Aller au contenu principal

Python SDK

Status: early access. The Python client is scaffolded in app/sdk/python/ and covers the hosted API surface. Not yet published to PyPI; install from source.

Install from source​

git clone https://github.com/hulyasmath/zeq-framework
cd zeq-framework/app/sdk/python
pip install -e .

Quickstart​

import os
from zeq import Zeq

zeq = Zeq(api_key=os.environ["ZEQ_API_KEY"])

# Read pulse
p = zeq.pulse()
# {"zeqond": 72380021.4, "phase": 0.412, "f_hz": 1.287, "T_s": 0.777, "R_t": 0.999917}

# Hosted compute (signed CKO)
result = zeq.compute(
    domain="quantum_mechanics",
    inputs={"frequency": 5e14},
)
print(result.value, result.unit)
# 3.3126157e-19 J

print(result.ko42)
# {"mode": "KO42.1", "alpha": 0.00129, "error_band": 4.3e-4, "within_bound": True}

print(result.signature)
# ed25519:...

Parity with the TypeScript SDK​

The Python client is a thin wrapper over the hosted API. The function names match the TypeScript SDK:

TypeScriptPython
pulse()zeq.pulse()
unixToZeqond(t)zeq.unix_to_zeqond(t)
zeqondToUnix(z)zeq.zeqond_to_unix(z)
ZeqClient().compute(...)Zeq(...).compute(...)
ZeqMailClientzeq.mail(...)
ZeqMeshClientzeq.mesh(...)

What's missing​

  • Local compute. The Python SDK currently routes every compute() call through the hosted API. The TypeScript SDK can compute the 42-operator catalogue locally; the Python port is planned but not implemented.
  • PyPI publication. Install from source today; PyPI release targeted for the next framework minor version.
  • Async client. Current client is synchronous; an AsyncZeq class is on the roadmap.

Contributing​

The Python SDK is the highest-value open-source contribution right now. If you want to help:

  • Port the solvers.ts operator catalogue to Python (closest to TensorFlow-class work — clean numerical code).
  • Port the mail.ts, mesh.ts, daemon.ts vertical clients (straightforward — wrap HTTP endpoints).
  • Add pytest coverage for the full protocol registry.

See Contributing for the PR workflow.