Overview
Lens is a web console for running large language models on Kubernetes with llm-d. Use it to deploy a model, publish an API endpoint, and measure its performance.
First time using Lens?
Follow Quickstart to install Lens and get your first model response. The normal path is:
- Add a cluster.
- Prepare storage and download model weights.
- Deploy a model and wait for it to be ready.
- Publish a service and create an API key.
- Run an evaluation.
Find your task
| I want to... | Open |
|---|---|
| Choose and deploy a model | Model Market |
| Check deployment status or read pod logs | Deployments |
| Call a model or manage API keys | Model services |
| Compare latency and throughput | Evaluation |
| Replay a request trace | Simulation |
| Connect a cluster or maintain its nodes | Clusters |
| Create or remove a storage volume | Storage |
| Download models or free cache space | Model cache |
| Open monitoring dashboards | Observability |
| Connect an external AI provider | External providers |
| Ask the assistant about resources | Lens Assistant |
| Manage users and access | Administration |
Get help
For a failed operation, read its error message and logs, then check Troubleshooting. If a resource or action is missing, ask an administrator to check your permissions.
For automation, use the API reference. Implementation details are in Architecture, separate from the operating instructions.