spotwarp 3.0.0

Zero-Downtime Spot GPU Failover Guard & AI Acceleration Utility for Vast.ai & RunPod

SpotWarp: Zero-Downtime Spot GPU Failover Guard v3.0

SpotWarp is a lightweight, 100% local Python daemon that protects your AI inference & PyTorch training workloads on cheap Spot GPUs (Vast.ai, RunPod, AWS Spot) with zero downtime and automatic data restoration.

🔒 100% Local Security Guarantee

• Zero-Key Leakage: Your VAST_API_KEY and RUNPOD_API_KEY stay 100% local on your host machine.
• Zero Admin Rights Needed: Runs entirely in unprivileged user space.
• Auditable Open Source: 100% open-source Python code.

🚀 Quick Start

1. Installation

pip install spotwarp

2. Run Guard Daemon (With Auto-Sync and Stateful Resume)

export VAST_API_KEY="your_vast_api_key_here"
spotwarp start --license-key YOUR_LICENSE_KEY --resume-cmd "python /workspace/train.py --resume"

⚡ How It Works

1. 0.05s Eviction Detection: Monitors local/remote instance health 24/7.
2. Instant Failover: Automatically selects the cheapest alternative Spot GPU on Vast.ai or RunPod upon eviction warning.
3. High-Speed Rsync Delta Sync: Periodically backs up remote workspace files locally and automatically restores them to the replacement container before connection verification.
4. Stateful Workload Resume: Launches the training command in the background (nohup) on the new node automatically.