voltsdk 2.2.4

Python SDK for the Volt scientific computing platform

voltsdk

Python SDK for the Volt scientific computing platform.

Installation

pip install voltsdk

pip will select the platform wheel that matches the host operating system and architecture. VoltSDK wheels bundle the native Volt runtime under voltsdk/native/{bin,lib,share}, so users do not need to install the shipped TBB, hwloc, and related shared libraries manually for the bundled binaries.

Supported wheel targets published by CI:

• linux_x86_64 built on ubuntu-24.04
• macosx_14_0_arm64
• win_amd64

Linux wheels target the Ubuntu 24.04 build baseline. Older glibc releases are not supported; document that requirement in downstream READMEs when you ship Linux-native workflows.

Optional extras:

pip install "voltsdk[visualization]"
pip install "voltsdk[notebook]"

Authenticated client

from voltsdk import VoltClient

client = VoltClient(
    secret_key="your-secret-key",
    base_url="https://api.example.com",
)

Open In Volt

Use the Volt web canvas as the viewer for local GLBs, GLB sequences, or existing Volt resources:

from voltsdk import open_in_volt

open_in_volt('out/frame.glb')
open_in_volt({
    0: 'out/frame-0000.glb',
    5000: 'out/frame-5000.glb',
    10000: 'out/frame-10000.glb',
})

Resource helpers open the exact canvas route:

trajectory.open_in_volt(timestep=5000)
frame.open_in_volt()
analysis.open_in_volt(timestep=5000)
exposure.open_in_volt(5000)
run.open_in_volt('dislocations')

For plugin runs, open_in_volt(...) auto-assembles supported .json and .msgpack exporter payloads into a sibling .glb when needed.

For local files, VoltSDK serves them directly to /canvas/glb without creating fake analyses. In a Volt notebook it uses the Jupyter proxy; on a local machine it starts a tiny background file server. The browser still needs an authenticated Volt session. By default the viewer opens against https://app.voltcloud.dev; override it with volt_url=... or VOLT_APP_URL if needed.

SpatialAssembler

Volt exporter payloads can also be rendered locally from Python:

from voltsdk import SpatialAssembler, open_in_volt

assembler = SpatialAssembler()
glb_path = assembler.dislocations_glb(
    'output/ptm-dxa_dislocations.json',
    output_path='output/ptm-dxa_dislocations.glb',
)

open_in_volt(glb_path)

The Python API accepts either the full artifact file (.json or .msgpack) or the raw nested payload from export.AtomisticExporter, export.MeshExporter, or export.DislocationExporter.

You can also let a PluginRun do the conversion directly:

viewer_url = dxa_run.open_in_volt('dislocations', open_browser=False)
mesh_glb = dxa_run.glb('defect_mesh')

Plugin hub

VoltSDK ships a Hugging Face-style plugin hub. Plugins live in a static registry; bundles are downloaded and cached on first use:

from voltsdk import PluginHub

hub = PluginHub()                          # default registry, ~/.cache/volt
print(hub.list())                          # publisher-qualified marketplace listing

ptm = hub.get("voltlabs@polyhedral-template-matching")
result = ptm.run(
    "frame.dump",
    output_base="out/frame",
    crystal_structure="FCC",
    rmsd=0.1,
)
print(result.path("annotated.dump"))

The same hub is exposed on an authenticated client via client.plugins.

When a plugin subprocess is launched, VoltSDK also prepends its bundled native runtime directories to PATH, LD_LIBRARY_PATH, and DYLD_LIBRARY_PATH. This lets downloaded plugin bundles reuse the libraries installed with the wheel.

Configuration

Variable	Purpose	Default
VOLT_PLUGIN_REGISTRY	Registry base URL	https://server.voltcloud.dev/plugin-registry
VOLT_CACHE_DIR	Local plugin cache	$XDG_CACHE_HOME/volt

Pinning versions

Plugin identifiers are always publisher-qualified: publisher@plugin.

hub.get("voltlabs@opendxa", "1.0.0")  # explicit version
hub["voltlabs@opendxa"]               # shorthand for the latest version
hub.install("voltlabs@opendxa")       # pre-download the latest bundle
hub.uninstall("voltlabs@opendxa")     # drop every cached version

Registry layout

The hub expects a static index plus per-platform bundles:

<registry>/index.json
<registry>/<key>/<version>/<os>-<arch>.tar.zst

index.json example:

{
  "plugins": {
    "voltlabs": {
      "opendxa": {
        "publisher": "voltlabs",
        "latest": "1.0.0",
        "versions": {
            "1.0.0": {
              "linux-x86_64": {
                "url": "opendxa/1.0.0/linux-x86_64.tar.zst"
              }
            }
          }
      }
    }
  }
}

Each bundle is a flat archive containing plugin.json, bin/<binary>, and optional lib/ and scripts/ directories.