fourtheorem-tempo-client

Python client for the Tempo public API, generated from its OpenAPI specification. Tempo is a high-throughput compute platform for massive-scale modelling workloads.

Status: early development. The API surface is small and evolving, and authentication is not implemented yet — use Client; the generated AuthenticatedClient is boilerplate for a Bearer scheme the API does not support yet. Expect breaking changes until 1.0.

Install

pip install fourtheorem-tempo-client
# or
uv add fourtheorem-tempo-client

Requires Python 3.11+. Fully typed (py.typed); httpx-based, with sync and async variants for every operation.

Quickstart

from tempo_client import Client
from tempo_client.api.public import get_workload_status, health_check, submit_workload
from tempo_client.models import WorkloadSubmission

client = Client(base_url="https://your-tempo-host/api")

health_check.sync(client=client)

submitted = submit_workload.sync(
    client=client,
    body=WorkloadSubmission.from_dict(
        {
            "name": "demo",
            "jobs": [{"id": "j-1", "name": "Job 1", "work_type": "default", "predecessors": []}],
        }
    ),
)

status = get_workload_status.sync(client=client, workload_id=submitted.workload_id)
print(status.state)

Every operation also has an async variant:

status = await get_workload_status.asyncio(client=client, workload_id=submitted.workload_id)

What's covered

Workload submission (inline or via presigned uploads for large job graphs), workload lifecycle (status, cancel, delete, audit trail), job and attempt inspection, the resource-class catalog, and worker-artifact registry uploads — the full public API surface.

Documentation

The API reference (Scalar, with try-it-out) and narrative guides are served by your Tempo deployment at /docs/. The client is regenerated from the spec on every API change; see the repository for how the SDKs are produced.