Python Async gRPC Client#

The Python async gRPC client (cuopt.grpc.linear_programming.Client) is an explicit gRPC client for cuopt_grpc_server. It uses a job lifecycle: submit → wait / status → result → delete.

“Async” here means the job-based API (submit-and-wait / submit-and-poll). It is not Python asyncio.

For remote execution (zero code change via CUOPT_REMOTE_HOST / CUOPT_REMOTE_PORT), see Quick Start and the section overview in gRPC Remote Execution. Prefer this client when you need cancel, live log streaming, MIP incumbent streaming, or to manage multiple jobs yourself.

Prerequisites#

A running cuopt_grpc_server on a GPU host (see Quick Start):

cuopt_grpc_server --port 5001 --workers 1

Connect and Solve#

The LP below matches remote_lp_demo.py from the quick start (same constraint matrix and objective).

import numpy as np
from cuopt import linear_programming
from cuopt.grpc.linear_programming import Client, JobStatus

dm = linear_programming.DataModel()
dm.set_csr_constraint_matrix(
    np.array([3.0, 4.0, 2.7, 10.1], dtype=np.float64),
    np.array([0, 1, 0, 1], dtype=np.int32),
    np.array([0, 2, 4], dtype=np.int32),
)
dm.set_constraint_bounds(np.array([5.4, 4.9], dtype=np.float64))
dm.set_objective_coefficients(np.array([0.2, 0.1], dtype=np.float64))
dm.set_maximize(True)
dm.set_row_types(np.array(["L", "L"]))
dm.set_variable_lower_bounds(np.array([0.0, 0.0], dtype=np.float64))
dm.set_variable_upper_bounds(np.array([2.0, np.inf], dtype=np.float64))

settings = linear_programming.SolverSettings()
client = Client("localhost", 5001)  # tls=None uses CUOPT_TLS_* if set
job_id = client.submit(dm, settings)
try:
    if client.wait(job_id, timeout=120) != JobStatus.COMPLETED:
        raise RuntimeError("job did not complete")
    # Pass names if you want solution.get_vars() keyed by name.
    solution = client.result(job_id, variable_names=["x0", "x1"])
    print(solution.get_termination_reason(), solution.get_primal_objective())
finally:
    client.delete(job_id)

Client.submit() accepts either a DataModel or a Problem. Always call delete after you are done with the job so the server can release state.

Variable Names#

result(job_id, variable_names=...) builds the solution object. Pass a list of variable names (same order as the model columns) if you want solution.get_vars() keyed by those names. You can omit names and still use get_primal_solution() and other numeric accessors.

TLS / mTLS#

  • tls=None (default) — honor CUOPT_TLS_* environment variables.

  • tls=False — plain TCP; ignore CUOPT_TLS_*.

  • tls=TlsConfig(...) — explicit PEM text or file paths.

See Advanced Configuration for server-side TLS and which environment variables apply to remote execution versus this gRPC client.

Next steps#