Guides, scripts, and documentation for deploying, configuring, and testing model serving with llm-d.
This repository is a collection of practical resources for working with llm-d clusters. It covers deployment workflows, feature validation, and operational recipes that go beyond the core project documentation.
Test and validation scripts for llm-d features.
| Script | Description |
|---|---|
| test-priority-holdback.sh | Validates the priority holdback flow control feature against a live cluster. Creates InferenceObjectives at different priorities, drives load, and verifies that lower-priority traffic is gated before higher-priority traffic. |
kubectlconfigured and pointing at a Kubernetes cluster with llm-d installedcurlandjqfor HTTP and JSON operations- A running llm-d deployment (EPP, gateway, model servers)
Clone the repo:
git clone https://github.com/llm-d/llm-d-cookbook.git
cd llm-d-cookbookRun a script:
./scripts/test-priority-holdback.sh <NAMESPACE> [GATEWAY_URL] [METRICS_URL] [INFERENCE_POOL]Contributions are welcome. Each addition should be self-contained with clear prerequisites and usage instructions.