AIOS：本地AI研究的统一操作系统

AIOS is an Electron-based desktop application that integrates multi-AI research tool DeepDive and screenshot analysis tool Snipping Vault into a unified local AI work platform. Key features include local data storage, multi-provider AI integration, and privacy-first design. Original author/maintainer: jfoote22; Source platform: GitHub; Original title: AIOS - Artificial Intelligence Operating System; Original link: https://github.com/jfoote22/AIOS; Release time: May 2026.

AIOS is an ongoing unified local desktop app in Phase 2 development, using tech stack Electron + Vite + React 19 + TypeScript + Tailwind v4. Its core vision is to create a fully local AI workspace where users can conduct deep AI research, capture and organize screenshots, build personal knowledge bases without worrying about data privacy—all data stored locally, no account needed, no tracking.

DeepDive: Multi-AI Research Dialogue System

Features: Customizable panel layout; threaded multi-AI conversations; smart right-click menu (get details, simplify, generate examples, find links, search videos, ask questions, save to Vault); learning snippet capture; local SQLite storage. AI responses via local Express server, API keys read from Electron's safeStorage.

Snipping Vault: Screenshot & AI Analysis Tool

Features: System tray integration; global hotkey (Ctrl+Shift+S); screenshot overlay; Gemini intelligent analysis; Vault grid view; Ask-the-Vault dialogue; tag & chunk management. Deeply integrated with DeepDive—saved snippets carry originThreadId for bidirectional links.

Local-First Architecture

• IndexedDB for fragments, sessions, chat history;
• SQLite (better-sqlite3) for DeepDive sessions;
• Electron safeStorage for encrypted key management (Windows uses DPAPI);
• Local Express server (127.0.0.1 random port) simulating Vercel AI SDK endpoints.

Multi-Provider AI Integration

Supports OpenAI (DeepDive), Anthropic (Claude), Grok (xAI), Gemini (screenshot analysis/embeddings). Each provider's API key stored independently, activating corresponding models when added.

Security Storage Mechanism

1. safeStorage encryption; 2. On-demand key reading (not long驻留 in memory);3. Renderer process can't access keys directly (via IPC);4. Windows key file location: %APPDATA%/AIOS/provider-keys.json.

Completed (Phase 2)

• DeepDive module full移植;
• Snipping Vault function implementation;
• Vault & DeepDive cross-linking;
• Local Express API server;
• Multi-provider key management.

Planned (Phase3)

• Second Brain: Unified RAG system integrating fragments & dialogue threads;
• Hermes integration (pending);
• AI subscription dashboard (manage subscriptions/usage);
• Extra API endpoints (learning tool generation, voice transcription, image generation);
• Optional Firebase backup (encrypted for cloud sync users);
• SQLite migration from IndexedDB (support FTS5 full-text search).

AIOS is suitable for:

• AI Researchers: Compare multiple AI models' responses, save research results;
• Knowledge Workers: Capture & analyze screen info, organize visual data;
• Privacy-Sensitive Users: Prefer local tools without cloud data upload;
• Multi-AI Service Users: Manage multiple AI services in one interface, avoid switching between web pages.

AIOS reflects key trends:

1. Local AI Revival: Local AI tools become feasible with stronger models & lower running costs;
2. Privacy-First Design: Data privacy as core constraint from architecture level;
3. Multi-Model Strategy: Not tied to single provider, users choose models for tasks;
4. Closed-Loop Knowledge Management: From info capture (screenshot) → deep research (AI dialogue) → knowledge organization (Vault).

AIOS represents an important direction in AI tools—keeping powerful features while returning data control to users. Its local-first architecture, multi-provider support, and integrated workflow offer an attractive option for privacy-conscious users. With Phase3 features, AIOS is expected to be a strong competitor in personal AI workspaces.