Revornix
A self-hosted AI knowledge workspace with document chat, Markdown editing and podcast generation. Works with OpenAI-compatible providers.
Revornix is a self-hosted AI workspace for people who collect research, articles and meeting recordings and want to search or ask questions across them. It turns saved material into Markdown, connects it in a personal knowledge graph, and produces summaries, illustrated reports or two-voice podcasts.
You can host the stack on your own infrastructure, with documents and databases under your control, or use the hosted workspace. A browser interface and Electron desktop shell for macOS and Windows provide access. AI processing depends on your chosen providers: the default embedding service runs in the cloud, while local embedding is optional. It supports OpenAI-compatible providers, and you can choose separate engines for document conversion, embeddings, summaries, audio and illustrations.
The library accepts web pages, PDFs, Office documents, plain text and audio. Audio transcription includes speaker separation for meetings. Text and vector search work across private documents, while the assistant uses those documents and their relationships to answer questions. Milvus handles vector retrieval, and Neo4j stores the personal graph used by GraphRAG.
Collections can stay private, be shared with collaborators or appear on public pages. Automatic daily digests gather saved material, and notifications arrive through email, push or the app. The Markdown editor supports tables, Mermaid diagrams and math. MCP support lets the workspace use external tools and makes the library accessible to MCP-aware assistants; a public API, Python SDK and CLI also support automated collection.
