Skip to content

Albe83/llm-gw

Repository files navigation

llm-gw

LLM Gateway is a small container image that bundles the LiteLLM Proxy. It ships an opinionated entrypoint and a Helm chart to ease running the proxy locally or on Kubernetes.

Features

  • Reproducible build: the Dockerfile pins the Oracle Linux base image by digest and installs litellm with a specific version.
  • Minimal runtime: the container creates a dedicated non‑root user and copies only the entrypoint script and its configuration.
  • Configurable: point the LITELLM_CONFIG environment variable to a YAML file containing your model configuration. A placeholder file is provided in the repository.
  • Helm chart: under chart/llm-gw you will find a chart ready to deploy the container with common settings.

Building the image

docker build -t llm-gw .

Running

Supply a configuration file and map the service port (default 4000):

docker run --rm -p 4000:4000 \
  -v $(pwd)/litellm_config.yaml:/litellm_config.yaml:ro \
  llm-gw

The entrypoint prints basic information and forwards all arguments to litellm.

Helm installation

helm install my-gateway ./chart/llm-gw

Adjust values in chart/llm-gw/values.yaml to match your environment.

Tests

A small test script verifies the behaviour of entrypoint.sh:

./tests/test_entrypoint.sh

License

This project is licensed under the MIT License. See LICENSE for details.

About

Artificial Intelligence Gateway

Resources

License

Stars

Watchers

Forks

Packages

No packages published