HilbertRaum：离线问答本地文档

AI, fully local on your computer.

Your private AI workspace, fully offline Chat with a local AI, ask questions about your private documents and an offline Wikipedia, and keep everything on your own computer. Quick start: download the latest release, install it, and the app walks you through adding the AI engine and a model. HilbertRaum is a private AI workspace that runs locally on your computer and can operate entirely offline. Chat with a local AI, ask questions about your documents, and keep your data under your control. There is no cloud fallback, no web search, and no telemetry. Your prompts, documents, embeddings, and chat history stay on local storage. The workspace is encrypted and portable. The app, models, and data can all live on a hard drive or USB drive, allowing you to move the entire setup between machines. HilbertRaum is open source and designed to run on a wide range of hardware, from CPU-only systems to single-GPU and unified-memory machines. The model catalog is intentionally small: we test each model for quality, speed, and hardware compatibility. HilbertRaum then selects an appropriate model for your machine, so you don't need to compare models or tune technical settings. 🔒 Private by design. No cloud, no telemetry, no analytics. Nothing you type or import is uploaded. 🧳 Portable and encrypted. Keep models and a password-encrypted workspace on an external drive and move between computers. 🧠 Local models. A llama.cpp runtime with GGUF models from a curated open-weight catalog (Qwen3, Ministral, Gemma, Granite). The app benchmarks your machine and recommends one. 📄 Document Q&A with citations. Import PDFs, Word files, or text, ask questions, and get answers grounded in your files. Retrieval combines vector and keyword search with a reranker and can be scoped to your library, a project, a section, specific documents, or the files attached to a chat. 📚 Knowledge packs — an offline Wikipedia (optional). Register ZIM archives — Wikipedia in ~100 languages, Wiktionary, Wikivoyage — as per-chat sources searched and cited alongside your documents, with the article readable offline in the app. Needs the kiwix-tools binaries (GPL-3.0-or-later) on the drive — the app offers to install them with your consent, the first time you need them, or you can place them yourself. See Knowledge packs. 🖼️ Image understanding. Ask questions about a picture with a local vision model. The analysis history is encrypted at rest and can be deleted. 🎙️ Audio and voice. Transcribe audio files with Whisper, dictate prompts, and OCR scanned pages. 🌐 Translation. A dedicated screen for text and whole documents (local TranslateGemma sidecar, 51 languages). Translated documents land back in your library. ✅ Evidence review. Turn a document answer into a reviewable record: every statement is checked against its frozen source snippets, decisions and notes are saved, and the result exports as an HTML or PDF evidence pack. 🛠️ Document tasks and skills. Summarize, translate, and compare documents, or install reusable skills for structured extraction (bank statements, invoices, meeting protocols, redaction). The skills overview describes each bundled skill. 🪟 Cross-platform. Runs on Windows, macOS, and Linux. 🔌 Local API (opt-in). Other programs on the same computer can use your running model through an OpenAI-compatible loopback endpoint: point any client at http://127.0.0.1:4980/v1. It is off by default, never touches the internet, exposes only completions (no documents, no conversations), and keeps no record of what was asked. See docs/local-api.md. Table of contents Status Which path are you on? What you need Getting started (DIY / from source) Supported models Knowledge packs — an offline Wikipedia Two distribution paths Documentation For developers Privacy & security Contributing License Status The app is feature-complete. You can explore the whole interface without downloading a model, and real AI answers start as soon as you add one (step 2 below). What remains before the first polished release is manual release testing: signed installers and a live demo run. BUILD_STATE.md tracks the details, and docs/known-limitations.md lists the gaps we have accepted for now. Which path are you on? You set it up yourself (free, open source). Keep reading: you install the app, download the models, and point the app at them. You want it ready-made. We are preparing the HilbertRaum AI Kit, a preconfigured drive with tested hardware and preloaded models: plug it in, double-click Start HilbertRaum, and follow the user guide. Join the waitlist at hilbertraum.ai. What you need A computer running Windows, macOS, or Linux, with at least 8 GB of RAM. Linux: Ubuntu 22.04, Debian 12 or Fedora 36, or newer. The AI engine needs glibc 2.34, GCC 12's C++ library and OpenSSL 3, and uses the system's OpenMP library (libgomp1), which desktop installations include. Windows: the AI engine uses Microsoft's Visual C++ 2015–2022 runtime (x64), which most PCs already have. If either is missing, the app names it on the AI Model screen; see The AI engine can't run on this computer. Memory decides which model you get. The app detects your RAM and graphics memory (VRAM) automatically, benchmarks your machine, and recommends the model that fits best. On a computer without a usable graphics card the RAM tiers below decide directly; on a computer with one, the recommendation only moves off the RAM tier when that pick would not fit in the card's free memory, and it never recommends a model below the card's RAM floor either way. RAM (no usable graphics card) Recommended model 8-11 GB Qwen3.5 4B 12-15 GB Gemma 4 E2B 16-23 GB Qwen3.5 9B 24 GB Qwen3.8 27B (UD-Q4_K_M) 32 GB and up Qwen3.8 27B (UD-Q5_K_M) These are best-fit recommendations, not hard minimums. Each model's actual floor is the Min RAM column in the model table below; Qwen3.5 9B, for example, already runs from 12 GB. The MoE models stay opt-in. Disk space: about 3 GB for the smallest hand-built setup (the 4B chat model plus