Automate Foundry toolkit extension and de novo design workflows.
npx skills add https://github.com/LiorZ/protein-design-skills --skill foundry-liorz
Please help me install this Agent Skill. Skill: foundry Source: https://github.com/LiorZ/protein-design-skills/tree/main/skills/foundry Command: npx skills add https://github.com/LiorZ/protein-design-skills --skill foundry-liorz
Foundry provides a unified infrastructure to run, extend, and manage three model families (RFD3, RF3, MPNN variants) for de novo design and structure prediction, reducing setup overhead and enabling reproducible experiments.
Install a checkpoint and run a quick design through the rfd3, rf3, and mpnn workflows.
High-intent search queries and answers about installing and using this skill.
You register new MPNN checkpoints in a de novo design workflow by using a registry-based discovery system that enforces metadata-driven extension points, allowing teams to plug in new architectures without modifying existing tooling. This ensures scalable collaboration and reproducible experiments.
Yes, you can run RFD3 and RF3 inference campaigns from a centralized Python API. Foundry provides unified CLI and Python interfaces to install checkpoint data, run structure prediction, and manage end-to-end design workflows across multiple model families.
Hydra config wiring is a configuration management mechanism that connects custom inference engines to Foundry workflows. It enforces metadata-driven extension points, enabling researchers to integrate proprietary models into RFD3, RF3, and MPNN variants for scalable collaboration.
Yes, Foundry supports integrating custom inference engines for structure prediction. It provides dedicated extension pathways that allow teams to wire in proprietary models or new architectures, reducing setup overhead while maintaining reproducible experiment infrastructure.
The best way to manage checkpoints across RFD3, RF3, and MPNN variants is through a centralized registry-based discovery system. Foundry enforces metadata-driven checkpoint registration, enabling reproducible experiments and scalable collaboration across all three model families.
NVIDIA-BioNeMoOrchestrate end-to-end protein and ligand design pipelines with AF2, RF3, and MPNN backends.
zongtingweiGenerate flow-based protein backbones with fold conditioning for sequence design.
surus-latIntegrate AI model providers into the Benchy benchmarking framework via YAML configuration.
tizeeCreate and run standalone Python scripts with inline dependency metadata using uv.
