Articles
The article details the technical practices of Tencent's AI Data Agent tool, Lumos. Positioned as an AI-powered table-based big data analysis tool, Tomoro aims to address the pain points of non-technical users, such as difficulties in understanding traditional Business Intelligence data and the high barriers to using analysis tools. By combining big data engines, a familiar table-based User Interface, and AI capabilities, it achieves easy-to-use self-service analysis. Lumos employs a multi-agent architecture to handle complex data analysis tasks and resolves consistency issues between Agents through a shared working memory mechanism. To enhance user experience, the article elaborates on how the hierarchical computing framework achieves sub-second query responses for billion-level data and the strategy of combining AI Coding with tool invocation to optimize functions such as charting. Furthermore, Lumos enhances business adaptability by collaboratively building expert Agents through the MCP (Multi-Agent Collaboration Platform), supporting custom advanced tools for businesses. The article also introduces how question recommendation and clarification functions improve the effectiveness of user queries and emphasizes the importance of building a continuously effective evaluation mechanism for the iterative improvement of Agent capabilities. Overall, Lumos is committed to achieving human-machine collaborative data analysis and eliminating analysis obstacles for non-technical users through technological innovation.
The article delves into the challenges and practices of implementing AI Coding tools in enterprise R&D teams. First, by diagnosing the pain points of various roles such as Project Manager (PM), product managers, UI/UE designers, testers, and R&D engineers in the traditional R&D process, it analyzes the potential of AI-powered efficiency improvement. Then, the article proposes a decision-making framework for AI Coding tools and introduces the product features of Tencent's self-developed AI programming tool CodeBuddy, including its four major AI agents: Plan Agent (requirement planning), Design Agent (design production), Coding Agent (coding implementation), and Deploy Agent (delivery deployment), detailing how they cover the entire R&D process. The core practice section highlights a 'document-driven' AI R&D workflow. It emphasizes efficient human-CodeBuddy collaboration through defined Context and standardized Rules documents (e.g., requirements, design, technical solution templates). The article provides specific practical cases in the requirement and solution design stages, demonstrating how AI can assist in generating PRD, UI design drafts, and technical solutions, aiming to provide enterprises with an actionable AI Coding implementation guide to significantly enhance R&D efficiency.
This article details the latest advancements in Tencent's Kuikly cross-platform development framework, announcing the open-source release of its Web version, which adds support for H5 and WeChat Mini-Programs, enabling unified development for Android, iOS, HarmonyOS, H5, and WeChat Mini-Programs. It begins by highlighting Kuikly's widespread use within Tencent as a cross-platform framework built on Kotlin Multiplatform technology, demonstrating its success across various projects. The article then compares Kuikly Web to mainstream frameworks like Flutter and Compose Multiplatform, emphasizing its superior performance in compiled artifact size and page loading speed due to its DOM rendering approach rather than a self-drawing engine. Finally, it elaborates on Kuikly Web's technical architecture, including the Web Container Runtime Abstraction Layer and specific adaptations for H5 and WeChat Mini-Programs, while also outlining future directions for performance optimization and ecosystem expansion.