OR-Tools CP-SAT Methods

Constraint programming is a strong natural fit for job-shop scheduling. JobShopLib already includes an OR-Tools CP-SAT solver, so the OptPilot example reuses that implementation instead of building its own model.

In upstream JobShopLib, this solver is exposed from job_shop_lib.constraint_programming as ORToolsSolver. The bundled OptPilot wrapper imports it in method code and emits the same schedule-solution contract used by the other solver methods.

The included method is:

examples/methods/ortools_cpsat_solver/

It targets:

examples/environments/job_shop_scheduling/environment_schedule_solution.yaml

Install Optional Dependency

JobShopLib is intentionally not a core OptPilot dependency. Install the example extra before running this study:

uv sync --extra examples

Run It

uv run optpilot validate examples/studies/job_shop_ortools_cpsat.yaml
uv run optpilot run examples/studies/job_shop_ortools_cpsat.yaml

What The Method Produces

The method reads the job-shop case references exposed through methodContext.references, runs JobShopLib's OR-Tools solver for each case, and emits schedule solutions:

Candidate spec payload fragment:

solutions:
  ft06_small:
    operations:
      - job: 0
        operation: 0
        machine: 0
        start: 0
        end: 3

The method owns the CP-SAT call:

from job_shop_lib.constraint_programming import ORToolsSolver

The evaluator independently validates the returned schedule. This keeps the environment reusable for any solver that can produce the same schedule-solution shape.

Why OR-Tools, Not Gurobi

Gurobi can model job-shop scheduling, but it is not the best default example for this release because it adds licensing and installation friction. OR-Tools CP-SAT is open, widely used for scheduling, and already exposed by JobShopLib. That makes it the better demonstration of OptPilot's ability to wrap a solver method while keeping the tutorial reproducible.

Users who already have a Gurobi implementation can connect it either as a method that emits schedule-solution parameters or through the solver.py file-candidate contract. In both cases, the Gurobi dependency stays outside the environment.