WeKnora
Open-source knowledge platform that turns documents into searchable RAG knowledge, autonomous agent workflows, and a self-maintaining wiki.
WeKnora is an open-source, self-hostable knowledge platform for teams that need to work with documents and make the resulting knowledge searchable. Its quick Q&A mode uses retrieval-augmented generation, while a ReAct agent can orchestrate retrieval, MCP tools, the skill catalog, sandboxes and web search for multi-step tasks. Wiki Mode turns source documents into an interlinked Markdown knowledge base with an interactive graph, manual editing, revision history and rollback. The platform also supports folder-structured uploads, editable retrieval chunks with revision history, and long-term memory across sessions. The repository describes ingestion from sources such as Feishu, GitLab, Tencent IMA, Notion, Yuque, DingTalk Docs and RSS, plus Q&A access through WeCom, Feishu, Slack and Telegram. It supports documents including PDF, Word, images, Excel and XMind. Website embed widgets, scoped API keys, multiple storage backends, LLM provider integrations, Langfuse observability and multi-workspace RBAC are also documented. To try it, follow the official repository's Docker Compose or development setup and configure the required LLM and storage services. WeKnora is self-hostable, but the deployment has multiple infrastructure and provider dependencies; the exact integrations and limits should be checked against the current documentation.