Kwipu

Kwipu helps people query a document folder or Obsidian vault and trace answers back to their notes. It connects information across documents so you can investigate project decisions, people and their roles, or relationships between ideas.

• Answers with citations: Ask questions across your documents and check the source passages behind each answer; search combines semantic similarity, keyword matching and time metadata.
• Connections between notes: Kwipu reads wikilinks and YAML frontmatter, then uses an LLM to extract relationships from prose. It resolves aliases and links to note sections while keeping identically named notes in different folders separate.
• Interactive graph: Explore connections in a 3D browser view and open nodes to inspect attributes, backlinks, source passages and the lines behind structural relationships.
• Terminal and MCP access: Query the same index from a terminal alongside Obsidian or through an AI client that supports MCP. The terminal watches your folder for document changes.

Kwipu accepts Markdown, plain text, PDF and Word documents without rewriting the originals. It supports relation patterns in English, Italian, French, German, Spanish and Portuguese.

It's open source under the MIT license and runs on Windows, macOS and Linux using Ollama. Indexing and answers stay on your machine when both the language model and embedding model run locally. The default language model, gpt-oss:20b-cloud, sends document chunks, retrieved context and questions to its provider; qwen2.5:7b is a supported local choice.