WeKnora
LLM-powered AI agent framework for deep document understanding, semantic retrieval, and context-aware answers using a RAG AI system. Latest release v0.2.1 (December 8, 2025) brings Qdrant vector search, auto retriever configuration, and steadier Docker profiles.
What is WeKnora?
WeKnora is Tencent's open-source framework for document understanding, hybrid retrieval, and grounded question answering. It pairs RAG pipelines with tool-using agents so teams can index PDFs, Word, Markdown, and HTML and answer questions with citations.
Key Features
Deep Document Understanding
Layout-aware parsing and chunking for PDFs, Word, Markdown, and HTML with semantic structure preserved.
Hybrid Retrieval (Qdrant)
Vector + keyword + graph search with native Qdrant collections and automatic retriever selection.
Agent Mode (ReACT)
Grounded agents that call knowledge bases, MCP tools, and web search for multi-step answers.
Knowledge Graphs
Entity and section linking to visualize relationships across a corpus.
Multi-tenant Security
Tenant isolation, shared model catalog, and access controls for enterprise deployments.
Observability
Tracing and logging hooks for retrieval chains, migrations, and agent tool calls.
What's new in v0.2.1
- Qdrant vector database support with hybrid vector/keyword search and automatic collection sizing.
- Retriever engines auto-resolve from the
RETRIEVE_DRIVERenvironment variable for quicker setup. - Docker Compose profiles for optional services (minio, qdrant, neo4j, jaeger, full) with Qdrant 1.16.2 pinned.
- Safer migrations with dirty-state recovery plus Neo4j retry and backoff for stability.
- Fixes for Chinese keyword search and stricter image URL validation in the UI.
Quick Start
Get WeKnora up and running in minutes with our simple installation process:
For detailed installation instructions and configuration, visit our Getting Started Guide.
Use Cases
Enterprise Knowledge Base
Build intelligent knowledge bases for your organization's documentation, policies, and procedures.
See example ?Customer Support Chatbots
Create AI-powered support agents that can answer customer questions using your product documentation.
See example ?Research & Documentation
Transform research papers and technical documentation into searchable, queryable knowledge systems.
See example ?WeChat Integration
Deploy intelligent Q&A services within the WeChat ecosystem through the WeChat Dialog Open Platform.
See example ?AI Research Assistant
An AI-powered research tool that processes documents and delivers accurate insights using RAG AI systems.
See example ?Enterprise Knowledge Base AI
Transform company data into an intelligent AI knowledge base for faster decision-making and search.
See example ?AI Customer Support Automation
Automate responses using AI trained on your internal data and documents for reliable customer support.
See example ?Why Choose WeKnora?
- Open Source: Fully open-source under MIT License, free to use and modify
- Production Ready: Battle-tested framework used by Tencent in production environments
- Comprehensive: Complete solution from document parsing to intelligent Q&A
- Extensible: Plugin architecture and MCP server support for custom integrations
- Modern Stack: Built with Go, Vue.js, and modern AI technologies
- Active Community: Growing community of contributors and users
Resources
Documentation & Guides
- Complete Documentation - Comprehensive guides and references
- Getting Started Guide - Step-by-step setup instructions
- API Reference - Complete API documentation
- Examples & Tutorials - Real-world use cases and code samples
Community & Support
- Community Guidelines - How to contribute and participate
- Blog & Articles - Latest news, updates, and technical articles
- Contact Us - Get help and connect with the team
- GitHub Repository - Source code and issue tracking
Get Involved
WeKnora is an open-source project and we welcome contributions from the community. Whether you're fixing bugs, adding features, improving documentation, or sharing use cases, your contributions help make WeKnora better for everyone.