OpenMindAI

OpenMindAI is a local-first, offline-first AI desktop application. It downloads and runs an open-weight language model directly on your computer — no cloud account, no API key, no subscription. After first-run setup, it works fully offline.

Official repository: https://github.com/smshagor-dev/OpenMindAI

Quick Download

Platform	File	Status
Windows — Recommended	OpenMindAI-Setup-v1.0.1-x64.exe	Tested / Stable
Windows — Git Bootstrap	OpenMindAI-Setup.bat	Tested / Stable
Linux	openmindai-setup.sh	Tested / Stable
macOS	OpenMindAI-Setup.command	Tested / Stable

The Windows .exe installer is the simplest path once it's published as part of the v1.0.1 GitHub Release. The bootstrap scripts (.bat/.sh/ .command) work today — they clone the official source and build/launch OpenMindAI directly, and will automatically switch to downloading a prebuilt release once one is published. See Support Matrix below for validation status.

Windows Installation

1. Download OpenMindAI-Setup-v1.0.1-x64.exe.
2. Run it. It's currently unsigned (no code-signing certificate yet), so Windows SmartScreen will warn — click "More info" → "Run anyway" if you trust the source, or verify it against the published SHA256SUMS.txt first.
3. Choose your AI storage location and a local profile name.
4. OpenMindAI configures its local database, detects your hardware, downloads the matching AI engine, then downloads and verifies the AI model (Qwen3 4B).
5. OpenMindAI opens. Chat now works fully offline.

Windows Git Bootstrap

1. Download OpenMindAI-Setup.bat.
2. Put it in the folder/drive you want OpenMindAI's source and build in — any drive, including a USB or external SSD, with or without spaces in the path.
3. Double-click it.
4. It checks for Git (installing it via winget if missing and available), then clones the official repository: https://github.com/smshagor-dev/OpenMindAI.git
5. It looks for a compatible prebuilt release; if none exists yet, it builds OpenMindAI from source automatically (requires Node.js and Rust/Cargo — it tells you clearly if either is missing, with where to get it, rather than failing silently).
6. OpenMindAI launches and first-run setup begins.
7. Running the .bat again later reuses the existing installation — it does not re-clone, re-install dependencies, or rebuild unless something actually changed.

You never need to type a Git command yourself.

Linux Installation

chmod +x openmindai-setup.sh
./openmindai-setup.sh

It detects your package manager (apt, dnf, yum, pacman, zypper, or apk), installs Git if missing (asking for sudo only when actually needed), clones the official repository, and either downloads a compatible prebuilt release or builds from source. x86_64 is the only architecture currently targeted — ARM64 is not yet built or tested. This script has been written and reviewed but not yet run-tested on a real Linux machine; treat it as beta until that's done.

macOS Installation

Double-click OpenMindAI-Setup.command, or from Terminal:

chmod +x OpenMindAI-Setup.command
./OpenMindAI-Setup.command

Detects Intel vs. Apple Silicon automatically. Uses Apple's own Xcode Command Line Tools for Git if it's missing (or Homebrew, if you already have it installed — OpenMindAI never installs Homebrew for you without asking). A downloaded/built OpenMindAI app is not currently signed or notarized, so Gatekeeper will warn on first open; you'll need to allow it via System Settings → Privacy & Security. This script has been written and reviewed but not yet run-tested on a real Mac; treat it as experimental until that's done.

First Launch

Every platform follows the same setup: choose AI storage → create a local profile → OpenMindAI detects your hardware → downloads the AI engine → downloads and verifies the AI model (Qwen3 4B, Q4_K_M) → ready to chat.

Portable Mode

If OpenMindAI (or one of the bootstrap scripts) finds an openmindai.marker file near its own location, it treats that folder as its data root instead of asking where to store data. This makes the whole install portable — models, runtime, and database move with the folder. If a portable install moves to a different drive letter (e.g. G:\OpenMindAI becomes H:\OpenMindAI after being plugged into a different port), it keeps working: paths are resolved relative to the marker's location, not a fixed drive letter.

Offline Mode

Needs internet: first-run setup (downloading the AI engine and model), the Git bootstrap's source clone/update step, and update checks.

Works fully offline: chatting, conversation history, local settings, GPU/CPU inference, and the Maintenance Center's diagnostics/repair/backup tools — once setup has completed once.

If a bootstrap script or OpenMindAI itself can't reach the internet on a later run, it skips the update/clone step entirely and launches your existing installation immediately. It never refuses to start just because GitHub or an update server is unreachable.

Updates

Application updates and model updates are independent, and both are notify-first, not silent:

• Application updates — OpenMindAI checks for a newer release in the background and lets you download, verify, and install it from inside the app, with an explicit restart step you control.
• Model updates — checked separately; large model downloads are never started automatically unless you opt in.

The bootstrap scripts (.bat/.sh/.command) intentionally don't implement their own separate update-checking logic — they only handle getting OpenMindAI installed and launched. All update checking happens inside the app itself, so there's one update system, not two competing ones.

Storage

All of OpenMindAI's data — models, runtime, database, chat history, cache, generated files, workspaces, knowledge, backups, logs — lives under a folder you choose during setup, referred to as OPENMINDAI_ROOT. This is kept separate from wherever the application itself or its source code is installed.

Model

OpenMindAI ships against Qwen3 4B (Q4_K_M) by default. Model weights are never stored in this Git repository — they're downloaded and checksum-verified during setup.

Support Matrix

Platform	Status
Windows x64	Tested / Stable
Linux x64	Implemented / Not yet run-tested on real hardware
macOS (Intel / Apple Silicon)	Implemented / Not yet run-tested on real hardware
Linux ARM64	Not implemented

Troubleshooting

• Git not found / can't install it automatically — install it yourself (Windows: https://git-scm.com/download/win · Linux: your distro's package manager · macOS: xcode-select --install), then run setup again.
• No internet on first run — setup can't complete without it (the AI engine and model have to be downloaded once). Connect and try again.
• Model download interrupted — resumes automatically from where it left off; no need to start over.
• No compatible AI runtime / GPU not detected — OpenMindAI falls back automatically (GPU backend → Vulkan → CPU); chat still works, just slower without GPU acceleration.
• Storage location unreachable (e.g. an external drive unplugged) — OpenMindAI won't silently create a new, different storage location; it tells you the original one is missing so you don't lose track of your data.
• Windows SmartScreen warning — expected for the unsigned installer; verify against SHA256SUMS.txt instead.

Build From Source

For developers only — normal users should use the downloads above.

git clone https://github.com/smshagor-dev/OpenMindAI.git
cd OpenMindAI
npm install
npm run tauri dev

Requires Node.js/npm and Rust/Cargo (Windows also needs MSVC Build Tools). Production build: npm run build && npm run tauri -- build.

License

Apache License 2.0. Third-party attributions: THIRD_PARTY_NOTICES.txt.