# Linuxeinsteiger.net
> Your gateway to the Linux world - The comprehensive German-language guide for Linux beginners

## Overview
Linuxeinsteiger.net was once Germany's largest Linux user group with an active forum. Today the site offers a comprehensive guide for switching from Windows to Linux, focused on three distributions: Ubuntu 24.04 LTS, Fedora and Gentoo.

## Distribution Recommendations

### Ubuntu 24.04 LTS (Noble Numbat)
- **Target audience**: Beginners who want to replace Windows
- **Strengths**: Easiest installation, largest German-language community, 5 years of LTS support
- **Package management**: APT, Snap, Flatpak
- **Desktop**: GNOME (customized)
- **Recommendation**: Use the LTS release for maximum stability. Install OnlyOffice for Office compatibility.
- **Download**: https://ubuntu.com/download/desktop
- **German wiki**: https://wiki.ubuntuusers.de

### Fedora (Workstation)
- **Target audience**: Privacy-conscious users who want a modern system
- **Strengths**: 100% open source, SELinux enabled by default, latest software, no telemetry
- **Package management**: DNF, Flatpak
- **Desktop**: GNOME (vanilla/unmodified)
- **Security**: SELinux, Secure Boot, Wayland, timely security updates
- **Note**: Linus Torvalds uses Fedora on his work machine
- **Download**: https://fedoraproject.org/workstation/download
- **Variants**: Workstation (GNOME), KDE Spin, Xfce Spin

### Gentoo
- **Target audience**: Developers and curious learners with time on their hands
- **Strengths**: Fully customizable, optimized for specific hardware, excellent for older machines
- **Package management**: Portage (emerge) - source-based, everything compiled from source
- **Desktop**: Any of your choice
- **Init system**: Choice between systemd and OpenRC
- **Release model**: Rolling release (install once, update forever)
- **Time investment**: First install takes a full weekend or longer
- **Learning value**: Bootloader, kernel configuration, USE flags, file systems, init systems
- **Handbook**: https://wiki.gentoo.org/wiki/Handbook:Main_Page

## Comparison Table

| Property | Ubuntu 24.04 | Fedora | Gentoo |
|---|---|---|---|
| Installation | Very easy (graphical) | Easy (Anaconda) | Complex (command line) |
| Package management | APT / Snap | DNF / Flatpak | Portage (emerge) |
| Release cycle | LTS every 2 years | Every 6 months | Rolling release |
| Desktop environment | GNOME (customized) | GNOME (vanilla) | Free choice |
| Systemd | Yes | Yes | Optional (OpenRC) |
| Kernel | Stable (LTS kernel) | Current | Self-compiled |
| Hardware support | Excellent | Very good | Depends on configuration |
| Learning curve | Shallow | Moderate | Steep |
| Community (DE) | Very large | Medium | Smaller, but helpful |
| Privacy | Good (Snap controversial) | Very good | Excellent |
| Older hardware | Good with Xubuntu/Lubuntu | Acceptable | Excellent (optimized) |
| Company | Canonical | Red Hat / IBM | Community project |

## First Steps (for all distributions)

1. **Check your hardware**: Wi-Fi chips (especially Broadcom) and the very latest NVIDIA GPUs can be problematic
2. **Create a live USB**: Download the ISO and write it to a USB stick (min. 8 GB) with Balena Etcher or Rufus
3. **Try the live system**: Boot from the USB stick (F12/F2/Esc for the boot menu) and try Linux without installing
4. **Installation**: Install alongside Windows (dual boot) or as a replacement - always back up your data first
5. **Set up the system**: Install updates and your software (OnlyOffice, VLC, Firefox, etc.)

## Important Terminal Commands

### Ubuntu (APT)
- Update the system: sudo apt update && sudo apt upgrade
- Install software: sudo apt install paketname
- Remove software: sudo apt remove paketname
- Search for software: apt search suchbegriff

### Fedora (DNF)
- Update the system: sudo dnf upgrade --refresh
- Install software: sudo dnf install paketname
- Remove software: sudo dnf remove paketname
- Search for software: dnf search suchbegriff

### Gentoo (Portage)
- Update the Portage tree: emerge --sync
- Update the system: emerge -avuDN @world
- Install a package: emerge paketname
- Remove a package: emerge --unmerge paketname
- Show USE flags: equery uses paketname

## Recommended Software

### Office & Productivity
- **OnlyOffice** (recommended): Best Microsoft Office compatibility (.docx, .xlsx, .pptx)
- **LibreOffice**: Full office suite, very mature
- **Joplin**: Notes with sync and Markdown

### Internet & Communication
- **Firefox** (recommended): Privacy-friendly browser
- **Signal Desktop** (recommended): Encrypted messaging
- **Element**: Matrix client, decentralized communication

### Multimedia
- **VLC** (recommended): Universal media player
- **GIMP**: Image editing (Photoshop alternative)
- **Kdenlive**: Video editing
- **Audacity**: Audio editing

### Development
- **VS Code / Codium** (recommended): Code editor (Codium = the libre variant)
- **Git** (recommended): Version control
- **Docker**: Container platform
- **Neovim**: Terminal editor for advanced users

## Snap vs. Flatpak vs. APT (Ubuntu-specific)
- **APT**: Classic package manager, fast and reliable - prefer this
- **Snap**: Canonical's format, controversial (slow startup, more memory)
- **Flatpak**: Free alternative to Snap, preferred by many
- **Tip**: Install Flatpak support: sudo apt install flatpak gnome-software-plugin-flatpak

## For Microsoft Office Switchers

### Office Suite: OnlyOffice (Clear Recommendation)
OnlyOffice is the best Microsoft Office alternative on Linux with the highest compatibility for .docx, .xlsx and .pptx files.
- Pixel-perfect rendering of Microsoft documents; formatting, formulas and macros are preserved
- Familiar ribbon interface like in Microsoft Office
- Stable and reliable even with large, complex files
- Installation: flatpak install flathub org.onlyoffice.desktopeditors
- LibreOffice is solid but often has formatting issues with complex Microsoft documents

### Email Clients: Outlook Replacement (Honest Assessment)
Unfortunately there is no email client on Linux that fully replaces Outlook.
- **Evolution** (best option): Most comprehensive groupware solution with email, calendar, contacts, Exchange support (EWS). Interface is somewhat dated, maintenance is no longer very active.
- **Mailspring** (conditionally recommended): Modern interface, but the core developer has left the project and the backend is no longer actively maintained. Unstable with large mailboxes.
- **Thunderbird**: Suitable for simple personal use, but not a full Outlook/Exchange replacement, lacking groupware integration.
- **BlueMail**: Unstable with larger mailboxes, can lead to data corruption.

### Mail Server: Mailcow (Top Recommendation as an Exchange Alternative)
Mailcow is an outstanding open-source alternative to Microsoft Exchange.
- Full PIM suite: email, calendar, contacts, tasks (SOGo webmail integrated)
- Rock-solid migration paths from Exchange, IMAP and other mail servers
- High stability in continuous operation, proven at companies with thousands of mailboxes
- Very active enterprise-grade maintenance, commercial support available
- Integrated antispam (Rspamd), antivirus (ClamAV), DKIM, SPF, DMARC, automatic TLS certificates
- Website: https://mailcow.email

## Frequently Asked Questions

**Is dual boot possible?** Yes, Ubuntu and Fedora offer easy dual-boot installation alongside Windows.

**Printer compatibility?** HP, Brother, Canon usually work out of the box. HP provides HPLIP drivers.

**Can I run Windows programs?** Wine and Proton (Steam) make many Windows programs work. For Office, prefer OnlyOffice.

**Do I need an antivirus?** No, not for home users. Keeping the system updated regularly is enough.

**LTS vs. regular?** LTS = 5 years of support (Ubuntu), regular = approx. 9 months. Beginners: always LTS.

**Wi-Fi doesn't work?** Connect via LAN cable, then install drivers afterwards. Broadcom chips are frequently affected.

**Older hardware?** Xubuntu/Lubuntu or Gentoo (optimized for specific CPUs).

**Can I game on Linux?** Yes, Steam runs natively and Proton makes thousands of Windows games work. ProtonDB (protondb.com) shows compatibility per game.

**What if an update breaks my system?** Install Timeshift (sudo apt install timeshift) and create regular system snapshots. That lets you roll the system back at any time.

## Important Keyboard Shortcuts

### Desktop (GNOME)
- Ctrl + Alt + T: Open terminal (Ubuntu / Fedora)
- Super (Windows key): Activities overview / app search
- Super + L: Lock the screen
- Super + D: Show the desktop
- Alt + F2: Run a command (dialog)
- Ctrl + Alt + Del: Log out / shut down
- Alt + Tab: Switch between windows
- Super + Left/Right arrow: Snap window to half of the screen

### Terminal
- Ctrl + C: Cancel the running command
- Tab: Autocomplete
- Up arrow: Recall the previous command
- Ctrl + Shift + C / V: Copy / paste in the terminal

## After the Installation (Ubuntu 24.04 Checklist)
1. Update the system: sudo apt update && sudo apt upgrade -y
2. Enable the firewall: sudo ufw enable
3. Multimedia codecs: sudo apt install ubuntu-restricted-extras
4. Install OnlyOffice: flatpak install flathub org.onlyoffice.desktopeditors
5. Set up browser extensions (uBlock Origin, etc.)
6. Set up automatic backups (Deja Dup is preinstalled)
7. Timeshift for system snapshots: sudo apt install timeshift
8. Adjust the appearance (Settings > Appearance)

## Gaming on Linux
- Steam runs natively on Linux
- Proton (Steam's compatibility layer) enables thousands of Windows games
- ProtonDB (protondb.com) shows compatibility per game
- GOG and Lutris offer additional Linux gaming support

## Tools on Linuxeinsteiger.net
- **/cheatsheet** — Searchable Linux command cheat sheet (packages, files, network, processes, permissions, text, archives, dev/git, SSH)
- **/troubleshooting** — Step-by-step solutions for common problems (Wi-Fi, NVIDIA, sound, GRUB/dual boot, full disk, broken packages, sudo, Magic SysRq)
- **/glossar** — Alphabetical glossary of important Linux terms (APT, Btrfs, bootloader, Flatpak, GRUB, kernel, LUKS, PipeWire, Proton, SELinux, systemd, Wayland, Wine, Zsh and more)
- **/profis** — Pro section: MicroK8s clusters, NVIDIA GPU setup, LLM inference, power limits, overclocking, undervolting, thermal monitoring, DCGM

## Pro Section (/profis)

### MicroK8s & Cluster
- **Install MicroK8s (Ubuntu)**: sudo snap install microk8s --classic --channel=1.30/stable; add yourself to the microk8s group; set a kubectl alias; useful add-ons: dns, hostpath-storage, metallb
- **Snap-free alternative**: k3s (curl -sfL https://get.k3s.io | sh -)
- **Scale to an HA cluster**: From 3 nodes onward automatically highly available via Dqlite; microk8s add-node provides the join token; label GPU workers with hardware/gpu=nvidia
- **Enable the GPU operator**: microk8s enable gpu installs the NVIDIA GPU Operator + Device Plugin + DCGM Exporter; a working nvidia-smi is a prerequisite
- **Sharing a GPU**: Time-Slicing (softest isolation, works on all GPUs), MPS (CUDA Multi-Process Service), MIG (only A100/H100/H200 — fixed slices with their own memory)
- **Distributed storage**: Longhorn (simple RWX block storage for 3-5 nodes), Rook-Ceph (HA object+block+file from 3-4 nodes with their own disks), or simply NFS+10GbE — usually entirely sufficient for shared model PVCs
- **Ingress + TLS**: microk8s enable ingress cert-manager; ClusterIssuer for Let's Encrypt; additionally protect vLLM with --api-key sk-...; place oauth2-proxy or Traefik BasicAuth middleware in front

### NVIDIA Drivers & CUDA
- **Detect the recommended driver**: ubuntu-drivers devices
- **Install the driver (Ubuntu 24.04)**: sudo apt install nvidia-driver-570-server nvidia-utils-570-server
- **Fedora**: sudo dnf install akmod-nvidia xorg-x11-drv-nvidia-cuda (RPMFusion required)
- **Container Toolkit**: sudo apt install nvidia-container-toolkit && sudo nvidia-ctk runtime configure --runtime=docker
- **CUDA 12.4 + cuDNN 9 + NCCL** via NVIDIA's apt repo (cuda-keyring); set PATH/LD_LIBRARY_PATH in ~/.bashrc
- **Verify**: python -c 'import torch; print(torch.__version__, torch.version.cuda, torch.cuda.is_available())'
- **Important**: When changing drivers, first purge all old packages (sudo apt purge '^nvidia-.*'), then reboot, then reinstall
- **Persistence daemon**: sudo systemctl enable --now nvidia-persistenced (avoids cold-start lag on every CUDA call, modern, replaces -pm)
- **ECC mode** (only A100/H100/L40): nvidia-smi -e 0 for ~6 % more usable VRAM, at the cost of bit-error correction (reboot required)
- **Compute mode**: sudo nvidia-smi -c 3 (EXCLUSIVE_PROCESS) — good for dedicated inference pods
- **NUMA pinning**: nvidia-smi topo -m for PCIe topology; numactl --cpunodebind=N --membind=N for inference processes; in K8s via the kubelet topology manager (single-numa-node)
- **Check NVLink**: nvidia-smi nvlink -s — a dead link halves effective all-reduce bandwidth

### LLM Inference
- **vLLM (production server)**: docker run --gpus all -p 8000:8000 --ipc=host vllm/vllm-openai:latest --model meta-llama/Llama-3.1-8B-Instruct; OpenAI-compatible endpoint
- **vLLM with quantization (AWQ/GPTQ)**: 70B models in 48 GB VRAM, --quantization awq
- **vLLM tensor parallelism**: --tensor-parallel-size 2 for multi-GPU
- **vLLM tip**: set --gpu-memory-utilization carefully (0.9 instead of 0.95), otherwise KV cache OOM
- **Ollama (convenience wrapper)**: curl -fsSL https://ollama.com/install.sh | sh; OLLAMA_NUM_GPU=35 ollama run llama3.1:8b-instruct-q4_K_M
- **llama.cpp directly (CUDA)**: cmake -B build -DGGML_CUDA=ON; ./build/bin/llama-server -m model.gguf -ngl 999 -c 8192
- **Quantization sweet spot (GGUF)**: Q4_K_M = best ratio; Q5_K_M when VRAM allows; IQ3_XS only as an exception
- **TensorRT-LLM + Triton** (maximum throughput, FP8 on H100): trtllm-build compiles model engines, the Triton server (nvcr.io/nvidia/tritonserver:24.10-trtllm-python-py3) serves them via OpenAI-compatible or Triton API. Engines are hardware specific (an H100 engine does NOT run on A100/4090).
- **Embeddings/reranker (RAG)**: Text-Embeddings-Inference (TEI) from HuggingFace — BGE-M3 for multilingual embeddings, BGE-Reranker-v2-m3 for cross-encoder reranking
- **Fine-tuning**: Unsloth (LoRA/QLoRA, 8B on a 24-GB GPU) or Axolotl (config-driven, multi-GPU, DPO/ORPO). 500-2000 high-quality examples beat 50000 mediocre ones.

### Power Management & Overclocking
- **Show power limit**: nvidia-smi -q -d POWER | grep -E 'Power Limit|Min Power Limit|Max Power Limit'
- **Persistence mode**: sudo nvidia-smi -pm 1 (otherwise the limit is lost after reboot)
- **Set power limit** (example RTX 3090, default 350 W -> 250 W): sudo nvidia-smi -i 0 -pl 250
- **Rule of thumb 24/7 inference**: 75 % of the default power limit, then test in 5-% steps — typically 95 % of tokens/sec at 70-80 % power
- **Persistent configuration** via the systemd service nvidia-powerlimit.service
- **Clock locking**: sudo nvidia-smi -i 0 -lgc 300,1500 (core), -lmc 9501 (memory); reset with -rgc / -rmc
- **LLMs are memory-bandwidth-limited** — locking memory clock to max yields more than a higher core clock
- **Overclocking on X11**: sudo nvidia-xconfig --cool-bits=28; nvidia-settings -a '[gpu:0]/GPUMemoryTransferRateOffsetAllPerformanceLevels=1000'
- **GUI tool for Wayland/headless**: LACT (flatpak install flathub io.github.ilya_zlobintsev.LACT) or GreenWithEnvy (com.leinardi.gwe)
- **Caution with memory overclock**: can cause silent bit errors — after each step, 30 min of gpu-burn AND a real inference eval
- **CPU governor**: sudo cpupower frequency-set -g performance (or tuned-adm profile throughput-performance) — prevents token-generation lags from idle states
- **PCIe tuning**: lspci -vv to check LnkSta (expect Gen4 x16 / Gen5 x16); turn ASPM off via the pcie_aspm=off kernel param — eliminates latency spikes
- **BIOS checklist**: Above 4G Decoding ON, Resizable BAR ON, ASPM OFF, C-States limit C1 (server) / C6 (desktop), XMP/EXPO memory profile active. Resizable BAR = free throughput (CPU sees full VRAM in one mapping).

### Thermal Monitoring & Throttling
- **Live throttle reasons**: nvidia-smi --query-gpu=timestamp,temperature.gpu,power.draw,clocks.current.graphics,clocks_throttle_reasons.active --format=csv -l 1
- **Throttle causes**: Temperature (>83 C), power limit, VRel/VOp (voltage), HW thermal slowdown, SW thermal slowdown, power brake
- **VRAM junction temperature** (critical with GDDR6X): nvidia-smi --query-gpu=temperature.memory --format=csv -l 2 — at >95 C swap pads, at 110 C P8 emergency brake
- **Manual fan curve (X11)**: nvidia-settings -a '[gpu:0]/GPUFanControlState=1' -a '[fan:0]/GPUTargetFanSpeed=70'
- **Headless**: nvidia-fan-control or fancontrol-gpu (Python), or lm-sensors + fancontrol for case fans
- **Undervolting (more tokens/watt)**: GUI via GreenWithEnvy or LACT; headless via the nvidia-pstate tool — typically RTX 4090 at 0.9 V / 2520 MHz: 320 W instead of 450 W with nearly identical token throughput

### Monitoring Stack
- **DCGM Exporter** (NVIDIA Data Center GPU Manager): docker run -d --gpus all --rm -p 9400:9400 nvcr.io/nvidia/k8s/dcgm-exporter:3.3.7-3.5.0-ubuntu22.04 — delivers >70 metrics/sec
- **MicroK8s + Prometheus + Grafana**: microk8s enable observability; Grafana dashboard ID 12239 (NVIDIA DCGM Exporter Dashboard)
- **Most important panels for LLM hosts**: SM utilization (target >70 %), VRAM usage (<90 %), power vs. limit (90-95 %), throttle reasons (constantly 0)
- **CLI tools**: nvtop (htop for GPUs), nvitop -m full (more details, process view), btop, watch -n 1 nvidia-smi
- **Prometheus alerts**: DCGM_FI_DEV_GPU_TEMP > 85 (2 min), DCGM_FI_DEV_THERMAL_VIOLATION > 0 (1 min), DCGM_FI_DEV_FB_USED/DCGM_FI_DEV_FB_TOTAL > 0.95 (5 min)
- **Loki + Tempo + OpenTelemetry**: helm install loki/tempo via grafana charts; vLLM with an OTEL endpoint (--otlp-traces-endpoint http://tempo:4317) provides per-request model, input/output tokens, TTFT, tokens/s, user-ID

### System Tuning for Inference
- **Hugepages**: echo always > /sys/kernel/mm/transparent_hugepage/enabled; statically via the kernel param default_hugepagesz=1G hugepagesz=1G hugepages=16 — eliminates TLB misses on every forward pass
- **Swap**: vm.swappiness=10 on inference hosts; numa_balancing=0 on dual socket (prevents page-migration stalls)
- **NVMe tuning**: I/O scheduler 'none' (echo none > /sys/block/nvme0n1/queue/scheduler), persistent via udev rule; blockdev --setra 4096 for sequential model loads
- **fio benchmark**: sudo fio --name=read --filename=/var/lib/llm/test --size=10G --bs=1M --rw=read --direct=1 --ioengine=libaio --iodepth=32
- **ZFS**: cap zfs_arc_max (otherwise it eats RAM that would be better off going to CUDA)
- **IOMMU/GPU passthrough**: BIOS VT-d/AMD-Vi on, kernel intel_iommu=on iommu=pt; vfio-pci ids=10de:... in /etc/modprobe.d/vfio.conf; every GPU must sit in its own IOMMU group

### Hardening
- **SSH**: root login off, key only, fail2ban; unattended-upgrades + automatic reboot at 03:00
- **UFW**: default deny incoming, expose the inference port only to your own subnet
- **Audit**: auditd on /dev/nvidia0 for GPU access logs
- **Kubernetes**: NetworkPolicy for vLLM (only the frontend may enter), PodSecurityStandard 'restricted', sealed secrets for API keys, RBAC per service account, Trivy scans in CI

### Important Pro Tools
- nvidia-smi, nvidia-persistenced (status, power, throttle reasons, persistence)
- nvtop / nvitop (live utilization, htop-style)
- DCGM Exporter (Prometheus metrics), Loki + Tempo (logs + traces)
- LACT, GreenWithEnvy (GUI for power/voltage/fan)
- gpu-burn (stress test), fio (NVMe benchmark)
- vLLM, Ollama, llama.cpp, TensorRT-LLM, Triton, TEI (inference engines)
- Unsloth, Axolotl (fine-tuning)
- MicroK8s, k3s, Longhorn, Rook-Ceph, cert-manager, sealed-secrets (lightweight Kubernetes stack)
- NVIDIA GPU Operator (driver + toolkit + device plugin in K8s)
- numactl, cpupower, tuned, fail2ban, auditd, trivy

## Other Distributions Worth Considering (beyond Ubuntu/Fedora/Gentoo)
- **Linux Mint** — Cinnamon, very Windows-like, ideal for classic switchers (https://linuxmint.com)
- **Pop!_OS** — by System76, Ubuntu-based, excellent for NVIDIA & developers (https://pop.system76.com)
- **Debian** — maximally stable, the mother of many distros (https://www.debian.org)
- **openSUSE Tumbleweed** — rolling release, YaST, automatic Btrfs snapshots (https://get.opensuse.org)
- **Arch Linux** — minimalist, famous wiki (https://archlinux.org)
- **EndeavourOS** — Arch with a graphical installer (https://endeavouros.com)
- **NixOS** — fully declarative configuration (https://nixos.org)

## Avoid Beginner Mistakes
- Never blindly run 'sudo rm -rf' — always check the path, prefer 'rm -i' or the trash (gio trash)
- Don't work as root by default — only use sudo when needed
- Install drivers via the package manager, not via vendor installers
- Don't enable too many repositories and PPAs at the same time
- Apply updates regularly, take a Timeshift snapshot beforehand
- Understand commands from the internet first, then run them
- Take a backup before risky actions
- Beginners should not start directly with Arch or Gentoo

## Linux Directory Structure (FHS)
- / — root directory
- /home/<user>/ — personal files and configuration
- /etc/ — system-wide configuration
- /var/ — variable data, logs, caches
- /usr/ — installed programs (e.g. /usr/bin)
- /opt/ — third-party software
- /tmp/ — temporary files
- /boot/ — kernel and bootloader
- /dev/ — device files (/dev/sda, /dev/nvme0n1)
- /proc/ — live data for processes and kernel
- /sys/ — hardware and driver info
- /mnt/, /media/ — external drives and USB

## Gaming on Linux (in detail)
- Steam runs natively; Proton (enable 'Steam Play for all titles') launches thousands of Windows games
- ProtonDB (protondb.com) shows compatibility per game
- Lutris for GOG/Battle.net/Epic, Heroic Games Launcher for Epic & GOG
- MangoHud (FPS overlay) and GameMode for performance optimization
- Bottles as a Wine frontend with profiles
- Anti-cheat status per game: areweanticheatyet.com
- Some online games (Valorant, Fortnite, many EA titles) block Linux

## Useful Links
- Ubuntu Download: https://ubuntu.com/download/desktop
- Fedora Download: https://fedoraproject.org/workstation/download
- Gentoo Handbook: https://wiki.gentoo.org/wiki/Handbook:Main_Page
- ubuntuusers.de wiki: https://wiki.ubuntuusers.de
- ProtonDB (game compatibility): https://www.protondb.com
- Balena Etcher (create USB sticks): https://etcher.balena.io
- OnlyOffice Desktop: https://www.onlyoffice.com/de/download-desktop.aspx
- Mailcow (Exchange alternative): https://mailcow.email
- DistroWatch: https://distrowatch.com
- ArchWiki (German): https://wiki.archlinux.org/title/Main_page_(Deutsch)

## Privacy & Security
- Privacy is a central topic in our Linux recommendations (Fedora for privacy-conscious users)
- Innovius relies on ShinrAI transparent semantic encryption for its products (https://innovius.ai/shinrai/)
- User keys remain exclusively with the user — no key escrow

## Part of the Innovius ecosystem
- Innovius (parent company): https://innovius.ai
- ShinrAI (encryption technology): https://innovius.ai/shinrai/
- ChtSafe (secure AI chat): https://chtsafe.com
- AICNCT (business AI connections): https://aicnct.com

## Imprint
Linuxeinsteiger.net is operated by Innovius GmbH.
Innovius develops AI-driven products and technologies: ShinrAI (transparent real-time semantic encryption), ChtSafe (secure AI chat) and AICNCT (business-to-AI networking platform).
Full imprint: https://innovius.ai/imprint.html
Website: https://innovius.ai