A minimal Python 3.11 prototype for delivery order-rider assignment.

python main.py eval_cases/small/sample.json

Optional time limit and solver:

python main.py eval_cases/small/sample.json --time-limit 10 --solver seed_solvers.greedy_single

python -m pytest -q

{
  "orders": [{"id": "o1"}],
  "riders": [{"id": "r1"}],
  "pair_scores": [
    {"order_id": "o1", "rider_id": "r1", "score": 2.0, "accept_prob": 0.9}
  ],
  "max_riders_per_order": 1,
  "max_orders_per_rider": 1
}

The first baseline evaluates solutions by:

1. Maximizing assigned order count.
2. Minimizing total assigned pair score when assigned order count ties.

seed_solvers/greedy_single.py — sorts available order-rider pairs by:

score ascending, accept_prob descending, order_id, rider_id

It adds each pair when doing so stays within the simple capacity constraints.

Additional baseline solvers are in seed_solvers/. LLM-generated solver variants go in generated_solvers/. The evolution framework is in evolve/.