unified_interface.py

"""
Unified Interface for Algorithm Recommender System
===================================================

This module provides a single, unified interface for all algorithm
recommendation functionality. Users can simply provide raw input
(arrays or graphs) and receive intelligent algorithm recommendations.

The system supports three domains:
1. SORTING: Recommends among Insertion, Selection, Merge, Quick, Heap Sort
2. ARRAY SEARCHING: Recommends among Linear, Binary, Jump, Exponential Search
3. GRAPH SEARCHING: Recommends among BFS, Dijkstra, Bellman-Ford, A*

Key Features:
- Automatic feature extraction from raw input
- ML-based algorithm selection with correctness constraints
- Human-readable explanations for all recommendations
- Optional execution of recommended algorithm

Usage:
    from algorithm_recommender import AlgorithmRecommender

    recommender = AlgorithmRecommender()
    recommender.load_models()  # Or train_all() if models don't exist

    # Sorting recommendation
    result = recommender.recommend_sorting([5, 2, 8, 1, 9])
    print(result['algorithm'], result['explanation'])

    # Array search recommendation
    result = recommender.recommend_array_search([1, 2, 3, 4, 5])
    print(result['algorithm'])

    # Graph search recommendation
    graph = {0: [(1, 1.0), (2, 4.0)], 1: [(2, 2.0)], 2: []}
    result = recommender.recommend_graph_search(graph)
    print(result['algorithm'])

Author: Algorithm Recommender Research Team
Version: 1.0.0
"""

import os
from typing import List, Dict, Tuple, Optional, Union, Any, Callable

# Import recommenders
from sorting.recommender import SortingRecommender
from sorting.algorithms import SORTING_ALGORITHMS
from searching.array_search.recommender import ArraySearchRecommender
from searching.array_search.algorithms import ARRAY_SEARCH_ALGORITHMS
from searching.graph_search.recommender import GraphSearchRecommender
from searching.graph_search.algorithms import GRAPH_SEARCH_ALGORITHMS

# Import new modules
from searching.string_search.recommender import StringSearchRecommender
from searching.string_search.algorithms import STRING_SEARCH_ALGORITHMS
from searching.tree_search.recommender import TreeSearchRecommender
from searching.tree_search.algorithms import TREE_SEARCH_ALGORITHMS, BST, AVLTree

# Import structure inference
from utils.structure_inference import (
    StructureInference, StructureType, StructureInferenceResult,
    infer_structure, infer_from_problem
)


class AlgorithmRecommender:
    """
    Unified interface for all algorithm recommendation functionality.

    This class provides a single entry point for recommending algorithms
    across three domains: sorting, array searching, and graph searching.

    The recommender:
    1. Accepts raw input (no precomputed features required)
    2. Automatically extracts relevant features
    3. Uses trained ML models to recommend algorithms
    4. Enforces correctness constraints (especially for searching)
    5. Provides human-readable explanations

    Attributes:
        sorting: SortingRecommender instance
        array_search: ArraySearchRecommender instance
        graph_search: GraphSearchRecommender instance

    Example:
        >>> recommender = AlgorithmRecommender()
        >>> recommender.load_models()
        >>>
        >>> # Sorting
        >>> result = recommender.recommend_sorting([3, 1, 4, 1, 5])
        >>> print(f"Use {result['algorithm']}: {result['explanation']}")
        >>>
        >>> # Array search
        >>> result = recommender.recommend_array_search([1, 2, 3, 4, 5])
        >>> print(f"Use {result['algorithm']}")
        >>>
        >>> # Graph search
        >>> graph = {0: [(1, 1.0)], 1: [(2, 1.0)], 2: []}
        >>> result = recommender.recommend_graph_search(graph)
        >>> print(f"Use {result['algorithm']}")
    """

    def __init__(self, models_dir: Optional[str] = None):
        """
        Initialize the unified recommender.

        Args:
            models_dir: Directory containing trained models.
                        Defaults to 'models/' relative to this file.
        """
        if models_dir is None:
            self.models_dir = os.path.join(
                os.path.dirname(os.path.abspath(__file__)), 'models'
            )
        else:
            self.models_dir = models_dir

        # Initialize recommenders
        self.sorting = SortingRecommender()
        self.array_search = ArraySearchRecommender()
        self.graph_search = GraphSearchRecommender()
        self.string_search = StringSearchRecommender()
        self.tree_search = TreeSearchRecommender()

        # Initialize structure inference
        self.structure_inference = StructureInference()

        self._models_loaded = False

    def load_models(self, verbose: bool = True):
        """
        Load all pre-trained models from disk.

        Args:
            verbose: Whether to print loading progress.

        Raises:
            FileNotFoundError: If any required model file is missing.
        """
        sorting_path = os.path.join(self.models_dir, 'sorting_model.pkl')
        array_search_path = os.path.join(self.models_dir, 'array_search_model.pkl')
        graph_search_path = os.path.join(self.models_dir, 'graph_search_model.pkl')
        string_search_path = os.path.join(self.models_dir, 'string_search_model.pkl')
        tree_search_path = os.path.join(self.models_dir, 'tree_search_model.pkl')

        # Check required models
        missing = []
        if not os.path.exists(sorting_path):
            missing.append('sorting_model.pkl')
        if not os.path.exists(array_search_path):
            missing.append('array_search_model.pkl')
        if not os.path.exists(graph_search_path):
            missing.append('graph_search_model.pkl')

        if missing:
            raise FileNotFoundError(
                f"Missing required model files: {missing}. "
                f"Run train_all() to generate models."
            )

        if verbose:
            print("Loading models...")

        self.sorting.load_model(sorting_path)
        self.array_search.load_model(array_search_path)
        self.graph_search.load_model(graph_search_path)

        # Load optional new models if they exist
        if os.path.exists(string_search_path):
            try:
                self.string_search.load_model(string_search_path)
                if verbose:
                    print("  ✓ String search model loaded")
            except Exception as e:
                if verbose:
                    print(f"  ⚠ String search model not loaded: {e}")
        elif verbose:
            print("  ⚠ String search model not found (optional)")

        if os.path.exists(tree_search_path):
            try:
                self.tree_search.load_model(tree_search_path)
                if verbose:
                    print("  ✓ Tree search model loaded")
            except Exception as e:
                if verbose:
                    print(f"  ⚠ Tree search model not loaded: {e}")
        elif verbose:
            print("  ⚠ Tree search model not found (optional)")

        self._models_loaded = True

        if verbose:
            print("All required models loaded successfully!")

    def train_all(
        self,
        sorting_instances: int = 800,
        array_search_instances: int = 400,
        graph_search_instances: int = 300,
        verbose: bool = True
    ):
        """
        Train all models from scratch.

        This generates training data, trains RandomForest classifiers,
        and saves the models to disk.

        Args:
            sorting_instances: Number of sorting training instances.
            array_search_instances: Number of array search instances.
            graph_search_instances: Number of graph search instances.
            verbose: Whether to print progress.
        """
        os.makedirs(self.models_dir, exist_ok=True)

        # Train sorting model
        if verbose:
            print("\n" + "=" * 60)
            print("TRAINING SORTING MODEL")
            print("=" * 60)

        from sorting.train import train_sorting_model
        train_sorting_model(
            num_instances=sorting_instances,
            output_dir=self.models_dir,
            verbose=verbose
        )
        self.sorting.load_model(models_dir=self.models_dir)

        # Train array search model
        if verbose:
            print("\n" + "=" * 60)
            print("TRAINING ARRAY SEARCH MODEL")
            print("=" * 60)

        from searching.array_search.train import train_array_search_model
        train_array_search_model(
            num_instances=array_search_instances,
            output_dir=self.models_dir,
            verbose=verbose
        )
        self.array_search.load_model(models_dir=self.models_dir)

        # Train graph search model
        if verbose:
            print("\n" + "=" * 60)
            print("TRAINING GRAPH SEARCH MODEL")
            print("=" * 60)

        from searching.graph_search.train import train_graph_search_model
        train_graph_search_model(
            num_instances=graph_search_instances,
            output_dir=self.models_dir,
            verbose=verbose
        )
        self.graph_search.load_model(models_dir=self.models_dir)

        # Train string search model
        if verbose:
            print("\n" + "=" * 60)
            print("TRAINING STRING SEARCH MODEL")
            print("=" * 60)

        from searching.string_search.train import train_string_search_model
        string_search_path = os.path.join(self.models_dir, 'string_search_model.pkl')
        train_string_search_model(
            num_instances=300,
            output_path=string_search_path,
            seed=42,
            verbose=verbose
        )
        self.string_search.load_model(models_dir=self.models_dir)

        # Train tree search model
        if verbose:
            print("\n" + "=" * 60)
            print("TRAINING TREE SEARCH MODEL")
            print("=" * 60)

        from searching.tree_search.train import train_tree_search_model
        tree_search_path = os.path.join(self.models_dir, 'tree_search_model.pkl')
        train_tree_search_model(
            num_instances=300,
            output_path=tree_search_path,
            seed=42,
            verbose=verbose
        )
        self.tree_search.load_model(models_dir=self.models_dir)

        self._models_loaded = True

        if verbose:
            print("\n" + "=" * 60)
            print("ALL MODELS TRAINED SUCCESSFULLY!")
            print("=" * 60)

    def recommend_sorting(
        self,
        arr: List[Union[int, float]],
        return_probabilities: bool = False
    ) -> Dict:
        """
        Recommend the best sorting algorithm for a raw input array.

        The system automatically extracts features and uses the ML model
        to select the optimal algorithm.

        Args:
            arr: Raw input array to be sorted.
            return_probabilities: Include probability distribution.

        Returns:
            Dictionary containing:
            - 'algorithm': Recommended algorithm name
            - 'explanation': Why this algorithm was chosen
            - 'features': Extracted features
            - 'feature_summary': Human-readable feature description
            - 'probabilities': (if requested) Probability per algorithm

        Example:
            >>> result = recommender.recommend_sorting([5, 2, 8, 1, 9])
            >>> print(result['algorithm'])
            'insertion_sort'
            >>> print(result['explanation'])
            'Array is small, nearly sorted → Insertion Sort selected.'
        """
        self._ensure_models_loaded()
        return self.sorting.recommend(arr, return_probabilities)

    def recommend_array_search(
        self,
        arr: List[Union[int, float]],
        return_probabilities: bool = False
    ) -> Dict:
        """
        Recommend the best array search algorithm.

        Correctness Constraints (STRICT):
        - Linear Search: Always valid
        - Binary/Jump Search: Only if array is sorted
        - Exponential Search: Only if sorted and uniform

        The recommender will NEVER recommend an invalid algorithm.

        Args:
            arr: Raw input array.
            return_probabilities: Include probability distribution.

        Returns:
            Dictionary with algorithm, explanation, valid_algorithms, etc.

        Example:
            >>> result = recommender.recommend_array_search([1, 2, 3, 4, 5])
            >>> print(result['algorithm'])
            'binary_search'
            >>> print(result['valid_algorithms'])
            ['linear_search', 'binary_search', 'jump_search', 'exponential_search']
        """
        self._ensure_models_loaded()
        return self.array_search.recommend(arr, return_probabilities=return_probabilities)

    def recommend_graph_search(
        self,
        graph: Dict[Any, List[Tuple[Any, float]]],
        has_heuristic: bool = False,
        return_probabilities: bool = False
    ) -> Dict:
        """
        Recommend the best graph search algorithm.

        Correctness Constraints (STRICT):
        - BFS: Only for unweighted graphs
        - Dijkstra: Only for non-negative weights
        - Bellman-Ford: Required for negative weights
        - A*: Only when heuristic is available

        The recommender will NEVER recommend an invalid algorithm.

        Args:
            graph: Adjacency list {node: [(neighbor, weight), ...]}.
            has_heuristic: Whether a heuristic function is available.
            return_probabilities: Include probability distribution.

        Returns:
            Dictionary with algorithm, explanation, valid_algorithms, etc.

        Example:
            >>> graph = {0: [(1, 1.0)], 1: [(2, 1.0)], 2: []}
            >>> result = recommender.recommend_graph_search(graph)
            >>> print(result['algorithm'])
            'bfs'
        """
        self._ensure_models_loaded()
        return self.graph_search.recommend(graph, has_heuristic, return_probabilities)

    def recommend_string_search(
        self,
        text: str,
        patterns: Union[str, List[str]],
        is_prefix_query: bool = False,
        return_probabilities: bool = False
    ) -> Dict:
        """
        Recommend the best string search algorithm.

        Correctness Constraints (STRICT):
        - Naive String Search: Always valid
        - Trie Search: Only valid for multiple patterns OR prefix queries

        Args:
            text: The text to search in.
            patterns: Single pattern or list of patterns.
            is_prefix_query: Whether this is a prefix-based search.
            return_probabilities: Include probability distribution.

        Returns:
            Dictionary with algorithm, explanation, valid_algorithms, etc.

        Example:
            >>> result = recommender.recommend_string_search("hello world", ["hello", "world"])
            >>> print(result['algorithm'])
            'trie_search'
        """
        self._ensure_models_loaded()
        if self.string_search.model is None:
            raise RuntimeError("String search model not loaded. Train with train_all().")
        return self.string_search.recommend(text, patterns, is_prefix_query, return_probabilities)

    def recommend_tree_search(
        self,
        tree: Union[BST, AVLTree, Dict],
        keys: Optional[List[Any]] = None,
        return_probabilities: bool = False
    ) -> Dict:
        """
        Recommend the best tree search algorithm.

        Correctness Constraints (STRICT):
        - BST Search: Always valid (but O(n) worst case on skewed trees)
        - Balanced Tree Search: Only valid on balanced trees

        Args:
            tree: BST, AVLTree, or dict representation.
            keys: Optional list of keys if tree is dict.
            return_probabilities: Include probability distribution.

        Returns:
            Dictionary with algorithm, explanation, valid_algorithms, etc.

        Example:
            >>> bst = BST()
            >>> bst.from_list([5, 3, 7, 1, 9])
            >>> result = recommender.recommend_tree_search(bst)
            >>> print(result['algorithm'])
        """
        self._ensure_models_loaded()
        if self.tree_search.model is None:
            raise RuntimeError("Tree search model not loaded. Train with train_all().")
        return self.tree_search.recommend(tree, keys, return_probabilities)

    def recommend_auto(
        self,
        data: Any = None,
        problem_statement: str = "",
        operation_hint: str = "",
        **kwargs
    ) -> Dict:
        """
        Automatically detect input type and recommend appropriate algorithm.

        This is the smart entry point that uses structure inference to
        determine what kind of problem the user has and routes to the
        correct recommender.

        Args:
            data: Input data (array, graph, tree, string, etc.)
            problem_statement: Natural language problem description.
            operation_hint: Hint about operation (sort, search, etc.)
            **kwargs: Additional arguments passed to specific recommender.

        Returns:
            Dictionary containing:
            - 'task_type': Detected task type
            - 'structure_type': Detected structure type
            - 'algorithm': Recommended algorithm
            - 'explanation': Why this algorithm was chosen
            - 'inference_result': Structure inference details

        Example:
            >>> result = recommender.recommend_auto([5, 2, 8, 1, 9], operation_hint="sort")
            >>> print(result['task_type'], result['algorithm'])
        """
        self._ensure_models_loaded()

        # Infer from problem statement if provided
        if problem_statement:
            inference_result = self.structure_inference.infer_from_problem(problem_statement)
        elif data is not None:
            inference_result = self.structure_inference.infer(data)
        else:
            raise ValueError("Either data or problem_statement must be provided")

        structure_type = inference_result.structure_type
        task_type = self.structure_inference.get_task_type(structure_type, operation_hint)

        result = {
            'task_type': task_type,
            'structure_type': structure_type.value,
            'inference_result': inference_result.to_dict(),
        }

        try:
            if task_type == 'sorting':
                recommendation = self.sorting.recommend(data, **kwargs)
            elif task_type == 'array_search':
                recommendation = self.array_search.recommend(data, **kwargs)
            elif task_type == 'graph_search':
                has_heuristic = kwargs.get('has_heuristic', False)
                recommendation = self.graph_search.recommend(data, has_heuristic, **kwargs)
            elif task_type == 'string_search':
                if isinstance(data, tuple) and len(data) >= 2:
                    text, patterns = data[0], data[1]
                else:
                    text = data
                    patterns = kwargs.get('patterns', [''])
                is_prefix = kwargs.get('is_prefix_query', False)
                recommendation = self.string_search.recommend(text, patterns, is_prefix)
            elif task_type == 'tree_search':
                keys = kwargs.get('keys', None)
                recommendation = self.tree_search.recommend(data, keys)
            else:
                recommendation = {'algorithm': 'unknown', 'explanation': 'Could not determine task type'}

            result.update(recommendation)
        except Exception as e:
            result['error'] = str(e)
            result['algorithm'] = 'unknown'
            result['explanation'] = f'Error during recommendation: {e}'

        return result

    def sort(
        self,
        arr: List[Union[int, float]]
    ) -> Tuple[List[float], str, str]:
        """
        Recommend and execute sorting on an array.

        Convenience method that recommends the best algorithm and
        executes it in one call.

        Args:
            arr: Raw input array.

        Returns:
            Tuple of (sorted_array, algorithm_name, explanation).

        Example:
            >>> sorted_arr, algo, why = recommender.sort([5, 2, 8, 1, 9])
            >>> print(sorted_arr)
            [1, 2, 5, 8, 9]
        """
        self._ensure_models_loaded()
        return self.sorting.execute_recommendation(arr)

    def search_array(
        self,
        arr: List[Union[int, float]],
        target: Union[int, float]
    ) -> Tuple[int, str, str]:
        """
        Recommend and execute array search.

        Args:
            arr: Raw input array.
            target: Value to search for.

        Returns:
            Tuple of (index, algorithm_name, explanation).
            Index is -1 if target not found.

        Example:
            >>> idx, algo, why = recommender.search_array([1, 2, 3, 4, 5], 3)
            >>> print(idx, algo)
            2 binary_search
        """
        self._ensure_models_loaded()
        return self.array_search.execute_recommendation(arr, target)

    def search_graph(
        self,
        graph: Dict[Any, List[Tuple[Any, float]]],
        start: Any,
        goal: Any,
        heuristic: Optional[Callable] = None,
        node_positions: Optional[Dict] = None
    ) -> Tuple[Optional[List[Any]], float, str, str]:
        """
        Recommend and execute graph search.

        Args:
            graph: Adjacency list representation.
            start: Starting node.
            goal: Target node.
            heuristic: Optional heuristic function for A*.
            node_positions: Optional positions for Euclidean heuristic.

        Returns:
            Tuple of (path, cost, algorithm_name, explanation).
            Path is None if goal is unreachable.

        Example:
            >>> graph = {0: [(1, 1.0)], 1: [(2, 1.0)], 2: []}
            >>> path, cost, algo, why = recommender.search_graph(graph, 0, 2)
            >>> print(path, cost)
            [0, 1, 2] 2.0
        """
        self._ensure_models_loaded()
        return self.graph_search.execute_recommendation(
            graph, start, goal, heuristic, node_positions
        )

    def _ensure_models_loaded(self):
        """Ensure models are loaded before making recommendations."""
        if not self._models_loaded:
            raise RuntimeError(
                "Models not loaded. Call load_models() or train_all() first."
            )

    def get_available_algorithms(self) -> Dict[str, List[str]]:
        """
        Get all available algorithms organized by domain.

        Returns:
            Dictionary mapping domain to algorithm names.
        """
        return {
            'sorting': list(SORTING_ALGORITHMS.keys()),
            'array_search': list(ARRAY_SEARCH_ALGORITHMS.keys()),
            'graph_search': list(GRAPH_SEARCH_ALGORITHMS.keys()),
            'string_search': list(STRING_SEARCH_ALGORITHMS.keys()),
            'tree_search': list(TREE_SEARCH_ALGORITHMS.keys()),
        }

    def get_feature_importance(self, domain: str) -> Dict[str, float]:
        """
        Get feature importance for a specific domain's model.

        Args:
            domain: One of 'sorting', 'array_search', 'graph_search'.

        Returns:
            Dictionary mapping feature names to importance scores.
        """
        self._ensure_models_loaded()

        if domain == 'sorting':
            return self.sorting.get_feature_importance()
        elif domain == 'array_search':
            # Array search recommender doesn't have this method yet
            raise NotImplementedError("Feature importance not yet implemented for array_search")
        elif domain == 'graph_search':
            raise NotImplementedError("Feature importance not yet implemented for graph_search")
        else:
            raise ValueError(f"Unknown domain: {domain}")


def demo():
    """
    Demonstrate the unified algorithm recommender.

    This function shows how to use the system for all three domains.
    """
    print("=" * 70)
    print("ALGORITHM RECOMMENDER SYSTEM - DEMONSTRATION")
    print("=" * 70)

    # Initialize recommender
    recommender = AlgorithmRecommender()

    # Check if models exist, train if not
    try:
        recommender.load_models()
    except FileNotFoundError:
        print("\nModels not found. Training all models...")
        recommender.train_all(
            sorting_instances=500,
            array_search_instances=300,
            graph_search_instances=200
        )

    print("\n" + "-" * 70)
    print("SORTING DEMONSTRATIONS")
    print("-" * 70)

    # Demo 1: Small random array
    arr1 = [5, 2, 8, 1, 9, 3, 7]
    result1 = recommender.recommend_sorting(arr1)
    print(f"\nArray: {arr1}")
    print(f"Recommended: {result1['algorithm']}")
    print(f"Explanation: {result1['explanation']}")

    # Demo 2: Large sorted array
    arr2 = list(range(1000))
    result2 = recommender.recommend_sorting(arr2)
    print(f"\nArray: [0, 1, 2, ..., 999] (sorted, n=1000)")
    print(f"Recommended: {result2['algorithm']}")
    print(f"Explanation: {result2['explanation']}")

    # Demo 3: Execute sorting
    arr3 = [10, 5, 8, 3, 1]
    sorted_arr, algo, explanation = recommender.sort(arr3)
    print(f"\nSorting {arr3}")
    print(f"Result: {sorted_arr}")
    print(f"Used: {algo}")

    print("\n" + "-" * 70)
    print("ARRAY SEARCH DEMONSTRATIONS")
    print("-" * 70)

    # Demo 4: Sorted array
    arr4 = list(range(100))
    result4 = recommender.recommend_array_search(arr4)
    print(f"\nArray: [0, 1, 2, ..., 99] (sorted)")
    print(f"Recommended: {result4['algorithm']}")
    print(f"Valid options: {result4['valid_algorithms']}")
    print(f"Explanation: {result4['explanation']}")

    # Demo 5: Unsorted array
    arr5 = [5, 2, 8, 1, 9, 3, 7]
    result5 = recommender.recommend_array_search(arr5)
    print(f"\nArray: {arr5} (unsorted)")
    print(f"Recommended: {result5['algorithm']}")
    print(f"Valid options: {result5['valid_algorithms']}")
    print(f"Note: Only linear_search is valid for unsorted arrays!")

    # Demo 6: Execute search
    arr6 = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
    idx, algo, explanation = recommender.search_array(arr6, 7)
    print(f"\nSearching for 7 in {arr6}")
    print(f"Found at index: {idx}")
    print(f"Used: {algo}")

    print("\n" + "-" * 70)
    print("GRAPH SEARCH DEMONSTRATIONS")
    print("-" * 70)

    # Demo 7: Unweighted graph
    graph1 = {
        0: [(1, 1.0), (2, 1.0)],
        1: [(3, 1.0)],
        2: [(3, 1.0)],
        3: []
    }
    result7 = recommender.recommend_graph_search(graph1)
    print(f"\nUnweighted graph: 0 -> 1,2; 1,2 -> 3")
    print(f"Recommended: {result7['algorithm']}")
    print(f"Valid options: {result7['valid_algorithms']}")

    # Demo 8: Weighted graph
    graph2 = {
        0: [(1, 4.0), (2, 1.0)],
        1: [(3, 1.0)],
        2: [(1, 2.0), (3, 5.0)],
        3: []
    }
    result8 = recommender.recommend_graph_search(graph2)
    print(f"\nWeighted graph (non-negative)")
    print(f"Recommended: {result8['algorithm']}")
    print(f"Valid options: {result8['valid_algorithms']}")

    # Demo 9: Graph with negative weights
    graph3 = {
        0: [(1, 4.0), (2, 5.0)],
        1: [(2, -3.0)],  # Negative weight!
        2: []
    }
    result9 = recommender.recommend_graph_search(graph3)
    print(f"\nGraph with NEGATIVE weights")
    print(f"Recommended: {result9['algorithm']}")
    print(f"Valid options: {result9['valid_algorithms']}")
    print(f"Note: Only bellman_ford can handle negative weights!")

    # Demo 10: Execute graph search
    path, cost, algo, explanation = recommender.search_graph(graph2, 0, 3)
    print(f"\nFinding path from 0 to 3")
    print(f"Path: {path}")
    print(f"Cost: {cost}")
    print(f"Used: {algo}")

    print("\n" + "=" * 70)
    print("DEMONSTRATION COMPLETE")
    print("=" * 70)


if __name__ == '__main__':
    demo()