If Lab Foundation is useful to you, please consider giving the project a star:
Documentation & community
Lab Foundation is the container image foundation behind LabNow. It packages operating-system dependencies, language runtimes, and workload tools into reusable images that platforms can launch as agent execution environments or offer as on-demand environments.
Use these images as building blocks: select a lightweight runtime, compose a purpose-built toolchain, and provide it to an agent or user when needed. This makes it easier to standardize environment templates and avoid rebuilding the same dependencies for every session.
- Runtime building blocks — start from a base image with common system utilities and a Python runtime, or use a richer image with more tools preinstalled.
- Repeatable environment templates — build images from shared layers and package profiles so agent sessions can start from consistent environments.
- Fit the workload — choose focused language and toolchain images, broader stacks, GPU variants, or PostgreSQL images with additional extensions.
- EaaS-ready substrate — use the published images as templates for an environment service that provisions environments on demand.
Scope: This repository builds and publishes container images. Agent orchestration, per-session lifecycle, authentication, and sandbox security policies (such as filesystem, network, and resource limits) are responsibilities of the runtime or platform that launches these images.
The following image names are built by this repository's workflows. On the main branch, the default image namespace is quay.io/labnow; images are also published or mirrored through the registries linked above.
Image availability and tags may vary by registry. Browse the current images on Quay.io or DockerHub.
| Environment building block | Images | Intended role |
|---|---|---|
| Foundation and language runtimes | atom, base, python-3.7, python-3.13, python-3.14, r-base, r-std, node, bun, jdk, go, rust, julia |
Minimal starting points for a custom agent runtime or environment template |
| Focused execution environments | py-data, py-std, py-chem, py-jdk |
Ready-made Python environments with different tool and dependency profiles |
| Broad toolchain environments | core, data-science-stack, full-stack |
Multi-language environments for agents or users that need a wider set of installed tools |
| Accelerated execution environments | torch, tf2, paddle-3.0, py-nlp, py-cv, core-cuda |
GPU/framework-oriented variants (see the GPU workflow) |
| Service and operations images | postgres-16, postgres-17, postgres-18, *-ext, docker-kit |
PostgreSQL environments with optional extensions, plus Docker Compose and image-sync utilities |
Start a container from the base runtime image and mount a workspace into it:
docker run --rm -it \
-v "$PWD":/workspace \
-w /workspace \
quay.io/labnow/base:latest \
bash
An agent runtime or environment service can use the same pattern to launch a selected image with its own workspace, resource limits, and lifecycle policy. Replace base with another image from the catalog to provide a different toolchain. For GPU images, configure the NVIDIA Container Toolkit on the host and pass --gpus all:
docker run --rm -it --gpus all quay.io/labnow/torch:latest bash
The image definitions are organized by layer: docker_atom/ (OS setup), docker_base/ (Python runtime), docker_core/ (language and workload packages), and additional directories for CUDA, PostgreSQL, and Docker utilities. For example, build a Python environment with database packages on top of the published base image:
docker build \
-f docker_core/core.Dockerfile \
--build-arg BASE_NAMESPACE=quay.io/labnow \
--build-arg ARG_PROFILE_PYTHON=base,database \
-t lab-foundation:custom \
docker_core/
Available profiles and package lists are defined in docker_core/work/. See the build workflows for the published image configurations.