app/analysis/summaries.py

from __future__ import annotations

import numpy as np

from app.core.schemas import TopEdge


def compute_outgoing_importance(matrix: np.ndarray) -> list[float]:
    sentence_count = matrix.shape[0]
    scores: list[float] = []
    for source_idx in range(sentence_count):
        column = matrix[source_idx + 1 :, source_idx]
        scores.append(float(column.mean()) if column.size else 0.0)
    return scores


def compute_incoming_importance(matrix: np.ndarray) -> list[float]:
    sentence_count = matrix.shape[0]
    scores: list[float] = []
    for target_idx in range(sentence_count):
        row = matrix[target_idx, :target_idx]
        scores.append(float(row.mean()) if row.size else 0.0)
    return scores


def compute_top_edges(matrix: np.ndarray, top_k: int = 10) -> list[TopEdge]:
    sentence_count = matrix.shape[0]
    candidates: list[TopEdge] = []
    for target_idx in range(sentence_count):
        for source_idx in range(target_idx):
            candidates.append(
                TopEdge(
                    source_sentence_idx=source_idx,
                    target_sentence_idx=target_idx,
                    score=float(matrix[target_idx, source_idx]),
                )
            )
    candidates.sort(key=lambda edge: edge.score, reverse=True)
    return candidates[:top_k]