Tencent/WeKnora

WeKnora – AI‑powered knowledge management platform

What it is

• An open‑source framework built by Tencent that lets teams store, search, and reason over their documents with large language models (LLMs). It combines three main modes:
  1. RAG (Retrieval‑Augmented Generation) – hybrid text‑plus‑vector search with citations and multimodal parsing.
  2. Agent – a multi‑step AI assistant that can call tools, run sandboxed code, and even control a browser.
  3. Wiki – a collaborative knowledge‑base with a graph view, versioning, and rollback.

Key capabilities

• Multi‑source ingestion – sync from Feishu, Confluence, GitLab, Notion, DingTalk, RSS, etc.; supports >10 file types (PDF, Word, Excel, images, XMind, …) via the built‑in anydoc parser.
• Toolbox for agents – install skills from ClawHub/SkillHub or custom ZIP/Git repos; run them in Docker/E2B/Cube sandboxes or a local “Lite” host; browser automation via the open‑source BrowserSkill extension.
• Long‑term memory & curation – per‑user profile, preferences, confirmed facts, and editable retrieval chunks with diff/rollback.
• Rich integrations – Q&A bots for WeCom, Feishu, Slack, Telegram, embed widget for websites, MCP server for inter‑tool communication, scoped API keys.
• Model‑agnostic – 29 built‑in vendor connectors (OpenAI, Gemini, Ollama, DeepSeek, Qwen, etc.) and a generated model catalog; supports multimodal embedding models for image‑text retrieval.
• Enterprise‑grade security – RBAC with four roles, per‑resource ownership, audit logs, encrypted credentials, OIDC/JWKS verification.
• Pluggable infrastructure – swap vector stores (pgvector, Milvus, OpenSearch, Pinecone, etc.), storage back‑ends (MinIO, S3, Tencent IMA), and run via Docker Compose, Helm/Kubernetes, or a single‑binary “Lite” mode.

Typical use cases

• Internal Q&A bot that answers employee questions using up‑to‑date company documents.
• Automated research assistant that gathers information, runs code, and writes reports.
• Knowledge‑base wiki that stays synchronized with source repositories and visualizes relationships as a graph.
• Customer‑support tool that retrieves relevant policy documents and suggests actions.

Quick start (Docker Compose)

git clone https://github.com/Tencent/WeKnora.git
cd WeKnora
cp .env.example .env   # edit as needed
docker compose pull      # fetch images
docker compose up -d    # start services

• UI: http://localhost
• API: http://localhost:8080
• Tracing (Langfuse): http://localhost:3000

Installation options

• Docker Compose – default all‑in‑one setup.
• Kubernetes (Helm chart) – production clusters.
• Lite binary – single‑process SQLite + in‑memory queue for low‑resource environments.
• Desktop app – GUI wrapper around the Lite runtime (macOS only, build from source).

Documentation & community

• Full docs: https://weknora.weixin.qq.com/docs/
• Release notes, upgrade guides, and a demo video are linked in the README.
• Integrations listed on the site include a WeChat Dialog platform, Chrome extension, and npm package @wxg-prc-cpg/dsh-weknora.

License

• MIT License (per the repository badge).

Bottom line WeKnora is a production‑ready, extensible platform for building LLM‑driven enterprise knowledge systems, offering out‑of‑the‑box RAG, autonomous agents, and a wiki‑style UI while letting you plug in your own models, storage, and tools.