High-performance Serverless GPU task orchestration system
Waverless is a high-performance Serverless GPU task orchestration system designed for AI inference and training workloads, powered by WaveSpeedAI.
- 🚀 Pull-based Architecture - Workers actively pull tasks for better load balancing and fault tolerance
- 🔌 RunPod Compatible - Fully compatible with runpod-python SDK, no code modification needed
- ☸️ Kubernetes Native - Built-in K8s application management, supports deploying GPU workloads via API
- 📊 Multi-Endpoint Routing - Supports multiple independent task queues and worker pools
- 🌐 Web Management Interface - React-based modern UI for visual deployment and monitoring
- ⚡ Auto Scaling - Automatically adjusts worker count based on queue depth
# Clone repository
git clone https://github.com/wavespeedai/waverless.git
cd waverless
# Deploy complete environment
./deploy.sh install
# Access Web UI
kubectl port-forward -n wavespeed svc/waverless-web-svc 3000:80
# Visit http://localhost:3000 (default: admin/admin)For detailed deployment, configuration, and usage, see User Guide.
┌─────────────┐ ┌──────────────────┐ ┌─────────────┐
│ Client │ submit │ Waverless │ pull │ Worker │
│ ├────────>│ API Server │<────────┤ (RunPod) │
│ (V1 API) │ │ │ │ Endpoint: A │
└─────────────┘ │ - Task Queue │ └─────────────┘
│ - Worker Mgmt │
┌─────────────┐ API │ - K8s Manager │ ┌─────────────┐
│ Web UI │ Request │ │ pull │ Worker │
│(React+Nginx)├────────>│ Redis + MySQL │<────────┤ (RunPod) │
│ │ │ │ │ Endpoint: B │
└─────────────┘ └──────────────────┘ └─────────────┘
See System Architecture for detailed design.
Waverless provides RunPod-compatible V1/V2 APIs and K8s management APIs.
Quick Example:
# Submit task
curl -X POST http://localhost:8080/v1/wan22/run \
-H "Content-Type: application/json" \
-d '{"input": {"prompt": "a beautiful landscape"}}'
# Query status
curl http://localhost:8080/v1/status/{task_id}See User Guide for complete API documentation and usage examples.
Waverless documentation has been streamlined into 3 core documents:
| Document | Description | Audience |
|---|---|---|
| User Guide | Quick start, configuration, autoscaling, Web UI, and troubleshooting | Users, Operators |
| Architecture | System architecture, components, data models, statistics, and GPU tracking | Architects, System Designers |
| Developer Guide | Advanced topics, graceful shutdown, concurrency safety, task tracking internals | Developers, Contributors |
New Users: Start with User Guide → Quick Start section
Operators: User Guide → Configuration & Troubleshooting sections
Developers: Architecture → Developer Guide
Architects: Architecture for complete system design
MIT License