ARGOS: Autonomous Repository Guardian & Operational Swarm

ARGOS is an intelligent, agentic system designed to act as your autonomous Site Reliability Engineer (SRE). It uses a swarm of specialized AI agents to scan repositories, audit dependencies, execute safe sandboxed code, and generate real, functional security patches for your codebase.

🚀 Key Features

• 🕵️ Explorer: Scans local or remote (GitHub) repositories to understand codebase structure and language.
• 🛡️ Auditor: Runs security tools (Bandit for Python, npm audit for Node) to identify vulnerabilities.
• 🧠 Supervisor: Orchestrates the swarm, plans missions, and routes blocking issues to humans.
• 📦 Sandboxed Execution: All code analysis and testing happens inside secure E2B sandboxes, ensuring safety.
• 🛠️ Fix Generation: Automatically generates functional code fixes for Python and JavaScript vulnerabilities.
• ✅ Human-in-the-Loop: Critical security fixes require manual approval via a dedicated UI modal.
• 📥 Patch Downloads: Approved fixes are exported as standard .patch files for easy application.
• 📊 Live Telemetry: Real-time tracking of Token Usage, Latency, and Cost ($).

🚀 Core Missions

ARGOS provides specialized autonomous missions:

1. 🛡️ Security Scan:

   • Goal: Identify and fix security vulnerabilities (OWASP Top 10).
   • Tools: Bandit (Python), npm audit (Node.js).
   • Action: Generates functional code patches for review.

2. 📦 Dependency Audit:

   • Goal: Analyze project dependencies for outdated or risky packages.
   • Tools: pip-audit, npm audit.
   • Action: Reports dependency health and suggests upgrades.

3. 🐛 Bug Analysis:

   • Goal: Deep static analysis to find code quality issues and potential bugs.
   • Tools: Pylint, ESLint (planned).
   • Action: Provides a detailed report of code smells and refactoring opportunities.

4. ⚡ Custom Task:

   • Goal: Execute natural language requests (e.g., "Refactor the login class").
   • Action: The Agentic Swarm plans and executes the request in the sandbox.

🏗️ Architecture & Tech Stack

ARGOS is built on a battle-tested, modern stack designed for reliability and speed:

Backend

• Core: Python 3.10+, FastAPI (High-performance Async API)
• Orchestration: LangGraph (Stateful multi-agent workflows)
• AI/LLM: LangChain + Groq (OSS-120B) for <1s inference
• Runtime: E2B Code Interpreter (Secure, cloud-based sandboxing)
• Communication: WebSockets (Real-time agent streaming)

Frontend

• Framework: React 18 + Vite (Blazing fast build tool)
• Styling: TailwindCSS (Utility-first styling)
• Visualization: React Flow (Interactive Graph/DAG view of agent swarm)
• Terminal: xterm.js (Live streaming terminal logs)
• Icons: Lucide React

Infrastructure

• Telemetry: LangSmith (Traceability) + Custom Token Cost Tracking
• Security: Environment-isolated process execution

🛠️ Setup Instructions

Prerequisites

• Python 3.10+
• Node.js 16+
• Git

1. Clone the Repository

git clone https://github.com/vyom-modi/ARGOS.git
cd ARGOS

2. Backend Setup

cd backend
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

Configuration: Create a .env file in the backend/ directory:

GROQ_API_KEY=your_groq_key
E2B_API_KEY=your_e2b_key
LANGSMITH_API_KEY=your_langsmith_key (Optional)
LANGCHAIN_TRACING_V2=true (Optional)

3. Frontend Setup

cd ../frontend
npm install

🏃‍♂️ Running the System

1. Start Backend:

   # From project root
   python run_backend.py

   Backend runs on http://localhost:8000.

2. Start Frontend:

   # In a new terminal, from frontend/
   npm run dev

   Frontend accessible at http://localhost:5173.

🎮 Usage Guide

1. 🛡️ Security Scan

Goal: Find vulnerabilities and generate fixes.

1. Click New Mission -> Security Scan.
2. Paste a target URL (e.g., https://github.com/adeyosemanputra/pygoat).
3. Approve: When the Supervisor finds an issue, review the code diff and approve it.
4. Download: Click the generated "Patch Available" button to save the .patch file locally.

2. 📦 Dependency Audit

Goal: Check for outdated or vulnerable packages.

1. Click New Mission -> Dependency Audit.
2. Target: Repo URL (e.g., https://github.com/vuejs/vue).
3. Result: View realtime logs as the Auditor scans requirements.txt or package.json.

3. ⚡ Custom Task

Goal: Ask the Agent Swarm to do anything.

1. Click New Mission -> Custom Task.
2. Input: "Refactor the models.py file to use dataclasses" or "Explain the authentication flow in auth.py".
3. Result: The Explorer and Executor agents will research the codebase and execute your request.

🤝 Contributing

Contributions are welcome! Please start a discussion or open an issue before making major changes.

📄 License

MIT License