You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Production-ready scientific workflows for the Horus Runtime. Molecular dynamics, virtual screening, docking, LLM training, and more. Each workflow is a multi-stage pipeline that routes every stage to the right cluster type automatically: HPC, cloud, or local.
Why Pantheon? Stop writing glue code. Declare your pipeline once and Horus runs it across heterogeneous compute, streams logs live, retries failures, and produces typed artifacts you can trace.
Star the repo ⭐ if you build scientific pipelines, it helps more people find these workflows.
Each workflow directory contains:
README.md: purpose, pipeline, install steps, and configuration guide
workflow.yaml or run.py: a plain Horus workflow definition or Python workflow builder
scripts/: small stage scripts and helper code when needed
Getting Started
curl -LsSf https://astral.sh/uv/install.sh | sh
git clone https://github.com/temple-compute/pantheon
cd pantheon
cd workflows/drug-discovery/w01-boltz2-virtual-screening
uv sync
uv run horus run workflow.yaml
BioExcel workflows also require a conda-family tool (micromamba, mamba, or conda) on your PATH. Their executors build conda environments to provision GROMACS, AMBER, and related simulation tools. More on executors at docs.templecompute.com.
One measurable task per execution mechanism (subprocess, venv, container, in-process) as an acceptance harness for per-task resource measurement
Contributing
See CONTRIBUTING.md for how to add new workflows or improve existing ones.
About
Production-ready scientific workflows for the Horus Runtime: molecular dynamics, virtual screening, docking, LLM training and more. HPC pipelines that actually run at scale.