The constraint scheduler

Model the resources that must fit

An exclusive resource can serve one overlapping claim; a cumulative resource can serve several claims up to its capacity. Requirements also describe physical fit, skills valid during the task, precedence between tasks and travel between sites. Active holds consume capacity until their expiry.

findOptions searches the supplied candidate resources and start-time grid depth first. It checks partial assignments and prunes branches that already violate constraints, then checks complete plans again. It cannot discover a time or resource omitted from its input.

Verified means the verifier accepted

verifyPlan is independent of the search strategy. Only an accepted plan gets the feasible-plan marker. A stopped search is not proof that no solution exists; an exhausted evaluation budget or unknown facts can leave the answer unknown.

For a partial day, placeWhatYouCan tries tasks in order and returns a checked placement plus reasons for work left out. This is a useful partial schedule, not a promise to maximize the number of jobs or the business value.

The search does not yet provide objective bounds. A feasible result proves the supplied constraints for its snapshot, not that it is the best possible schedule or that a provider has accepted it.

// Schematic engine calls; problem and budget are typed inputs.
const options = findOptions(problem, budget);
const partialDay = placeWhatYouCan(problem, budget);

Planned

Compare solvers on the same fixtures

An OR-Tools CP-SAT comparison is planned. Any alternative solver would still have to pass the same independent verifier. No comparative speed or production throughput claim is made here.

Try the marina schedule

This interactive trial uses fictional boats and slips. Change availability or fit and run the deterministic engine. It makes no paid AI calls and creates no real reservation.