Tencent WeKnora: An Open-Source LLM RAG Knowledge Platform

TL;DR: Tencent/WeKnora is an open-source LLM knowledge platform from Tencent that turns raw documents into a queryable RAG store, an autonomous reasoning agent, and a self-maintaining Wiki. Go codebase, 25,380 stars.

What it is

Tencent/WeKnora describes itself as “Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.” The repo lives under the Tencent GitHub organisation, the language is Go, the license is NOASSERTION (GitHub could not auto-detect an SPDX identifier), and the repository has 25,380 stars and 3,482 forks. It is not archived.

Topics include agent, agentic, ai, chatbot, dsh-plugin, embeddings, evaluation, generative-ai, golang, knowledge-base, llm, multi-tenant, ollama, openai, question-answering, rag, reranking, semantic-search, vector-search, and wiki. The homepage is https://weknora.weixin.qq.com.

What it does well

WeKnora bundles three workflows that RAG projects often piece together by hand.

• Document ingestion. Raw documents go in and a queryable RAG store comes out — chunking, embedding, and indexing are handled inside the platform.
• Reasoning agent. Beyond retrieval, the platform can drive an agent that uses the knowledge base as a tool during multi-step reasoning.
• Self-maintaining Wiki. New content updates itself over time, which is the part most RAG projects struggle with once the underlying data starts drifting.
• Multi-tenant. Topics include multi-tenant support, so the platform can serve several teams or customers from a single deployment.
• Model pluggable. Both Ollama and OpenAI are listed in the topics, so the same platform works against local and hosted models.

How to try it

The Go install path follows the standard Go toolchain:

git clone https://github.com/Tencent/WeKnora.git
cd WeKnora
go build ./...

The docs on the homepage walk through configuring the document store, picking a model backend, and starting the agent service.

When it fits, and when it doesn’t

WeKnora fits teams that need a multi-tenant RAG + agent + Wiki stack in a single Go-deployable service, especially when they prefer local models. It is less attractive if your RAG use case is narrow enough that a single-purpose vector database plus a thin agent wrapper is enough, or if your stack is firmly Python-based and you would rather stay there.

25k stars and the topic list including both Ollama and OpenAI suggest the project is being kept current with the model ecosystem on both hosted and local fronts.

Source and Accuracy Notes

• Project page: https://github.com/Tencent/WeKnora
• Source repository: Tencent/WeKnora (verified via GitHub API html_url)
• License: NOASSERTION (GitHub could not auto-detect an SPDX identifier)
• Stars at time of writing: 25,380
• Language: Go
• Homepage: https://weknora.weixin.qq.com
• Topics: agent, agentic, ai, chatbot, dsh-plugin, embeddings, evaluation, generative-ai, golang, knowledge-base, llm, multi-tenant, ollama, openai, question-answering, rag, reranking, semantic-search, vector-search, wiki