Prepare Your First Vendor Hardware Workflow

The workflow for preparing a qalgora-Q kernel for real vendor hardware follows the same shape on every provider: get credentials, write a kernel, transpile, and submit through the vendor's SDK to read results. The open reference build itself runs locally; this page is the universal recipe, and each provider's workflow page fills in the specifics.

The five steps

1. Get access — create an account with a provider and obtain an API token.
2. Set credentials — export the token as an environment variable.
3. Write & validate — build a kernel and run it on the qpp-cpu CPU statevector reference implementation first.
4. Submit — set the hardware target and submit through the vendor adapter (planned · not in the open reference build).
5. Retrieve — poll status, then read the counts (adapter workflow).

1–3 · Write and validate locally

import qalgora

@qalgora.kernel
def bell():
    q = qalgora.qvector(2)
    h(q[0])
    x.ctrl(q[0], q[1])
    mz(q)

qalgora.set_target("qpp-cpu")                   # reference implementation, local validation
# qalgora.set_target("gpu")                     # optional: requires qalgora-gpu plugin (planned)
print(qalgora.sample(bell, shots_count=1000))   # expect ~{00, 11}

4 · Submit to hardware (adapter workflow · planned)

# qalgora.set_target("ibm", machine="<ibm_backend_name>")  # planned adapter — needs vendor SDK + credentials
# job = qalgora.sample_async(bell, shots_count=4000)        # returns immediately (adapter future)
# print(job.id(), job.status())                             # queued / running / done

5 · Retrieve results (adapter workflow · planned)

# counts = job.get()              # blocks until the job finishes
# print(counts.most_probable(), counts.probability("11"))

Good habits

• Validate on a simulator before spending hardware time or money.
• Use async — hardware jobs queue; do other work while you wait.
• Mind the bitstring order and add explicit mz.
• Turn on mitigation for expectation values — see Error Mitigation.