This repository is no longer actively maintained.

The llm-d-deployer repository previously hosted monolithic Helm charts for installing components of the llm-d ecosystem. As of 25-07-2025, we are deprecating this repository in favor of a more modular and composable approach provided by llm-d-infra.

The llm-d-infra repository contains pre-curated deployment examples of the llm-d ecosystem using Helmfile. It enables flexible configuration and composition of the following:

• Similar quickstart pattern to deployer for creating namespace, deploying metrics infrastructure, creating secret for the HF_TOKEN, etc.
  • See the installer script for more information.
• Gateway Deployment and configurations are based on the install of the llm-d-infra charts
• Installation of the new modelservice helm charts
  • The controller pattern for modelservice was deprecated, same with the idea of baseConfig presets and sampleApplications. Instead the modelservice charts focus on clearer deployments with modularity, allowing people to opt in or out of all components (epp, P/D deployments or leaderWorkerSets, inferencepool, inferencemodel, etc.). For more information check out their getting-started docs
• Compatibility with upstream GIE charts
  • Most of our examples feature this, but the simple example would be the easiest place to start to experience the composability with upstream GIE charts

You are encouraged to migrate any existing deployments to the examples provided in llm-d-infra, or use it as a reference to build your own Helmfile stacks.

• The charts in this repo are no longer updated.
• Issues and PRs will be closed with a deprecation notice.
• Historical references are preserved, but users should not use this repo for new deployments.
• The repo contents will remain if people want to use the existing monolithic installs.
  • For information on this refer to our old docs

If you're currently using the llm-d-deployer Helm charts and need help migrating to llm-d-infra, feel free to reach out via slack in the #sig-installation channel or file an issue in the llm-d-infra repo.

Thanks for supporting the llm-d project!