Overview
Product introduction
The CleverSee web-retrieval Q&A agent is an end-to-end Q&A service built on web search capabilities, integrating search, reasoning, and generation. It uses intent recognition, query rewriting and enhancement, real-time web retrieval, multi-turn reasoning, and iterative "think-and-search" to generate final answers with source references. It also returns rich media structured cards for specific categories, such as weather, stock prices, hot topics, sports news, and nearby points of interest (POIs). This service is ideal for applications such as AI Q&A assistants, in-vehicle smart conversation, and mobile AI assistants.
The service is available in the following versions:
Lite: Delivers high-quality answers quickly. It is optimized for performance without compromising on quality, making it ideal for most smart conversation scenarios.
Pro (Coming soon): Designed for complex, multi-hop research tasks. It provides an Agent Loop for web retrieval, allowing the agent to autonomously search, read web pages, and organize evidence to generate production-grade research reports for complex questions with traceable evidence.
Use cases
Scenario | Description |
AI Q&A assistant | Provides end-to-end web-based Q&A by retrieving real-time information and generating answers with source references. It covers hot topics and real-time external knowledge across news, lifestyle, and encyclopedias, ensuring verifiable responses and supporting conversational extension through follow-up question recommendation. |
AI search box / Direct answer summary | Enhances existing search portals by adding a direct answer summary, upgrading the user experience from "getting links" to "getting answers with sources." |
In-vehicle smart conversation | Provides conversational answers and rich media structured cards for queries related to weather, nearby POIs, routes, lifestyle services, and news. |
Mobile AI assistant | Adds real-time web Q&A capabilities to AI assistants on mobile phones and other terminal devices. It can be used for system-level AI assistants, voice assistants, "minus-one" screen AI search boxes, and visual search queries. |
Internal enterprise Q&A chatbot | Enhances Q&A, market research, and fact-checking tasks that require real-time external information, reducing hallucinations caused by outdated or missing knowledge. |
Agent tool integration | When building a custom agent, you can integrate the web-retrieval Q&A feature as a tool that is invoked only when needed, without altering the primary workflow. |
Feature comparison
Feature | Lite | Pro (Coming soon) |
Search capability | It balances response speed and search quality, making it suitable for most business scenarios. | It performs in-depth content retrieval from high-quality sources for broader coverage and more comprehensive results. |
Advanced reasoning | It uses a standardized reasoning process. | It uses intelligent "think-and-search" reasoning. |
Answer generation capability | It generates rich and detailed answers quickly. | It delivers superior performance for complex and planning-related questions. |
Vertical tools and data sources | News headlines, encyclopedia, trending searches, weather, stocks, and sports. | Integrates a wider range of Alibaba ecosystem data sources, such as e-commerce products, travel and mobility, education, and streaming video. |
Multimodal capability | It supports image-based Q&A. | It supports image-based Q&A, reverse image search, and mixed text-and-image output. |
Try the product
Activate the service: If you are a new user, follow the instructions in Activate a new service to get started.
Try it online: Log on to the product console, go to the PLAYGROUND, and select web-retrieval Q&A to create and configure an application with custom search, tool, and answer generation capabilities.
Manage your quota: You receive an initial free trial quota. You can check your real-time usage and remaining quota on the Overview page.
Features and benefits
Benefit | Description |
End-to-end delivery | The service understands user query intent and combines web search with tool capabilities, so users move from "finding information" to "getting answers." |
High-quality source aggregation | It integrates capabilities like general web search, hot topics, encyclopedia, weather, stocks, local search, and news headlines. Answers include source references. |
Out-of-the-box and easy to configure | You can configure scenario templates, introductory hot questions, persona prompts, search modes, sources, and tools. You can also customize the reply experience with rich media structured cards and follow-up questions. Deploy your application in three steps (configure, debug, and publish) and call it quickly via API. |
Simplified integration | It provides a standardized interface that allows you to access multiple search capabilities through a single API or SDK. Agent-friendly integration methods such as MCP and SKILL.md will also be supported. |
Low latency and streaming direct answers | It supports both Server-Sent Events (SSE) for streaming responses and single JSON responses. The time to first token (TTFT) for streaming is approximately 1.3s, making it suitable for various applications such as AI chat, mobile or in-vehicle assistants, and smart terminals. |
Integration
API and billing: For API definitions and call examples, see Intelligent Conversation API. For pricing information, see the web-retrieval Q&A billing details.
Integration method: The service supports integration via HTTP, and more methods are planned.