An AI-Powered Personal Knowledge Base with Semantic Section Search.
KnowledgeDump is designed for individuals who need to capture information freely — notes, PR comments, research, code snippets, meeting dumps — and retrieve exactly the right section later using natural language queries.
- Semantic Search: Find exactly what you are looking for using conceptual searches, not just keyword matches. Returns section-level results with AI-generated descriptions. Semantic search embeddings are computed on-device using Transformers.js.
- Auto-tagging: Extracts 1-4 concept tags from your text automatically in the background as you write.
- Knowledge Graph: Visualize connections between your notes natively in a beautiful 2D force-directed graph.
- Offline First: All data is stored locally in SQLite. The vector embeddings run entirely on-device (via Transformers.js), and AI functions can transparently fall back to local Ollama if offline.
- Zero Ongoing Cost: Designed to operate entirely on free-tier models (Gemini Flash for APIs, local HuggingFace/Ollama models for everything else).
Currently, KnowledgeDump is built from source. Ensure you have Node.js and Rust installed.
npm install
npm run devOn your first launch, the app will welcome you and ask for a Gemini API Key.
- Get a free Gemini API key from Google AI Studio.
- Enter it into the Onboarding screen or the Settings gear (⚙️) in the sidebar.
If you want to run completely offline without Gemini:
- Install Ollama.
- Pull the Llama 3 model:
ollama run llama3 - KnowledgeDump will automatically fallback to the local Ollama instance running on port
11434if it fails to reach Gemini.
- Capture: Click the
+icon to dump a new note. Write in plain text or Markdown. - Search: Press
Ctrl+Kto open the semantic search overlay. - Graph View: Press
Ctrl+Gto open the interactive note graph.
Built with Tauri, React, SQLite, and LanceDB.