#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""analyze-model-lang-matrix.py — 解析 sysvoices-dom,生成 model × lang × voice 矩阵"""
import re, json, sys
from collections import defaultdict

PATH = r"C:/Users/caoyg/AppData/Local/Temp/sysvoices-dom.txt"
text = open(PATH, encoding='utf-8').read()

# 阿里云页面有 4 个章节(每个章节都重复了"本页导读"导航,所以取第一个匹配即可)
SECTIONS = [
    ('cosyvoice-v3-flash',    r'cosyvoice-v3-flash\s*音色列表'),
    ('cosyvoice-v3-plus',     r'cosyvoice-v3-plus\s*音色列表'),
    ('cosyvoice-v2',          r'cosyvoice-v2\s*音色列表'),
    ('cosyvoice-v1',          r'cosyvoice-v1\s*音色列表'),
]

# 取每个模型标题第一个出现的位置(跳过右侧栏的目录)
positions = []
for model, pat in SECTIONS:
    matches = list(re.finditer(pat, text))
    # 选最小位置的(主体内容不是侧栏目录)
    if matches:
        m = min(matches, key=lambda x: x.start())
        positions.append((m.start(), model))

positions.sort()
all_voices = defaultdict(list)
all_voices_per_model = {}

for i, (start, model) in enumerate(positions):
    end = positions[i + 1][0] if i + 1 < len(positions) else len(text)
    section = text[start:end]
    voices = []
    pos = [mm.start() for mm in re.finditer(r'名称[：:]\s*.+?\n', section)]
    pos.append(len(section))
    for j in range(len(pos) - 1):
        chunk = section[pos[j]:pos[j + 1]]
        v = {}
        n = re.search(r'^名称[：:]\s*(.+?)\s*$', chunk, re.MULTILINE)
        if n: v['name'] = n.group(1).strip()
        vp = re.search(r'voice参数[：:]\s*(\w+)', chunk)
        if vp: v['voice_id'] = vp.group(1)
        t = re.search(r'特质[：:]\s*(.+?)\n', chunk)
        if t: v['desc'] = t.group(1).strip()[:80]
        a = re.search(r'年龄[：:]\s*(.+?)\n', chunk)
        if a: v['age'] = a.group(1).strip()
        l = re.search(r'语言[：:]\s*(.+?)\n', chunk)
        if l: v['lang'] = l.group(1).strip()[:60]
        v['instruct'] = '支持' if 'Instruct：支持' in chunk else '不支持' if 'Instruct：不支持' in chunk else '?'
        if v.get('name'):
            voices.append(v)
    all_voices_per_model[model] = voices

# 输出汇总
sys.stdout.reconfigure(encoding='utf-8')
total_voices = 0
all_langs = set()
for model, voices in all_voices_per_model.items():
    print(f'\n=== {model}: {len(voices)} 个音色 ===')
    lang_count = defaultdict(int)
    for v in voices:
        lang = v.get('lang', '?')
        for l in re.split('[、，]', lang):
            l = l.strip()
            if not l: continue
            all_langs.add(l)
            lang_count[l] += 1
    for l, c in sorted(lang_count.items(), key=lambda x: (-x[1], x[0])):
        print(f'    {l}: {c}')
    total_voices += len(voices)

print(f'\n========== 总览 ==========')
print(f'总音色数: {total_voices}')
print(f'涉及语种 ({len(all_langs)} 种):')
for l in sorted(all_langs): print(f'  - {l}')

# 写 JSON 给前端用
with open(r"C:/Users/caoyg/AppData/Local/Temp/alicloud-voices-matrix.json", 'w', encoding='utf-8') as f:
    json.dump(all_voices_per_model, f, ensure_ascii=False, indent=2)
print(f'\nJSON: {len(all_voices_per_model)} 个模型 → C:/Users/caoyg/AppData/Local/Temp/alicloud-voices-matrix.json')