dmr

dmr, published by Docker Inc., is the standalone Docker Model Runner, a tool designed to run AI models locally without requiring Docker Desktop or a running Docker Engine. It is distributed as a single binary that bundles two complementary components: an inference daemon, launched via the "dmr serve" command, and a full model management command-line interface used to communicate with that daemon. The CLI supports the familiar model lifecycle operations, including run, pull, ls, rm, and ps, allowing users to download models, execute them, list available models, remove them, and inspect running instances. This design makes dmr a practical option for developers and operators who want local AI inference capabilities in lightweight environments, on systems where installing the complete Docker Desktop suite is impractical or unnecessary, or in minimal setups where a container engine is not otherwise present. By eliminating the dependency on Docker's broader platform components, dmr reduces setup complexity and resource overhead while retaining a workflow that will feel familiar to anyone accustomed to Docker-style commands, since pulling, listing, and running models mirrors the container management experience. The software falls within the developer tools and local AI inference categories, sitting at the intersection of container tooling and machine learning model serving. Typical use cases include experimenting with AI models on a workstation, integrating local inference into development workflows, and managing a local library of models through a consistent command-line interface. The current version of dmr is 0.1.0, and it is the only version presently listed in the catalog, indicating an early-stage release. As version 0.1.0, the software represents the initial public iteration of this standalone runner, offering the core daemon and CLI functionality in a single self-contained package from Docker Inc.

Tags: