WeKnora

Overview

WeKnora is an open-source, LLM-powered knowledge framework for enterprise document understanding, semantic retrieval and reasoning. It brings a team's documents together so they can be searched, reasoned over and kept up to date.

https://github.com/user-attachments/assets/de0e490a-e738-405e-bb3a-64b63edd7278

Use RAG to look things up, the agent for multi-step tasks, and the wiki to organize knowledge. All three work on the same knowledge bases.

The agent's toolbox. Skills installed from ClawHub / SkillHub / Git / ZIP run in session-persistent Docker / E2B / Cube sandboxes, with an interactive terminal and graphical desktop beside the chat. Through the BrowserSkill extension the agent operates the user's own Chrome or Edge, and external MCP services (OAuth included) can be connected and enabled tool by tool.

Beyond the three modes:

• Memory and curation: cross-session long-term memory keeps the profile, preferences and facts a user has confirmed. Folder uploads keep their directory tree, and retrieval chunks can be edited, diffed and rolled back.
• Data sources and formats: auto-sync from Feishu wiki / Feishu Drive / Confluence / GitLab / Tencent IMA / Notion / Yuque / DingTalk Docs / Seafile / RSS, with more on the way. 10+ document formats including PDF, Word, images, Excel and XMind; Office files are parsed in-process by anydoc.
• Channels and integrations: Q&A in WeCom, Feishu, Slack, Telegram and other IM apps; an embed widget for external websites; a built-in MCP Server for Cursor, Claude and other AI tools; scoped API keys with a principal model for programmatic access.
• Models: 29 built-in vendors with a generated model catalog, including OpenAI, DeepSeek, Qwen (Alibaba Cloud), Zhipu, Hunyuan, Gemini, MiniMax, NVIDIA, LiteLLM and Ollama.
• Permissions and operations: multi-workspace RBAC (four roles, per-resource ownership, per-workspace audit log), several storage instances per workspace, a runtime task-queue dashboard with worker-pool governance, and Langfuse tracing for agent steps, token usage and pipelines.
• Deployment: LLMs, vector databases and storage backends are all swappable. Deploy locally or on a private cloud and keep the data in your own environment.

Quick Start

Run with Docker Compose

Requires Docker, Docker Compose and Git.

git clone https://github.com/Tencent/WeKnora.git
cd WeKnora
cp .env.example .env    # Edit .env as needed, see comments in the file
docker compose pull     # Pull the latest images
docker compose up -d    # Start core services

Then open http://localhost and follow the onboarding guide. A walkthrough with sample data is in the Quickstart.

[!TIP] To use a local Ollama model, run ollama serve > /dev/null 2>&1 & first. For the Ollama embedding model name, OLLAMA_BASE_URL, and RAM notes, see Configuration.

Service	URL
Web UI	http://localhost
Backend API	http://localhost:8080
Langfuse Tracing	http://localhost:3000

Optional services

Add --profile flags to enable additional components; multiple profiles can be combined.

Profile	Adds
(default)	Core services
full	All features
neo4j	Knowledge Graph (Neo4j)
minio	Object Storage (MinIO)
langfuse	Tracing (Langfuse)

docker compose --profile neo4j --profile minio pull
docker compose --profile neo4j --profile minio up -d
docker compose down     # Stop services

Upgrading

If you already have WeKnora running and downloaded a newer release:

# Set WEKNORA_VERSION in .env to the target release (e.g. 0.9.0), or keep latest
docker compose pull     # Pull images matching WEKNORA_VERSION
docker compose up -d    # Recreate containers with new images

[!NOTE] docker compose up -d alone reuses locally cached images and may leave the UI version out of sync with the release you downloaded. Read the upgrade notes before upgrading. Deployments older than v0.8.2 should also read the v0.8.2 notes.

Other ways to deploy

Option	When to use it
Docker Compose	The standard deployment above: all features, multiple services
Kubernetes (Helm)	Production clusters; the chart is in helm/
Lite single binary	Local or low-resource use with no external dependencies (SQLite + in-memory queue); see Lite vs. standard
Desktop app	The Lite runtime with a GUI, login-free start and a macOS host sandbox; no installer is published yet, so build it from source

All options, hardware requirements and deployment topologies: Installation guide.

[!WARNING] WeKnora ships with login authentication, but for production deployments we strongly recommend that you:

• 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;
• configure proper firewall rules and access controls for the deployment environment;
• regularly update to the latest version for security patches and improvements.

What's New

v0.9.0 · 2026-10-29 · release notes

• WeKnora assistant: Ask about WeKnora releases, features and configuration, then work from the current page: create agents, schedule tasks and manage knowledge, acting as you within your permissions.
• Multi-agent collaboration: Delegate independent subtasks, follow progress and bring the results together.
• Automations: Run one-off or recurring tasks with execution history, results and approval notifications.
• Execution permissions: Read-only, confirm-before-change and automatic-write modes, with scoped session and automation grants.
• Sandboxes and background work: Optional per-user sandbox reuse, managed background shell jobs and skill rollout controls.
• Parsing, sync and retrieval: Configurable image pipelines, OCR preprocessing, deeper Feishu/Lark sync, Confluence PAT and retrieval improvements.
• Upgrade notes: Review database migrations, Feishu resync, the sandbox allocation gate and configuration changes.

v0.8.3 · 2026-10-14 · release notes

• Image gallery: browse document images and filter by source, caption, OCR and image attributes.
• Multimodal Embedding: support for multimodal embedding models that encode text and images into a shared vector space, including Tencent WeMM-Embedding and Qwen3-VL-Embedding, with opt-in image-vector retrieval and compatible image-aware reranking.
• Wiki graph: clearer group layouts and theme captions, with group focus, switching and return to the full graph.
• Source citations: navigate from an answer to the cited passage in the original document.
• Chat and tools: on-demand attachment reading without a sandbox, unread replies, @ mention search, Docker terminals, Lite host skills and MCP interoperability updates.
• Administration and integrations: platform user management, invitation-only signup mode, administrator model catalog, WeMM embeddings, TEI/Pinecone/Cohere reranking, Seafile and Traditional Chinese.
• Reliability: broad retrieval, parsing, datasource, Wiki, OAuth and CLI fixes. Review upgrade requirements for registration modes, empty tool lists, database migrations and reindexing.

v0.8.2 · 2026-09-24 · release notes

Agents can operate the browser on your computer, knowledge bases can be published to other AI tools over MCP, and a running conversation can be steered, forked or rewound.

• Local Browser (BrowserSkill): agents drive the user's own Chrome / Edge through the open-source BrowserSkill extension, with a live task preview, pause / resume and hand-off for logins and CAPTCHAs.
• Built-in MCP Server: per-workspace /mcp/<endpoint_id> endpoints over Streamable HTTP, each with its own token, knowledge-base scope, rate limit and tool groups. The Python mcp-server/ is deprecated.
• Conversation control: append requirements to a running turn, fork from any earlier question, rewind in place with sandbox checkpoints, and pick reasoning effort per session. Generated files are collected in a new artifacts library.
• Sandbox: interactive terminal and graphical desktop; macOS Lite host sandbox with project folders; a sidebar Toolbox for skills, MCP services and the browser connection.
• Models: rebuilt model catalog (27 built-in vendors with generated context-window, max-output, reasoning and vision metadata). Agent retrieval tools are consolidated into search_knowledge / read_document / list_documents.
• Knowledge and platform: Confluence and DingTalk Docs data sources; Bocha and Serply web search; Japanese UI; per-channel IM reply language; whitelist-only outbound mode.

[!IMPORTANT] Breaking: DingTalk channels are Stream-only, and sandbox commands run as root. See the upgrade notes.

v0.8.0 · release notes

• Skill sandbox runtime: session-persistent Docker / E2B / Cube backends with per-tenant network policy; the Local host-process backend is removed and Docker is opt-in.
• Tenant skill catalog: install from ClawHub / SkillHub / git / zip, with per-sandbox snapshots, live progress, file browse/edit, and personal and workspace env vars.
• Cross-session long-term memory: profile / preference / fact / task / interest, auto-extracted with user confirmation, plus search_memory.
• Parsing and sources: in-process anydoc office parser; GitLab and Tencent IMA data sources; XMind parsing.
• Ecosystem: official DeepSeek Harness plugin @wxg-prc-cpg/dsh-weknora; LiteLLM; Exa and Metaso web search.
• Chat: chat artifacts, question outline and timestamps; context compaction and provider prompt-cache markers.
• Security: OIDC JWKS verification, optional complex passwords, document auto-tagging, and broad sandbox/security hardening.

• v0.7.2 — Product documentation site; knowledge base folder tree; chunk editing with revision history; wiki page revisions; directly loadable file URLs (resource_urls=public); Feishu Drive data source; batch tagging; MCP Server 1.1 (29 tools); AWS S3 default credential chain.
• v0.7.1 — Yunzhijia IM; Volcengine rerank; Zhipu AI web search; platform-scoped API keys; per-KB activity audit; FAQ filtering, tagging and export; Langfuse OTLP tracing; one-click Markdown export.
• v0.7.0 — Scoped API keys and principal model; task-queue dashboard and worker-pool governance; multiple storage instances per workspace; temporary chat attachments; @Skill / @MCP mentions; mid-conversation MCP OAuth; QQBot and Lark IM; Redis TLS; weknora CLI v0.10.
• v0.6.3 — Website embed widget & Integrations Center (secure-mode token exchange + rate limits); chat experience overhaul (citation popovers, RAG pipeline progress, streaming markdown); document multi-tag & batch reparse; Wiki folders & hierarchy navigation; RSS data source; MCP OAuth2; EPUB / MHTML parsing; agent model-readiness checks; model test debugger; session source filter; workspace deletion UI.
• v0.6.2 — Per-upload process configuration with upload-confirm dialog; document reparse with process_config; weknora CLI v0.9 (bundled Agent Skills, session stop, auth/profile harmonization); KB marquee multi-select; HNSW index for 1024-dim pgvector embeddings; chat resources store refactor; Langfuse-only tracing (Jaeger removed).
• v0.6.1 — Document parsing trace timeline (Langfuse-style span tree with stage-by-stage progress + stop-parse); OpenSearch vector store driver; declarative built-in models via YAML; system admin & consolidated platform settings + audit log; new-user onboarding guide; settings UI redesign; weknora CLI v0.7 / v0.8 (agent-first wire contract, NDJSON, --dry-run); OpenDataLoader + PaddleOCR-VL parsers; MCP server multi-transport (stdio / SSE / HTTP); per-model thinking-mode config; Tencent LKEAP rerank + native Gemini embeddings + MiniMax-M3.
• v0.6.0 — Workspace RBAC (4-tier role matrix Owner / Admin / Contributor / Viewer + per-KB ownership + per-workspace audit log), workspace member management & multi-workspace UX, self-service workspaces; weknora CLI v0.4 GA with mcp serve; KB retrieval fan-out across vector stores; AES-256-GCM credential encryption + docreader gRPC TLS + Token; Zhipu embedder + Huawei OBS; server-side user preferences; Go 1.26.0. See Tenants & auth.
• v0.5.2 — Wiki ingest scales to 40k-document KBs (task queue + DLQ); MCP human-in-the-loop tool approval; Anthropic / Apache Doris / Tencent VectorDB / KS3 / SearXNG backends; adaptive 3-tier chunking with live preview; global ⌘K command palette; Yuque connector + WeChat Mini Program; weknora CLI preview.
• v0.5.1 — Knowledge-base batch management; workspace-wide IM channels overview; session search + user-scoped pinning; unified Model / Web Search / MCP settings cards; per-agent LLM timeout; desktop workspace switching.
• v0.5.0 — Wiki Mode GA — agents auto-generate structured, interlinked Markdown wiki pages with a knowledge graph; wiki browser + visual graph in the UI.
• v0.4.0 — WeKnora Cloud (hosted LLM + parsing); Chrome Extension; ClawHub Skill; WeChat IM; attachment processing; Azure OpenAI / Alibaba OSS; Notion connector; Baidu + Ollama web search; VectorStore management.
• v0.3.6 — ASR (audio); Feishu data-source auto-sync; OIDC; IM quote-reply context + thread-based sessions; document summarization; Tavily search; parallel tool calling; agent @mention scope restriction.
• v0.3.5 — Telegram / DingTalk / Mattermost IM; IM slash commands + QA queue; suggested questions; VLM auto-describe MCP tool images; Novita AI; channel tracking.
• v0.3.4 — WeCom / Feishu / Slack IM; multimodal image support; NVIDIA model API; Weaviate; AWS S3; AES-256-GCM API-key encryption; built-in MCP service; hybrid-search optimization; final_answer tool.
• v0.3.3 — Parent-child chunking; KB pinning; fallback response; passage cleaning for rerank; storage auto-creation; Milvus.
• v0.3.2 — Knowledge Search entry; per-source parser & storage engine config; image rendering in local storage; document preview; Volcengine TOS; Mermaid rendering; batch session management; memory graph preview.
• v0.3.0 — Shared Space; Agent Skills + sandboxed execution; custom agents; Data Analyst agent; thinking mode; Bing / Google web search; API Key auth; Helm chart; Korean i18n; Qdrant.
• v0.2.0 — Agent Mode (ReACT); multi-type knowledge bases (FAQ + document); conversation strategy config; DuckDuckGo web search; MCP tool integration; new UI with agent mode switching; MQ async task management.

Full history: CHANGELOG.md.

Product Tour

Quick Q&A and smart reasoning

Two ways to ask. Quick Q&A answers from the knowledge base with RAG and cites the sources it used. In smart reasoning the agent plans multi-step work, searching, reading documents and calling tools and skills, and shows each step in the conversation. Docs →

Local browser

Operate the browser on your computer. Through Tencent's open-source BrowserSkill extension, the agent opens pages and fills in forms in your own Chrome or Edge, and hands over to you for logins and CAPTCHAs. Docs →

Skills and sandbox

Run skills and produce files. Docker, E2B and Cube backends are supported. Turns in the same session share one workspace, and generated files can be previewed and downloaded. Open the graphical desktop or interactive terminal beside the chat to follow each step and take over when needed. Docs →

Toolbox: MCP services and skills

Tools the agent can use. Connect external MCP services and choose, tool by tool, which are enabled and which calls need approval. Install skills from ClawHub, SkillHub, Git or ZIP, manage them per workspace, and reuse them across sandboxes. Docs →

Automatic wiki

Documents organized into a browsable wiki. With Wiki enabled, WeKnora extracts people, products and concepts from knowledge-base documents into pages with source citations, organized by directory. The knowledge graph shows how pages relate; pages can be edited directly and every change can be rolled back. Docs →

Observability

Tracing and runtime monitoring. Langfuse traces the reasoning, tool calls and token usage of each agent step. The document parsing timeline shows progress stage by stage, and the task-queue dashboard lists queued and failed tasks. Docs →

Architecture

A modular pipeline from document parsing, vectorization and retrieval to LLM inference, in which every component can be replaced or extended. It runs locally or on a private cloud, and the Web UI needs no setup to get started. More: Architecture overview · RAG pipeline · Extension points.

Features

Area	Highlights
Q&A and agent	Quick Q&A answers from knowledge bases with citations; smart reasoning runs a ReAct agent over knowledge bases, web search, MCP tools, skills and the local browser. Steer, fork or rewind a running conversation, and keep long-term memory across sessions
Wiki	Agent-generated, interlinked wiki pages with a knowledge graph; in-browser editing, revision diff and rollback
Skills and sandbox	Skill catalog installed from ClawHub / SkillHub / Git / ZIP; session-persistent Docker / E2B / Cube sandboxes with network policy; terminal and graphical desktop beside the chat
Knowledge bases	FAQ, document and wiki bases; folder tree; chunk editing with revisions; per-upload parsing, chunking and multimodal settings; auto-tagging
Retrieval	Keyword + vector hybrid search, rerank, parent-child chunking and GraphRAG (Neo4j); end-to-end evaluation with recall and BLEU / ROUGE
Access and security	Workspace RBAC with four roles and an audit log; scoped API keys; OIDC; AES-256-GCM credential encryption; SSRF-safe outbound requests with a whitelist-only mode
Operations	Langfuse tracing for agent steps, tokens and pipelines; document parsing timeline; task-queue dashboard with worker pools; automatic migrations on upgrade

Supported backends

Component	Options
LLMs	29 built-in vendors, including OpenAI / Azure OpenAI / Anthropic / DeepSeek / Qwen / Zhipu / Hunyuan / Doubao / Gemini / MiniMax / NVIDIA / SiliconFlow / OpenRouter / LiteLLM / Ollama
Embeddings	Ollama / BGE / GTE / Zhipu / OpenAI-compatible APIs
Vector databases	PostgreSQL (pgvector + ParadeDB) / Elasticsearch / OpenSearch / Milvus / Weaviate / Qdrant / Apache Doris / Tencent VectorDB
Object storage	Local / Tencent Cloud COS / MinIO / AWS S3 / Volcengine TOS / Alibaba Cloud OSS / Kingsoft Cloud KS3 / Huawei Cloud OBS
Document formats	PDF / Word / PPT / Excel / CSV / TXT / Markdown / HTML / EPUB / MHTML / JSON / XMind / images
Data sources	Feishu wiki / Feishu Drive / Lark / Confluence / GitLab / Tencent IMA / Notion / Yuque / DingTalk Docs / Seafile / RSS
IM channels	WeCom / Feishu / Lark / QQBot / Slack / Telegram / DingTalk / Mattermost / WeChat / Yunzhijia
Web search	DuckDuckGo / Bing / Google / Tavily / Baidu / Ollama / SearXNG / Keenable / Zhipu AI / Exa / Metaso / Bocha / Serply / SearchApi
Deployment	Docker Compose / Kubernetes (Helm) / Lite single binary / desktop app; offline and private-cloud installs; UI in Chinese, English, Japanese, Korean and Russian

Clients and Integrations

	Client	What it does
	CLI weknora	Agent-first command line for the full API, with a curated MCP tool surface and bundled Agent Skills
	Built-in MCP Server	Publishes knowledge bases to Cursor, Claude and other MCP clients over Streamable HTTP; the Python mcp-server/ is deprecated
	Local Browser (BrowserSkill)	Lets agents operate the user's own Chrome / Edge
	Chrome Extension	Select text, images, or entire pages in the browser and save them as knowledge entries with one click, without copy-paste or file upload
	WeChat Mini Program	Lightweight mobile client: configure API access, select knowledge bases, import URLs, and ask knowledge chat from WeChat
	ClawHub Skill	A WeKnora skill on ClawHub for document import, hybrid search and knowledge management via the REST API
	DeepSeek Harness plugin	Gives dsh coding agents four read-only tools: search, read document, ask and list knowledge bases
	Website Embed Widget	Publishes agents on external sites
	Go SDK	CRUD for knowledge bases, documents and sessions, plus SSE streaming Q&A
	WeChat Dialog Open Platform	Hosted Q&A built on WeKnora: upload knowledge and publish a Q&A service in WeChat without writing code

Command-Line Interface

weknora is the official CLI for driving the API from a terminal or an AI agent. It is agent-first: every command emits a stable JSON envelope by default (with typed error codes mapped to exit codes), and --format text renders for humans. It also serves a curated MCP tool surface (weknora mcp serve) and ships bundled Agent Skills.

weknora profile add prod --host https://kb.example.com --use
weknora auth login
weknora kb list
weknora link --kb my-knowledge-base    # bind the current directory
weknora doc upload notes.md
weknora chat "summarise the design doc"

For headless / CI use, set WEKNORA_API_KEY + WEKNORA_HOST and skip auth login; no credentials are written to disk. See cli/README.md for install + 5-minute quickstart and cli/AGENTS.md for the operational contract AI agents rely on.

Documentation

The full product documentation lives at weknora.weixin.qq.com/docs (in Chinese), organized as Getting Started → Architecture → Features → API → Clients → Development and covering ~360 API endpoints and ~150 environment variables.

Start here
Introduction	Capabilities overview
Installation	Docker Compose, Helm, Lite and desktop
Configuration	Environment variables and models
Troubleshooting FAQ	Common problems and fixes
API reference	REST API overview
Release notes	What changed in each release

Development

If you need to frequently modify code, you don't need to rebuild Docker images every time. Use fast development mode:

make dev-start      # Start infrastructure
make dev-app        # Start backend (new terminal)
make dev-frontend   # Start frontend (new terminal)

• Frontend modifications auto hot-reload (no restart needed)
• Backend modifications quick restart (5-10 seconds, supports Air hot-reload)
• No need to rebuild Docker images
• Supports IDE breakpoint debugging

See the Development guide for details.

The documentation site and product homepage are built from website-docs/. With Node.js 24, run cd website-docs && npm run setup && npm run build && npm run preview to preview both together; the unified static output serves the homepage at / and documentation at /docs/. See the directory's README for Nginx and Docker deployment.

Contributing

Welcome to submit Issues or Pull Requests.

• Process: Fork → Create branch → Commit changes → Open PR
• Standards: Format code with gofmt, follow Conventional Commits (feat: / fix: / docs: / test: / refactor:)

For a focused PR, validate the changed scope first:

git fetch origin main
git diff --check origin/main...HEAD
golangci-lint run --new-from-rev=origin/main ./...
go test ./path/to/changed/package -count=1

Run gofmt on changed Go files before committing. For frontend changes, run the relevant tests from frontend/ and use npm run type-check when the change affects TypeScript or Vue components.

The full maintainer gate remains:

make fmt
make lint
make test

make fmt formats the entire Go repository, so run it only with a clean worktree and review the resulting diff. Some full-suite tests require local infrastructure or service configuration. If a full check fails for an unrelated baseline or environment reason, include the exact command and failure in the PR while still providing passing targeted tests for your change.

Contributors

Thanks to everyone who has contributed:

License

This project is licensed under the MIT License. You are free to use, modify, and distribute the code with proper attribution.