Providers Guide¶

How backends plug into the SDK: vendor-neutral contracts plus one reference local provider. This guide describes what exists today.

Roles¶

Symbol	Role
`IProvider`	Registry of named backends: `capabilities()`, `health()`, `backend_names()`, `get_backend(name)`
`IBackend`	One executable backend: `name`, `capabilities()`, `is_available()`
`ProviderCapabilities`	Vendor-neutral description (kinds, max qubits, metadata)
`ProviderHealth`	Availability flag plus honest message
`ProviderRegistry`	Name → provider map with duplicate protection
`LocalSimulatorProvider`	Reference implementation over MicroQuantum local backends
`ProviderError` (+ `ProviderUnavailableError`, `BackendNotFoundError`, `BackendUnavailableError`)	Typed failures; SDK-wide catch via `QuantsMindError`

Local simulator¶

from quantsmind.providers import LocalSimulatorProvider, ProviderRegistry

registry = ProviderRegistry()
registry.register(LocalSimulatorProvider())
provider = registry.get("local_simulator")
backend = provider.get_backend("statevector")
print(provider.health())          # available=True when microquantum is installed
print(backend.capabilities())     # vendor-neutral view of the native model

Importing quantsmind.providers never requires MicroQuantum; the engine is imported lazily and absence surfaces as ProviderUnavailableError — never a silent substitution.

Boundaries (intentional)¶

No simulator is implemented here: execution delegates to MicroQuantum's backends (local today; IBM/IonQ hardware stays behind MicroQuantum, never a direct SDK dependency).
No credentials are stored: providers accept passthrough config only.
No network access in tests: unavailable paths are exercised with fakes and import blocking.
Circuit, algorithm, and optimizer logic stays in MicroQuantum; domain orchestration stays in quantsmind.quantum and quantsmind.runtime. This package only describes and locates backends.

Pinned by tests/unit/providers/test_provider_abcs.py.