"""Objective and solver configuration for OR-Tools CP-SAT solver.
This module handles objective function building and solver parameter configuration.
"""
from ortools.sat.python import cp_model
from chuk_mcp_solver.models import (
Objective,
ObjectiveSense,
SearchStrategy,
SolveConstraintModelRequest,
)
def build_objective(
model: cp_model.CpModel,
objective: Objective | list[Objective],
var_map: dict[str, cp_model.IntVar],
) -> None:
"""Build the objective function.
Supports both single and multi-objective optimization.
For multi-objective, uses lexicographic ordering by priority.
Args:
model: The CP-SAT model.
objective: Single objective or list of objectives.
var_map: Mapping from variable ID to CP-SAT variable.
"""
# Handle single objective
if isinstance(objective, Objective):
expr = sum(int(term.coef) * var_map[term.var] for term in objective.terms)
if objective.sense == ObjectiveSense.MINIMIZE:
model.Minimize(expr)
else: # MAXIMIZE
model.Maximize(expr)
return
# Handle multi-objective
# Sort by priority (higher = more important)
sorted_objectives = sorted(objective, key=lambda o: o.priority, reverse=True)
# For OR-Tools CP-SAT, we use weighted sum approach for multi-objective
# since lexicographic is not directly supported
# Higher priority objectives get much larger weights
total_expr = 0
weight_multiplier = 1000000 # Scale factor between priority levels
for obj in sorted_objectives:
expr = sum(int(term.coef) * var_map[term.var] for term in obj.terms)
# Apply priority scaling and user-defined weight
scaled_weight = (weight_multiplier**obj.priority) * obj.weight
total_expr += scaled_weight * expr
# Use the sense of the highest priority objective
if sorted_objectives[0].sense == ObjectiveSense.MINIMIZE:
model.Minimize(total_expr)
else:
model.Maximize(total_expr)
def configure_solver(solver: cp_model.CpSolver, request: SolveConstraintModelRequest) -> None:
"""Configure solver parameters.
Args:
solver: The CP-SAT solver.
request: The request containing search configuration.
"""
if not request.search:
return
# Time limit
if request.search.max_time_ms:
solver.parameters.max_time_in_seconds = request.search.max_time_ms / 1000.0
# Number of parallel search workers
if request.search.num_search_workers:
solver.parameters.num_search_workers = request.search.num_search_workers
# Search progress logging
if request.search.log_search_progress:
solver.parameters.log_search_progress = True
# Random seed for deterministic solving
if request.search.random_seed is not None:
solver.parameters.random_seed = request.search.random_seed
# Search strategy
if request.search.strategy != SearchStrategy.AUTO:
# Map our strategy enum to OR-Tools parameters
if request.search.strategy == SearchStrategy.FIRST_FAIL:
solver.parameters.search_branching = cp_model.FIXED_SEARCH
solver.parameters.preferred_variable_order = cp_model.CHOOSE_FIRST # type: ignore[assignment]
elif request.search.strategy == SearchStrategy.RANDOM:
solver.parameters.randomize_search = True
# Note: Warm start solution hints are handled separately in solve_constraint_model
# via solver.AddHint() after model construction
