LayerSmith
An open-source, self-hosted container image build tool designed for air-gapped network environments. It provides a web UI frontend, supports Docker and Podman, and offers features such as pinned base images, build metadata recording, and exportable bundles to ensure image reproducibility and trusted delivery in offline settings.
Timeline (last 90 days)
LayerSmith is an open-source project initiated by developer r0lfi that provides a self-hostable Web UI for building container images based on Docker or Podman and supporting migration across isolated environments.
LayerSmith allows users to select a Linux distribution and specify the intended use of the image (development environment, Linux operations, network tools, Ansible, Kubernetes, OpenShift, or custom configuration), and displays the generated Containerfile before the formal build.
LayerSmith supports users in manually editing the generated Containerfile, importing an existing Dockerfile, or appending their own packages, files, and scripts.
LayerSmith supports migration to isolated environments by pinning the base image, recording build metadata, and exporting a bundle containing the image, checksums, and installation instructions.
LayerSmith recently added a set of image configurations targeting large model training and fine-tuning, covering LoRA/QLoRA, advanced PyTorch training, and LLaMA-Factory.
LayerSmith's LLM training configurations use hash-locked dependencies and run an offline check that includes a small CPU training test after the build is complete.
LayerSmith's training configurations carry model weights and datasets separately rather than packaging them into the image, in order to prevent the image size from growing out of control.
LayerSmith is positioned to serve operations and infrastructure teams in sectors such as defense, finance, government affairs, and industrial control, which cannot freely connect to the internet yet need to frequently build and maintain container images.
All Facts (8)
LayerSmith is an open-source project initiated by developer r0lfi that provides a self-hostable Web UI for building container images based on Docker or Podman and supporting migration across isolated environments.
50%UnverifiedLayerSmith allows users to select a Linux distribution and specify the intended use of the image (development environment, Linux operations, network tools, Ansible, Kubernetes, OpenShift, or custom configuration), and displays the generated Containerfile before the formal build.
50%UnverifiedLayerSmith supports users in manually editing the generated Containerfile, importing an existing Dockerfile, or appending their own packages, files, and scripts.
50%UnverifiedLayerSmith supports migration to isolated environments by pinning the base image, recording build metadata, and exporting a bundle containing the image, checksums, and installation instructions.
50%UnverifiedLayerSmith is positioned to serve operations and infrastructure teams in sectors such as defense, finance, government affairs, and industrial control, which cannot freely connect to the internet yet need to frequently build and maintain container images.
50%UnverifiedLayerSmith's LLM training configurations use hash-locked dependencies and run an offline check that includes a small CPU training test after the build is complete.
50%UnverifiedLayerSmith's training configurations carry model weights and datasets separately rather than packaging them into the image, in order to prevent the image size from growing out of control.
50%UnverifiedLayerSmith recently added a set of image configurations targeting large model training and fine-tuning, covering LoRA/QLoRA, advanced PyTorch training, and LLaMA-Factory.
50%