff, .tif, .webp, .pcx, .ico, .psd, .dps, .et, .txt, .ofd, and .md.
  • Provides document parsing and splitting services and enhanced OCR for PDFs, scanned copies, images, and tables.
  • Supports multiple splitting methods: automatic splitting, hierarchical splitting, length-based splitting, and custom rules-based splitting.
  • Supports full-text search, vector search, and hybrid search.
  • Supports Q&A on structured documents, such as FAQs.
  • Supports tag and directory management.
• AI search

  An enhanced web search service gives LLMs access to up-to-date information available on the Internet. On top of this, AI search integrates LLMs and search planning to deliver a smart search experience.

• Flexible configuration
  • On KooSearch's model management page, you can choose any LLM that complies with OpenAI standards to provide the Q&A service.
  • KooSearch provides a prompt management function, where you can manage frequently used prompts for Q&A.

Product Architecture

Figure 1 KooSearch architecture

KooSearch supports practical use cases such as knowledge Q&A and AI search.

• KooSearch supports local knowledge bases. After intelligent parsing and splitting, local documents are vectorized and stored in the CSS vector database. The top-K documents are retrieved and ranked, and returned to an LLM, which then summarizes these results to generate answers.
• KooSearch supports AI search. It integrates an enhanced web search service, coupled with search planning, giving LLMs access to up-to-date information on the Internet and enhancing their performance.

Access Methods

KooSearch provides HTTPS-based APIs as well as a web-based console, which you can use to easily access the service.

• Using APIs

  If you want to integrate KooSearch into a third-party system for secondary development, use the APIs to access KooSearch.

• Web-based console

  For visualized operations, use the KooSearch web console.

  If you have already registered on the public cloud platform, log in to the cloud service management console and click Cloud Search Service.

  If you have not registered, Signing Up for a HUAWEI ID and Enabling Huawei Cloud Services.