CloudCull: The Autonomous Multi-Cloud GPU Sniper

CloudCull is an "Investor-Grade" autonomous governance framework designed to detect and eliminate GPU waste across AWS, Azure, and Google Cloud Platform. By 2026, it is the standard for multi-cloud cost optimization.

🔴 VIEW LIVE DEMO DASHBOARD

---

💰 The 2026 Problem: "GPU Bankruptcy"

Startups and AI companies lose thousands of dollars every month because expensive GPU instances are left running idle. Manual tagging and spreadsheets are not enough to stop this bleeding.

🔫 The Solution: CloudCull

CloudCull is not a dashboard; it is an Execution-First Sniper Agent. It proactively scans your multi-cloud environment, uses Multi-Model Intelligence (Claude/Gemini/Llama) to classify instance state, and provides a Kill-Switch to stop waste immediately.

---

🏛️ Architecture: "The Sniper Pattern"

CloudCull follows a robust, CLI-first automation flow designed for deep integration into DevOps pipelines.

graph LR
    Trigger["Cron / GitHub Action"] --> Probe["Probe: Multi-Cloud SDKs"]
    Probe --> Analyzer["Analyzer: Multi-Model AI"]
    Analyzer -- "Decision: Zombie Identified" --> UI["UI: Approval Notification"]
    UI -- "Approve" --> Execute["Execution: Boto3/SDK Terminate"]

Loading

🏗️ Key Features

• � High-Fidelity Brain: Pluggable AI (Claude 3.5, Gemini 1.5, Llama 3) for intelligent classification.
• 📡 Sniper Console: A technical Vite + React dashboard with AI Reasoning Callouts, Live Terminal Logs, and One-Tap Kill Actions.
• 👤 Identity Layer: Finds exactly who launched the instance for high-stakes accountability.
• 🛠️ IaC-Driven Remediation: Generates terraform state rm plans instead of raw, risky deletions.

---

🛠️ Usage

Note

CloudCull is a CLI-First Tool. The dashboard is a passive visualization layer.

1. Installation (via uv)

git clone https://github.com/daretechie/cloudcull.git
cd cloudcull
uv sync

2. Run a Demonstration (Simulated Mode)

uv run python main.py --simulated --dry-run

3. Execution (ActiveOps)

To run a real-world audit and trigger the automated remediation bundle:

uv run python main.py --region us-east-1 --active-ops

Caution

--active-ops will generate remediation.sh and prompt for execution. Use with high-stakes environments after dry-run verification.

---

🐳 Deployment & Docker

CloudCull is container-ready for consistent execution.

Running with Docker

# Build & Run
docker build -t cloudcull .
docker run --env-file .env cloudcull --simulated --dry-run

Running with Docker Compose

docker-compose up

🌐 Dashboard (GitHub Pages)

To enable the live dashboard, you must manually activate GitHub Pages in your repository settings:

1. Go to Settings > Pages.
2. Under Build and deployment > Source, select GitHub Actions.

📄 Documentation

• Architecture & System Design
• Design Principles & Rationale (Why we chose CLI-First & GitOps)
• Deployment Guide
• Developer Setup
• Dashboard Guide
• Operations Guide

📄 License

This project is licensed under the MIT License.