"""
Blog Similarity Finder
======================
讀取 blog_embeddings.json，用模糊搜尋選一篇文章，
列出最相似的 10 篇和最不相似的 10 篇。

需求：
  - pip install numpy iterfzf
  - 已經跑過 blog_embeddings.py 產生 blog_embeddings.json
"""

import json
import numpy as np
from iterfzf import iterfzf

# ============================================================
# 設定區
# ============================================================

# embedding 檔案路徑（blog_embeddings.py 的輸出）
EMBEDDINGS_FILE = "./blog_embeddings.json"

# ============================================================


def cosine_similarity(a: np.ndarray, b: np.ndarray) -> float:
    """計算兩個向量的 cosine similarity。"""
    dot = np.dot(a, b)
    norm = np.linalg.norm(a) * np.linalg.norm(b)
    if norm == 0:
        return 0.0
    return float(dot / norm)


def format_row(rank: int, sim: float, title: str, slug: str) -> str:
    """格式化一行結果。"""
    bar = "█" * int(sim * 30)
    return f"  {rank:3d}. {sim:.4f}  {bar}  {title}  ({slug})"


def main():
    # 讀取 embeddings
    with open(EMBEDDINGS_FILE, encoding="utf-8") as f:
        data = json.load(f)

    if not data:
        print("找不到任何文章資料。")
        return

    # 建立選項：title + slug，方便搜尋
    choices = []
    for item in data:
        title = item.get("title", "(無標題)")
        choices.append(f"{title}  |  {item['slug']}")

    # 模糊搜尋選擇
    selected = iterfzf(choices, prompt="選擇文章 > ")

    if not selected:
        return

    idx = choices.index(selected)
    target = data[idx]
    target_vec = np.array(target["embedding"])
    target_title = target.get("title", "(無標題)")

    print(f"\n以「{target_title}」為基準：\n")

    # 計算與所有其他文章的相似度
    similarities = []
    for i, item in enumerate(data):
        if i == idx:
            continue
        sim = cosine_similarity(target_vec, np.array(item["embedding"]))
        similarities.append({
            "slug": item["slug"],
            "title": item.get("title", "(無標題)"),
            "similarity": sim,
        })

    # 由高到低排序
    similarities.sort(key=lambda x: x["similarity"], reverse=True)

    # 最相似 10 篇
    top = similarities[:20]
    print("🔥 最相似的 20 篇：")
    for rank, s in enumerate(top, 1):
        print(format_row(rank, s["similarity"], s["title"], s["slug"]))

    # 最不相似 10 篇
    bottom = similarities[-5:]
    bottom.reverse()
    print(f"\n🧊 最不相似的 5 篇：")
    for rank, s in enumerate(bottom, 1):
        print(format_row(rank, s["similarity"], s["title"], s["slug"]))


if __name__ == "__main__":
    main()