Off Grid AI

Off Grid AI is an offline AI suite that runs natively on a phone's or Mac's hardware.

Verified

Core facts

Read by a script from the official API, not rewritten by a model.

License
MIT
Languages
TypeScript
Maintainer
off-grid-ai
Homepage
getoffgridai.co/pro
Verified

What it is

Off Grid AI is an offline AI suite covering text generation, image generation, vision AI, voice transcription, tool calling and document analysis, running natively on a phone's or Mac's hardware. Voice input uses on-device Whisper speech-to-text. Off Grid AI Pro is an optional, additive tier. One license covers a phone and a Mac.

Main features

  1. Models that support function calling can use built-in tools: web search, calculator, date/time, device info and knowledge base search.

  2. Tool calling has an automatic tool loop with runaway prevention.

  3. The remote LLM servers feature connects to OpenAI-compatible servers on the local network, such as Ollama, LM Studio and LocalAI.

  4. Image generation uses on-device Stable Diffusion with real-time preview.

  5. Image generation is NPU-accelerated on Snapdragon and uses Core ML on iOS.

Getting started

  1. The latest APK is available from GitHub Releases.

  2. On macOS, the iOS App Store version runs natively on Apple Silicon Macs via Mac Catalyst / iPad compatibility.

  3. Building from source requires Node.js 20 or newer, JDK 17 and Android SDK 36 for Android, and Xcode 15 or newer for iOS.

  4. Tests run across 3 platforms on every PR.

From the README · bash
git clone https://github.com/off-grid-ai/OGAM.git
cd OGAM
npm install

# Android
cd android && ./gradlew clean && cd ..
npm run android

# iOS
cd ios && pod install && cd ..
npm run ios
From the README · bash
npm test              # Run all tests (Jest + Android + iOS)

Who it is for

Anonymous stats and crash reports are off unless turned on, and never include prompts, replies or files. In voice mode, Pro adds on-device Kokoro text-to-speech, so the assistant talks back and can be run hands-free. The project says the app detects what the device has and defaults to the fastest backend that works, and the choice can be overridden in Settings.

Remote LLM server connections discover models automatically, stream responses via SSE and store API keys securely in the system keychain. A project's knowledge base accepts uploads of PDFs and text documents. Knowledge base documents are chunked, embedded on-device with a bundled MiniLM model, retrieved via cosine similarity and stored locally in SQLite.

Voice input uses on-device Whisper speech-to-text. Voice input is started and stopped by tapping the microphone, or used in Auto and Hands-free modes. Document analysis attaches PDFs, code files, CSVs and more to conversations.

The text model automatically enhances image generation prompts. The model manager shows the resident models and the RAM each one costs, with a per-model eject. The download manager runs 3 downloads at once, and the rest queue first-in-first-out and show as Queued.

Downloads can be paused, resumed, retried and cancelled. Performance results vary by model size and quantization.

Sources

This page is written from the official sources below. Open them to check the original text.

Verified ·

This page is written by AI from the official sources above and checked automatically. This site has not tested it. Found a mistake? Tell us.