Open Quantum Python SDK

A lightweight, modern Python SDK for the Open Quantum Platform.

Features:

• Management API — list organizations, backend classes, and providers
• Scheduler API — job categories, upload, prepare, submit, poll, and one-call submit_job()
• Input formats — OpenQASM (qasm) by default; other formats via input_format when the backend supports them
• Smart defaults — auto-select execution plan & queue priority from quote
• Thread-safe auth — Keycloak client-credentials with auto-refresh
• Automatic output download — download_job_output() returns parsed JSON
• Calibration download — download_job_calibration() when the job has calibration_data_url

Installation

pip install -U openquantum-sdk
# or local dev:
pip install -e .

Python 3.8+ required.

Source distributions include the tests/ tree so the unit tests can be run from an sdist. Wheels do not install those tests into site-packages. The Qiskit, PennyLane, Cirq, and Qrisp plugins, and openquantum-mcp, use the same split. Package versions live in each pyproject.toml and __version__ (core, Qiskit, and PennyLane 1.0.0; Cirq and Qrisp 0.5.0; MCP 0.2.0).

Authentication

Option 1: SDK Key File (Recommended)

Download from Open Quantum SDK Keys:

{
  "client_id": "s_abc123...",
  "client_secret": "e460c8..."
}

export OPENQUANTUM_SDK_KEY=/path/to/sdk-key.json

Option 2: Direct in Code

from openquantum_sdk.auth import ClientCredentials, ClientCredentialsAuth

auth = ClientCredentialsAuth(
    creds=ClientCredentials(client_id="s_...", client_secret="e460c8...")
)

Option 3: Environment Variables

export OPENQUANTUM_CLIENT_ID=s_...
export OPENQUANTUM_CLIENT_SECRET=e460c8...

OpenQuantumClient.from_env()

Builds a client from those environment credentials and from OPENQUANTUM_SCHEDULER_URL, OPENQUANTUM_MANAGEMENT_URL, OPENQUANTUM_KEYCLOAK_BASE, and OPENQUANTUM_KEYCLOAK_REALM. It does not autoload a saved account. It raises AuthenticationError when no credential resolves.

from openquantum_sdk import OpenQuantumClient

client = OpenQuantumClient.from_env()

OPENQUANTUM_ORGANIZATION_ID is not read by the core client constructor. PennyLane, Cirq, Qrisp, get_backend, and the MCP server use it when organization_id is omitted. The old names OQ_ORG_ID and OQ_SUBCATEGORY_ID are not read.

Quickstart: One-Call Submission (Recommended)

from openquantum_sdk.clients import SchedulerClient, ManagementClient, JobSubmissionConfig
from openquantum_sdk.enums import ExecutionPlanType, QueuePriorityType
import json

# 1. Instantiate clients (auto-loads auth from env)
scheduler = SchedulerClient()
management = ManagementClient()

try:
    # 2. Discover your Organization ID
    print("Discovering organization...")
    orgs = management.list_user_organizations()
    org_id = orgs.organizations[0].id
    print(f" > Using Org: {orgs.organizations[0].name}")

    # 2.5 Discover available backend providers
    print("\nAvailable providers:")
    providers = management.list_providers()
    for p in providers.providers:
        print(f" > {p.name} ({p.short_code or 'no short code'}) — {p.description}")

    # 2.6 Discover backend classes
    print("\nAvailable backend classes:")
    classes = management.list_backend_classes()
    for bc in classes.backend_classes:
        print(f" > {bc.name} ({bc.short_code or 'no short code'}) — {bc.type}")
        print(f"   ├ Status: {bc.status or 'Unknown'} | Queue depth: {bc.queue_depth or 'N/A'}")
        print(f"   └ Use: '{bc.short_code or bc.id}' in backend_class_id")

    # 3. Define config
    config = JobSubmissionConfig(
        organization_id=org_id,
        backend_class_id="iqm:emerald",
        name="Bell State SDK Quickstart",
        job_subcategory_id="fin:port", # Finance Portfolio Optimization
        shots=100,
        execution_plan="auto",
        queue_priority="auto",
        auto_approve_quote=True,
        verbose=True
    )

    # 4. Submit + wait + get result
    dummy_qasm = b"OPENQASM 2.0; include 'qelib1.inc'; qreg q[2]; creg c[2]; h q[0]; cx q[0],q[1]; measure q -> c;"
    job = scheduler.submit_job(config, file_content=dummy_qasm)

    # 5. Download and print parsed JSON output
    print("\n--- JOB SUCCEEDED ---")
    print(f"Job ID: {job.id}")
    print(f"Status: {job.status}")

    if job.output_data_url:
        output_json = scheduler.download_job_output(job)
        print("\n--- JOB OUTPUT (JSON) ---")
        print(json.dumps(output_json, indent=2))
    else:
        print("No output available.")

    if job.calibration_data_url:
        cal_json = scheduler.download_job_calibration(job)
        print("\n--- CALIBRATION (JSON) ---")
        print(json.dumps(cal_json, indent=2))

except Exception as e:
    print(f"\n--- JOB FAILED ---")
    print(f"Error: {e}")

finally:
    scheduler.close()
    management.close()

Handles everything in one call — including automatic JSON download.

Quickstart: Low-Level (Advanced)

from openquantum_sdk.clients import SchedulerClient, ManagementClient
from openquantum_sdk.models import JobPreparationCreate, JobCreate
from openquantum_sdk.utils import poll_for_status
import json

scheduler = SchedulerClient()
management = ManagementClient()

try:
    # 1. Get org
    org = management.list_user_organizations().organizations[0]

    # 1.5 List providers & backend classes
    print("Providers:")
    for p in management.list_providers().providers:
        print(f" - {p.name} ({p.short_code})")

    print("\nBackend classes:")
    for bc in management.list_backend_classes().backend_classes:
        print(f" - {bc.name} [{bc.type}] — {bc.short_code or bc.id}")

    # 2. Choose IDs
    MY_BACKEND = "iqm:emerald"
    MY_SUBCATEGORY = "fin:port"

    # 3. Upload
    upload_id = scheduler.upload_job_input(file_path="circuit.qasm")

    # 4. Prepare
    prep = JobPreparationCreate(
        organization_id=org.id,
        backend_class_id=MY_BACKEND,
        name="Low-Level Test",
        upload_endpoint_id=upload_id,
        job_subcategory_id=MY_SUBCATEGORY,
        shots=100,
        configuration_data={}
    )
    prep_resp = scheduler.prepare_job(prep)

    # 5. Poll preparation
    def get_prep_status(prep_id: str):
        result = scheduler.get_preparation_result(prep_id)
        done = result.status in ("Completed", "Failed")
        return done, result

    result = poll_for_status(
        get_status_fn=get_prep_status,
        resource_id=prep_resp.id,
        interval=2.0,
        timeout=300
    )

    if result.status == "Failed":
        raise RuntimeError(f"Preparation failed: {result.message or 'Unknown error'}")

    # Quote inspection
    print("\nQuote received:")
    for plan in result.quote:
        print(f" • {plan.name} — Base: {plan.price} credit{'s' if plan.price != 1 else ''}")
        for qp in plan.queue_priorities:
            total = plan.price + qp.price_increase
            print(f"    ├ {qp.name}: +{qp.price_increase} → Total: {total}")
        print(f"    └ Plan ID: {plan.execution_plan_id}")

    # 6. Select cheapest
    cheapest_plan = min(result.quote, key=lambda p: p.price)
    cheapest_prio = min(cheapest_plan.queue_priorities, key=lambda q: q.price_increase)

    # 7. Create job (organization comes from the preparation)
    job = JobCreate(
        job_preparation_id=prep_resp.id,
        execution_plan_id=cheapest_plan.execution_plan_id,
        queue_priority_id=cheapest_prio.queue_priority_id,
    )

    job_resp = scheduler.create_job(job)

    # 8. Poll completion
    def get_job_status(job_id: str):
        job_obj = scheduler.get_job(job_id)
        done = job_obj.status in ("Completed", "Failed", "Canceled")
        return done, job_obj

    final_job = poll_for_status(
        get_status_fn=get_job_status,
        resource_id=job_resp.id,
        interval=5.0,
        timeout=86_400,
    )

    print(f"\nJob finished: {final_job.status}")

    # 9. Download and print JSON output
    if final_job.output_data_url:
        output_json = scheduler.download_job_output(final_job)
        print("\n--- JOB OUTPUT (JSON) ---")
        print(json.dumps(output_json, indent=2))
    else:
        print("No output available.")

finally:
    scheduler.close()
    management.close()

High-Level API: submit_job()

JobSubmissionConfig

API Reference

SchedulerClient

ManagementClient

Data Models

Utils: poll_for_status()

from openquantum_sdk.utils import poll_for_status

Enums

from openquantum_sdk.enums import ExecutionPlanType, QueuePriorityType

CLI (Experimental)

python -m openquantum_sdk --sdk-key ./key.json \
                          --input bell.qasm \
                          --backend "iqm:emerald" \
                          --name "CLI Circuit" \
                          --subcategory "mathematics:linear-systems" \
                          --shots 100 \
                          --auto-approve

Automatically downloads and pretty-prints JSON output.

Short Codes

Backends

Note: This list may be modified as backends are added or removed from the platform. For the most up-to-date list, run the following code:

backends = management.list_backend_classes().backend_classes

print('| Name | Short Code | Online |')
for b in backends:
    print(f"| {b.name} | {b.short_code} | {b.accepting_jobs} |")

Subcategories

Note: This list may be modified as categories and subcategories are added or removed. For the most up-to-date list, run the following code:

cats = scheduler.get_job_categories().categories

print('| Category | Subcategory | Short Code |')
for cat in cats:
    subcats = scheduler.get_job_subcategories(cat.short_code).categories
    for subcat in subcats:
        print(f'| {cat.name} | {subcat.name} | {subcat.short_code} |')

Live smoke tests

.github/workflows/sdk-smoke.yml is a manual Actions workflow. It builds the core SDK and the Qiskit, PennyLane, Cirq, and Qrisp plugins from a git ref (empty means main), submits a Bell circuit through each one, and waits until the platform job completes. The QPU input defaults to rigetti:cepheus-1-108q. Choose the dev or prd environment when you run it. See sdk/smoke/README.md.

FAQ

Q: How do I get results as JSON?
A: Use scheduler.download_job_output(job) — returns parsed object.

Q: Why does auto mode fail on some QPUs? A: "auto" selects the cheapest plan and priority, which is often Public + Standard. It is not limited to those two. Friendly names ("public", "private", "standard", "priority", "instant"), enums, and UUIDs select that option exactly. If you need a specific plan or a faster queue, pass it explicitly. The MCP server passes prefer_public=True so Spark credits can apply; submit_job() does not.

Q: Can I hide the public plan citation notice?
A: Setting OPENQUANTUM_HIDE_CITATION_NOTICE=true will omit the citation notice from output.