mrsibe/KnowNote

A local-first AI knowledge base & NotebookLM alternative built with Electron.

What it solves

KnowNote is a local-first research workspace that allows users to create a knowledge base from their own documents (PDFs, Word, PowerPoint, and web pages) and query them using AI. It solves the problem of trusting AI-generated answers by providing traceable citations back to the original source passages, while ensuring privacy by keeping the parsing, embedding, and vector search local to the user's machine.

How it works

The application uses a Local RAG (Retrieval-Augmented Generation) architecture. It parses documents into structured text, splits them into overlapping passages, and generates embeddings using a local ONNX-based model (Xenova/multilingual-e5-small). These embeddings are stored in a local SQLite database with sqlite-vec for vector search. When a user asks a question, the app retrieves relevant passages and sends them to a user-configured model endpoint (such as Ollama, OpenAI, Anthropic, or Google).

Who it’s for

It is designed for researchers and individuals who need to analyze large sets of documents privately and want to verify AI claims against source material without deploying a complex server stack.

Highlights

  • Local-first processing: Parsing, splitting, and embedding happen entirely on the desktop app, requiring no Docker or server setup.
  • Traceable answers: Every AI response includes references to the specific passages and character offsets in the source text.
  • Model flexibility: Supports any OpenAI-, Anthropic-, or Google-compatible endpoint, including local servers like Ollama.
  • MCP Integration: Can serve a notebook as a read-only MCP server, allowing AI agents (like Claude Desktop or Cursor) to search the user's document library.
  • Knowledge organization: Includes tools for saving structured notes and generating mind maps from notebooks.

Related

  • Project
  • Project
  • Project
  • Dispatch
  • Project