WeKnora
Tencent/WeKnoraReviewed by AI
Checked by RepoPicks Radar on
What is WeKnora?
WeKnora, from Tencent, is a knowledge platform you run on your own machine or server. You upload documents (PDF, Word, Excel, images and more), and it answers questions about them with links to the passages it used. It can also act as an agent that does multi-step tasks and build an automatic wiki from your files. You start it with Docker Compose and use it in the browser. The AI models come from outside: an online provider with a key, or Ollama running on your own computer.
WeKnora is free and open source (Apache-2.0) and can replace NotebookLM, Chatbase, and Notion AI.
How easy is it to install WeKnora?
D · terminal, several tools or a specific graphics card · Windows
- Windows D
- macOS D
- Linux D
On all three systems the published way to install is the same: clone the repository, copy and edit the .env file, then run docker compose pull and up, which starts several services (app, web interface, PostgreSQL, Redis and a document reader). There is no ready installer; the README says the desktop app has no installer yet and must be built from source.
Coming soon: install in 1 click with the RepoPicks app.
WeKnora on GitHub
- Stars
- 37,106on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +721stars in 7 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Docker
- To build from source
- Docker, Go
- Disk space
- not determined
- Accounts or keys
- LLM provider (OpenAI, DeepSeek, Qwen, Gemini and 25 others), Web search provider (Tavily, Bing, Google and others)
- Cost
- Free · The software is free. The AI models are not included: you either pay an online provider per use or run a local model with Ollama.
- License
- Apache-2.0 (Permissive)
- Can replace
- NotebookLM, Chatbase, Notion AI
- Last change on GitHub
- Nov 9, 2026
WeKnora risk signals (5)
- Attentionowner warns not to expose publicly
README warning block: "deploy it in an internal / private network rather than on the public internet; avoid exposing the service directly to public networks, to prevent information leakage".
AI review
- Attentionoptional docker socket equals host root
docker-compose.yml comment next to the commented-out mount of /var/run/docker.sock: "Docker 沙箱默认关闭... 这等同宿主机 root" (the Docker sandbox is off by default; turning it on and mounting the socket is the same as root on the host).
AI review
- Attentionruns third party skills from the internet
README: "Skills installed from ClawHub / SkillHub / Git / ZIP run in session-persistent Docker / E2B / Cube sandboxes", and the agent can drive your own Chrome or Edge through the BrowserSkill extension.
AI review
- Infolicense text disagrees with detection
LICENSE opens with "WeKnora is licensed under the MIT except for the third-party components listed below" and the README says MIT, while the same file also embeds the full Apache-2.0 text for a bundled component, which is what the automatic check read.
AI review
- Infoexample file ships default passwords
.env.example ships working defaults such as DB_PASSWORD=postgres123!@#, REDIS_PASSWORD=redis123!@# and a default SearXNG secret marked "rotate-before-exposing-publicly".
AI review
Evidence is shown in its original language.
What Radar could not check about WeKnora
- How much memory and disk the full Docker stack really needs once PostgreSQL, Redis and the document reader are running
- Whether answers are correct for your own documents, and how good the automatic wiki and knowledge graph are
- The real monthly cost if you point it at a paid model provider instead of Ollama
- Whether the Lite single binary and the unpublished desktop app work as described
- How much of the documentation site is usable in English, since the full docs are described as being in Chinese