DocsICP FilingConsole
Documentation
Home Page Information Query Service Use Cases Vertical query Domain search - Overview

CleverSee - Domain search overview

Updated at:
Copy as MD
Product
My Favorites

1. Background

Domain search is part of the Domain Search and Question Answering section of CleverSee. It is a domain retrieval service for structured queries in specific domains. Unlike general web searches that return a list of web pages, domain search acts as an independent and composable retrieval tool that an agent can call on demand. It directly returns structured results from within the domain, such as POI information, conclusive medical answers, and health knowledge entities and their attributes. These results can be directly consumed by large language models without requiring web page parsing or secondary processing. Currently, three capabilities are available: nearby search, medical Q&A, and health knowledge graph.

Core advantages of domain search:

  • Structured, direct results: Directly returns structured domain fields, such as place names, types, administrative divisions, addresses, distances, business information, medical conclusions with their sources, and entity-attribute pairs. This eliminates the need for client-side web scraping or information extraction.

  • Authoritative sources and domain-specific standards: Built on authoritative data from specific domains. In scenarios that demand high expertise, such as local services and healthcare, its accuracy and completeness are significantly better than solutions that rely on reformulating general search queries.

  • Low latency for online integration: Each API has a server-side timeout of 5 seconds, making it suitable for direct integration into online conversational flows and agent orchestration.

Product advantages

Compared to a do-it-yourself approach that combines general search with query rewriting to assemble domain-specific results, domain search offers distinct advantages:

  • Zero rewriting, zero extraction: The interface directly accepts natural language queries (such as "high-end hotels within 5 km of Lingyin Temple in Hangzhou", "what to do for a stomach ache", or "Amoxicillin capsules"). The service handles intent recognition and field assembly, so you do not need to maintain rewriting prompts or parsing rules on the client side.

  • 按能力独立计量、成本可预估： 各能力按次计量、不绑定特定大模型或智能体框架，可与任意 LLM / Agent 编排方案自由组合，避免检索能力与生成链路强耦合带来的额外 Token 消耗。

  • Source traceability: Healthcare APIs return both conclusions and their evidence sources (title, link, abstract, and source type), which can be directly used for citations and professional validation.

2. API overview

Nearby Query API --- CommonQueryByScene

CommonQueryByScene is a natural language interface for nearby searches and supports searching all POI types. The publicly available enhanced version is POIAdvanced.

Key features:

  • Natural language input: The query parameter accepts full conversational descriptions. The service automatically identifies locations, ranges, and filtering criteria.

  • Optional scene constraint: The querySceneEnumCode parameter supports hotels, attractions, restaurants, and entertainment. If you omit this parameter, the model intelligently identifies the scene.

  • Comprehensive structured fields: Returns the name, type and type code, province/city/district and their codes, detailed address, distance from the center point, and business information such as phone number, business district, rating, opening hours, and average cost.

  • Controllable response size and mode: The limit parameter controls the number of returned items (default is 10). The searchModel parameter supports single (condensed fields) and normal (full fields, default).

For the complete API definition, including request and response parameters, JSON examples, HTTP/SDK call methods, and error codes, see the documentation for the Nearby Query API.

Medical Q&A API --- MedicalAnswer

MedicalAnswer provides conclusive professional answers to a wide range of colloquial health questions (such as "What causes motion sickness?", "What should I do for a stomach ache?", and "Is sugar-free cola good for you?"). The answers are generated by retrieving information from authoritative medical data and using a large language model to summarize and refine it.

Key features:

  • Conclusion and sources in one response: The messages[].content field contains the refined answer, where [n] is a citation marker. The raDocList provides the corresponding source title, link, and summary.

  • Identifiable source types: Sources are categorized by type, such as research papers, clinical guidelines, medical webpages, and medical knowledge bases. This allows you to weight and display information based on its professional level.

  • Minimal input: Only requires a query (1 to 200 characters; a length of 30 characters or less is recommended).

Health Knowledge Graph API --- MedicalKnow

MedicalKnow is designed for medical attribute and category queries, such as "which is the best dental hospital in Hangzhou," "calories in an apple," or "aspirin." It returns entities and their attributes as structured text in an "Entity + Attribute" format.

Key features:

  • Structured entity-attribute data: The content is returned in the format "Entity: X - Attribute[Attribute Name]: Attribute Value", which you can directly parse into key-value pairs.

  • Specific query types: The type parameter supports finding (symptoms), food, and medicine.

  • Traceable sources: The messages[].url field returns a link to the content source.

Note

  1. The Medical Q&A API and the Health Knowledge Graph API are two independent APIs, each with its own request path, parameters, and billing metrics.

  2. The server-side timeout for both APIs is 5 seconds.

  3. For complete API definitions, including response parameters, JSON examples, HTTP/SDK call methods, and error codes, see the documentation for the Medical Q&A API and Health Knowledge Graph API.

3. Use cases

  • Local services agents: Query and recommend nearby places such as stores, attractions, restaurants, and entertainment venues. You can directly use fields like ratings, opening hours, and average cost to sort results and render information cards.

  • Healthcare assistants: Provide conclusive answers and authoritative sources for symptom inquiries and medication questions. Use attribute queries for diseases, drugs, and foods to generate structured educational cards.

  • Enterprise Knowledge Assistant supplements external facts: In addition to internal enterprise knowledge, the assistant supplements real-time external facts in the location and health domains on demand. This prevents the large language model from generating responses based on its stored knowledge.

4. Try it out

In the CleverSee console, you can directly experience the domain search capability. Enter a natural language query to view the returned structured results.

5. Product integration

For API definitions and call examples, see the documentation for the Nearby Query API, Medical Q&A API, and Health Knowledge Graph API. For billing information, see Domain search - Billing.

For integration methods (HTTP / SDK / MCP / Skill), refer to the "Integration Methods" chapter.

Previous：Vertical queryNext：Developer reference