📚 Hands-on learning powered by vLLM Playground
A comprehensive workshop for learning vLLM — the high-performance LLM inference engine — through practical, hands-on exercises.
https://micytao.github.io/vllm-workshop
| Module | Topic | Key Skills |
|---|---|---|
| Module 1 | Getting Started | Deploy vLLM servers, use chat interface |
| Module 2 | Structured Outputs | JSON Schema, Regex, Grammar constraints |
| Module 3 | Tool Calling | Function calling with LLMs |
| Module 4 | MCP Integration | Agentic AI with human-in-the-loop |
| Module 5 | Performance Testing | Benchmarking with GuideLLM |
- Python 3.10+
- GPU with CUDA support (recommended) or CPU
- Podman or Docker
pip install vllm-playground
vllm-playground pull
vllm-playgroundThen open http://localhost:7860 and follow the workshop modules!
To run the documentation site locally:
# Install MkDocs
pip install mkdocs-material mkdocs-glightbox
# Serve locally
mkdocs serve
# Open http://localhost:8000docs/
├── index.md # Welcome page
├── overview.md # ACME Corporation narrative
├── details.md # Requirements & timing
├── workshop/
│ ├── module-01-getting-started.md
│ ├── module-02-structured-outputs.md
│ ├── module-03-tool-calling.md
│ ├── module-04-mcp-integration.md
│ └── module-05-benchmarking.md
└── conclusion.md # Summary & next steps
- vLLM Playground - The tool used in this workshop
- vLLM - High-throughput LLM serving engine
- GuideLLM - Performance benchmarking
Contributions welcome! Please feel free to submit issues and pull requests.
Apache 2.0 License
Built with ❤️ for the vLLM community