What it does
LLM-Wiki is a RAG system for your own documents: ask questions by text or voice and get answers backed by the sources they came from. It runs as a web app, a desktop app (Electron) and a REST API.
Highlights
- Hybrid search: semantic vector search (nomic-embed-text with Chroma) plus TF-IDF keyword search, merged with reciprocal rank fusion
- Many formats: PDF (native and OCR), images (Tesseract), Excel, Word, PowerPoint, CSV, HTML, Markdown and plain text
- Offline voice input with Vosk, streaming answers
- Automatic document tagging, comparison and summaries
- Charts straight from the chat, UI in German, English and Italian
Tech
JavaScript Python RAG Chroma Electron Vosk
Status
Actively maintained, public on GitHub. Last updated in October 2026.
Source code
The source code is public on GitHub: github.com/Mikweb2025-design/LLM-Wiki