SWX KI-Assistent

SWX KI-Assistent is a complete RAG system (Retrieval-Augmented Generation) that runs entirely on your own Windows PC or server. Import PDFs, Word, Excel, text/CSV/Markdown files or OTRS ticket exports; ask questions in plain language; get concise answers with source citations, generated by a local LLM (any GGUF model via llama.cpp). Retrieval combines semantic search (bge-m3 embeddings) with BM25 keyword search and a cross-encoder reranker, so rare technical terms and error codes are found reliably. If nothing matches, it honestly says so instead of hallucinating.

Highlights: native Windows app (English/German) with setup wizard, web UI and HTTP API, knowledge spaces, runs standalone / as a network server / as a Windows service, answer cache, optional INT8/ONNX search backend for CPU-only machines, incremental re-imports. GPU optional (CUDA/Vulkan). Two packages: Lite (~85 MB) or full offline bundle (~4 GB) for air-gapped networks. Binaries are Authenticode-signed. Free for individuals and small businesses (PolyForm Small Business License, source-available).