XNews

WeVisDoc: WeChat's document parser

WeVisDoc: WeChat's document parser

Tencent's WeChat team open-sourced WeKnora, an enterprise-grade LLM knowledge platform that transforms scattered documents into queryable RAG knowledge bases, ReAct agents, and self-evolving wikis. It supports private deployment, multi-tenancy, and role-based access control, and is the core technology behind WeChat's dialogue platform.

Key insights

WeKnora consolidates enterprise knowledge management

WeKnora, open-sourced by Tencent's WeChat team, addresses fragmented enterprise knowledge by converting documents from various sources into structured knowledge assets. It offers RAG-based Q&A, ReAct agents for complex tasks, and a Wiki mode that distills documents into linked Markdown knowledge bases. The platform supports private deployment and includes features like multi-tenancy, RBAC, and audit logs.

Advanced security and engineering features

WeKnora includes robust security measures such as AES-256-GCM encryption, SSRF-hardened HTTP clients, and sandbox isolation. It offers a four-tier RBAC model, API keys with scoped permissions, and a 30-day audit log. The platform treats knowledge bases like code repositories, with versioned diffs, rollback, and automatic re-indexing.

Source posts