Metafold SDK for Python

Installation

pip install metafold

Quickstart

Create a metafold.MetafoldClient to interact with the API. To do so you will need an access token and the ID of a project to perform operations against.

from metafold import MetafoldClient

access_token = "..."
project_id = "123"

metafold = MetafoldClient(access_token, project_id)

With the client initialized, you can now start running jobs. For example, we can evaluate the surface area of a single gyroid unit cell:

job = metafold.jobs.run("evaluate_metrics", {
    "graph": {
        "operators": [
            {
                "type": "GenerateSamplePoints",
                "parameters": {
                    "size": [1.0, 1.0, 1.0],
                    "resolution": [128, 128, 128],
                },
            },
            {
                "type": "SampleSurfaceLattice",
                "parameters": {
                    "lattice_type": "Gyroid",
                    "scale": [1.0, 1.0, 1.0],
                },
            },
            {
                "type": "Redistance",
                "parameters": {
                    "size": [1.0, 1.0, 1.0],
                },
            },
            {
                "type": "Threshold",
                "parameters": {
                    "width": 0.04,
                },
            },
        ],
        "edges": [
            {"source": 0, "target": [1, "Points"]},
            {"source": 1, "target": [2, "Samples"]},
            {"source": 2, "target": [3, "Samples"]},
        ],
    },
    "point_source": 0,
})

print(job.meta["surface_area"])

Jobs are expected to complete within minutes (with some exceptions). You may increase the timeout when calling metafold.jobs.JobsEndpoint.run():

job = metafold.jobs.run("...", params, timeout=600)  # 10 mins

Many jobs require assets. You can upload an asset to your project directly from disk:

asset = metafold.assets.create("/path/to/mesh.obj")
print(asset)

You can list the assets in the project. The list should include the asset you just uploaded:

assets = metafold.assets.list()
for a in assets:
    print(a)

You can also filter the list for assets with a specific name. Filter queries use a simple syntax, see the API documentation for more details.

assets = metafold.assets.list(q="filename:mesh.obj")
print(assets[0])

Assets are referenced in job parameters by filename. For example:

job = metafold.jobs.run("sample_triangle_mesh", {
    "mesh_filename": asset.filename,  # "mesh.obj"
    "max_resolution": 128,
})

Most jobs produce at least one asset. You may access them from the returned metafold.jobs.Job object:

print(job.assets[0])

For more complete examples, please refer to the examples in the source code repo.

Contents:

Indices and tables