Skip to content

joym-gits/airoost

Repository files navigation

airoost

airoost

The VLC of AI. Run any AI model on your computer.
No internet. No account. No data leaving your machine.

Your AI. Your machine. Your rules.


What is airoost?

airoost is a desktop application that lets anyone run open-source AI models locally on their own computer. No terminal commands. No configuration files. No cloud dependency. Install it, open it, and chat with AI within 60 seconds.

Think of it as VLC, but for AI models — it plays any model you throw at it, runs entirely offline, and your data never leaves your device.


Features

Chat

Open the app and start talking to AI immediately. Multi-session conversations, streaming responses that appear word by word, and full conversation history saved locally on your machine.

Model Library

Browse 7 hand-picked models from Microsoft, Meta, Google, and Mistral. Or search 500,000+ models on Hugging Face and download any of them with one click. airoost checks your hardware and tells you which models will run well before you download.

Document Chat

Drag a PDF, Word document, or text file into airoost. Ask it questions. Get answers grounded in your document. The file is processed entirely on your machine — not a single byte is uploaded anywhere.

Knowledge Base

Point airoost at a folder full of documents. It reads, indexes, and understands them all. Then ask questions across your entire collection. Sources are cited in every answer so you know exactly where the information came from.

Personas

Choose how your AI behaves. Five built-in personalities (Legal Reviewer, Code Reviewer, Medical Summariser, Writing Editor, General Assistant) or create your own with custom instructions and an emoji avatar.

Prompt Library

19 ready-to-use prompts across Legal, Medical, Coding, Writing, Research, and Business. Search, favourite, and create your own. One click inserts any prompt into your chat.

Model Comparison

Test two AI models side by side with the same question. See which one responds better, faster, or more accurately. Pick the winner and continue chatting with it.

Export

Save any conversation as a PDF (beautifully formatted with the airoost header), Word document, or Markdown file. Copy individual responses as formatted snippets with model attribution.

Performance Dashboard

Live CPU and RAM monitoring while your AI is thinking. Usage statistics, model benchmarks, response time charts. And at the top, a number you can screenshot and share: 0 bytes sent to the cloud. Ever.

Themes

Dark mode, light mode, or match your system. Six accent colours. Three font sizes. Compact or comfortable layout. Make it yours.

Keyboard Shortcuts

Cmd+N (new chat), Cmd+K (search), Cmd+L (models), Cmd+R (regenerate), Cmd+? (show all shortcuts). Everything is keyboard-accessible.


Privacy

This is the core of airoost. Every feature is designed around one principle: your data stays on your machine.

What happens Where it happens
AI inference (thinking) Your CPU/GPU
Document processing Your local disk
Conversation storage Your local disk
Model downloads Direct from Hugging Face to your disk
Voice processing Your device (coming soon)
Telemetry None. Zero. Off by default.

There is no account to create. No API key. No server to connect to. No analytics. The app works fully offline after you download your first model.


Tech Stack

Component Technology
Desktop runtime Electron 37
Frontend React 18 + TypeScript + Tailwind CSS
AI engine llama.cpp via node-llama-cpp (built-in)
Embeddings all-MiniLM-L6-v2 (local, 384 dimensions)
State management Zustand
Charts Recharts
Document parsing pdf-parse + mammoth
Export Electron print-to-PDF + docx package
Build electron-vite + electron-builder
Auto-update electron-updater via GitHub Releases

System Requirements

Minimum Recommended
RAM 4 GB 8 GB+
Disk 1 GB (app) + model size 10 GB+
OS macOS 12+, Windows 10+, Ubuntu 20.04+ Latest
GPU Not required (CPU inference) Apple Silicon / NVIDIA (faster)

Development

git clone https://github.com/joym-gits/airoost.git
cd airoost
npm install
npm run dev

Build for distribution

npm run build:mac    # macOS .dmg
npm run build:win    # Windows .exe (NSIS)
npm run build:linux  # Linux .AppImage
npm run build:all    # All platforms

Project Structure

src/
  main/              # Electron main process (10 service modules)
    llmService.ts       AI engine, model loading, inference
    ragService.ts       Knowledge base indexing and search
    documentService.ts  PDF/DOCX text extraction
    exportService.ts    PDF, Word, Markdown export
    huggingfaceService.ts  HuggingFace API integration
    personaService.ts   Persona management
    promptLibraryService.ts  Prompt library
    statsService.ts     Usage tracking and benchmarks
    updater.ts          Auto-update system
    index.ts            IPC handlers and app lifecycle

  preload/            # Electron preload (IPC bridge)

  renderer/           # React frontend
    pages/              10 page components
    components/         7 shared components
    store/              Zustand state stores
    hooks/              Custom React hooks
    styles/             Tailwind CSS

build/                # App icons and entitlements
resources/models/     # Bundled GGUF model (for offline distribution)
scripts/              # Build helper scripts

License

Proprietary. All rights reserved.


airoost
Your AI. Your machine. Your rules.

About

VLC for AI models — download and run open-source AI models locally with zero setup

Resources

Stars

Watchers

Forks

Packages

 
 
 

Contributors