Open-source AI research platform

Meet Amadeus, your personal research assistant

Save papers from anywhere. Get AI-powered deep reading notes with diagrams and math. Track trends across arXiv, Semantic Scholar, and Twitter/X. Launch autonomous research workflows that turn intent into execution.

Get Started View on GitHub

What Amadeus Does

1Save papers from arXiv, OpenReview, or any PDF with the Chrome extension.
2AI performs multi-pass deep reading and generates structured notes.
3Track emerging papers from your favorite authors and topics.
4Run autonomous research workflows on remote compute via SSH.

Key Features

Everything a researcher needs in one workspace

🔍

One-Click Paper Saving

Chrome extension detects papers on arXiv, OpenReview, and any PDF page. One click saves the paper, downloads the PDF, and queues it for AI analysis.

🤖

AI-Powered Deep Reading

Multi-pass analysis generates comprehensive notes with Mermaid diagrams, KaTeX math, method breakdowns, and code repository analysis.

📈

Trend Tracking

Subscribe to Semantic Scholar authors, keyword queries, or Twitter/X profiles. Daily crawl surfaces new papers in a prioritized feed.

💻

ARIS Autonomous Workflows

Launch research runs that execute on remote SSH servers — literature reviews, experiment monitoring, paper writing, and full research pipelines.

🛠

VS Code Companion

In-editor control surface for your paper library, ARIS run management, and quick actions without leaving your coding environment.

📚

Full-Text Search & Library

Search papers by title, tags, content, and reading status. Mark papers as read/unread and filter your growing library.

See It In Action

Real screenshots from a live instance

Latest paper feed with personalized recommendations from multiple sources — **Latest Feed** — Personalized daily paper feed aggregated from Semantic Scholar, arXiv, Twitter/X, and RSS sources. Filter by source, sort by relevance, and save papers with one click.

Paper library with saved papers, tags, and action buttons — **Paper Library** — Browse your saved papers in a grid view with tags, code links, and quick actions for generating AI notes, reading, and downloading PDFs.

ARIS autonomous research workspace with project launcher — **ARIS Workspace** — Link local projects, configure SSH deployment targets, and launch autonomous research workflows. Monitor runs with real-time logs and follow-up actions.

Tracker admin panel showing configured paper sources — **Tracker Admin** — Configure paper tracking sources including Semantic Scholar authors, arXiv categories, Twitter/X profiles, and RSS feeds with customizable crawl intervals.

Getting Started

Up and running in five minutes

Clone the repository

git clone https://github.com/CurryTang/Amadeus.git
cd Amadeus

Run the interactive installer

The installer walks you through deployment mode selection (all-in-one or proxy + local device) and generates environment files.

./scripts/install.sh
# Generates: backend/.env.generated, frontend/.env.generated

Apply environment configs

cp backend/.env.generated backend/.env
cp frontend/.env.generated frontend/.env

Start the backend and frontend

cd backend && npm install && npm run dev
# In another terminal:
cd frontend && npm install && npm run dev

Install the Chrome extension

Open chrome://extensions/, enable Developer mode, click "Load unpacked", and select the chrome-extension/ folder.

Optional: VS Code Companion

cd vscode-extension
npm install && npm run compile
# Launch with Extension Development Host in VS Code

Key Configuration

After setup, configure these in backend/.env:

# Database (local SQLite or Turso cloud)
TURSO_DATABASE_URL=file:./local.db

# Object storage for PDFs (AWS S3, MinIO, or Aliyun OSS)
OBJECT_STORAGE_PROVIDER=aws-s3
OBJECT_STORAGE_BUCKET=your-bucket
OBJECT_STORAGE_ACCESS_KEY_ID=your-key
OBJECT_STORAGE_SECRET_ACCESS_KEY=your-secret

# Authentication
ADMIN_TOKEN=your-secret-admin-token

See the full configuration guide for all options.

Under the Hood

Paper processing pipeline

Save

Chrome extension captures paper metadata and PDF from arXiv, OpenReview, or any URL.

Queue

Paper is added to the processing queue and PDF is uploaded to object storage (S3/MinIO/OSS).

Analyze

AI performs 3-pass deep reading: structure scan, content understanding, and deep analysis with math and diagrams.

View

Rendered notes with Mermaid diagrams, KaTeX math, method breakdowns, and linked code repositories.

Deployment Architecture

Flexible deployment: all-in-one or proxy mode

Run everything on a single machine, or split into a cheap cloud proxy (nginx + FRP) and your always-on local device for heavy AI workloads. No cloud GPU costs — Claude and Gemini CLI run on your own hardware.

┌──────────┐     ┌──────────────────────┐     ┌─────────────────────────┐
│  Browser │────►│  Cloud Server (proxy) │────►│  Local Device (WSL/PC)   │
│          │     │  nginx + frps         │     │  PM2: API + Frontend     │
└──────────┘     └──────────────────────┘     │  SQLite, S3 storage      │
                                                 └─────────────────────────┘

Client Cloud Proxy FRP Bridge Local Executor

Tech Stack

Built with modern, proven tools

Frontend

React 18 + Next.js (standalone)
React Markdown + KaTeX + Mermaid
Service worker for offline caching

Backend

Node.js + Express
Turso / local SQLite
AWS S3 / MinIO / Aliyun OSS

AI

Claude Code CLI (code analysis)
Google Gemini CLI (paper analysis)
Multi-pass deep reading pipeline

Ready to streamline your research?

Amadeus is free, open-source, and self-hosted. Get started in minutes.

Clone on GitHub Deployment Modes