#!/usr/bin/env python3
"""云蝠洞察录音导入示例：标准库、稳定分块请求、不自动重提未知的付费任务。"""

import argparse
import base64
import hashlib
import json
import os
import pathlib
import re
import sys
import time
import urllib.error
import urllib.parse
import urllib.request
import uuid


class NoRedirect(urllib.request.HTTPRedirectHandler):
    """密钥仅发送给明确指定的站点；部署重定向应先由操作者核对。"""

    def redirect_request(self, req, fp, code, msg, headers, newurl):
        return None


def main():
    parser = argparse.ArgumentParser(description='上传录音并查询识别状态；不自动确认角色或质检通过')
    parser.add_argument('file')
    parser.add_argument('--origin', required=True)
    parser.add_argument('--resource', default='')
    parser.add_argument('--model', choices=['fun-asr', 'qwen-audio-3.1-asr-flash-filetrans'], default='')
    parser.add_argument('--channels', type=int, choices=[1, 2], default=1)
    parser.add_argument('--source', default='API录音导入')
    parser.add_argument('--request-id', default=str(uuid.uuid4()))
    parser.add_argument('--job-id', default='', help='接续已经创建的清单/ZIP任务，不再调用create')
    parser.add_argument('--left-role', choices=['agent', 'customer', 'unknown'], default='unknown')
    parser.add_argument('--right-role', choices=['agent', 'customer', 'unknown'], default='unknown')
    parser.add_argument('--wait', type=int, default=0, help='最多等待秒数，0仅完成上传')
    args = parser.parse_args()
    key = os.environ.get('YUNFU_API_KEY')
    started = os.environ.get('YUNFU_RECORDING_STARTED_AT')
    file = pathlib.Path(args.file)
    target = urllib.parse.urlsplit(args.origin)
    if not key:
        parser.error('请先设置 YUNFU_API_KEY 环境变量')
    local_http = target.scheme == 'http' and target.hostname in ('127.0.0.1', 'localhost', '::1')
    if ((target.scheme != 'https' and not local_http) or not target.hostname or target.username or target.password
            or target.path not in ('', '/') or target.query or target.fragment):
        parser.error('origin必须为HTTPS站点根地址（本机回环可用HTTP），不能包含账号、路径或查询参数')
    if not re.fullmatch(r'[\w-]{8,100}', args.request_id, flags=re.ASCII):
        parser.error('request-id需为8–100位字母、数字、下划线或短横线')
    if not file.is_file() or not 0 < file.stat().st_size <= 100 * 1024 * 1024:
        parser.error('单条文件需为1字节至100MB')
    # 重跑同一 request-id 时文件元数据、开始时间及每个分块必须保持一致。
    if not started:
        parser.error('请设置带时区的 YUNFU_RECORDING_STARTED_AT，例如2026-10-17T09:00:00+08:00')
    origin = args.origin.rstrip('/')
    opener = urllib.request.build_opener(NoRedirect())

    def request(action, body=None, slot=''):
        path = '/v1/operations/insights.audio.' + action
        if body is None:
            path += '?' + urllib.parse.urlencode({'jobId': args.job_id})
        identity = args.request_id + ':' + action + ':' + slot
        headers = {
            'Authorization': 'Bearer ' + key,
            'Content-Type': 'application/json',
            'Idempotency-Key': hashlib.sha256(identity.encode()).hexdigest(),
        }
        req = urllib.request.Request(origin + path, data=None if body is None else json.dumps(body).encode(), headers=headers)
        try:
            with opener.open(req, timeout=135) as response:
                return json.load(response)
        except urllib.error.HTTPError as error:
            try:
                detail = json.loads(error.read(4096)).get('error', 'HTTP ' + str(error.code))
            except (ValueError, AttributeError):
                detail = 'HTTP ' + str(error.code)
            raise RuntimeError(str(detail)) from None

    metadata = {
        'requestId': args.request_id, 'filename': file.name, 'bytes': file.stat().st_size,
        'source': args.source, 'startedAt': started, 'channels': args.channels,
    }
    if args.resource:
        metadata['asrResourceId'] = args.resource
    if args.model:
        metadata['asrModel'] = args.model
    if args.channels == 2 and (args.left_role != 'unknown' or args.right_role != 'unknown'):
        metadata['channelMapping'] = {'left': args.left_role, 'right': args.right_role}
    job = request('read')['job'] if args.job_id else request('create', metadata)['job']
    if job['filename'] != file.name or job['bytes'] != file.stat().st_size:
        raise RuntimeError('接续文件名/大小与已有任务不一致；请核对原文件，不会覆盖任务')
    if job['status'] == 'uploading':
        with file.open('rb') as recording:
            index = 0
            while True:
                part = recording.read(1024 * 1024)
                if not part:
                    break
                request('chunk', {'jobId': job['id'], 'index': index,
                                  'data': base64.b64encode(part).decode()}, str(index))
                index += 1
        job = request('complete', {'jobId': job['id']})['job']
    print(json.dumps({'requestId': args.request_id, 'job': job}, ensure_ascii=False))
    if job.get('archiveChildren'):
        print('ZIP已解包；请逐条read查询archiveChildren，父任务completed不代表子任务已转写。', file=sys.stderr)
    end = time.monotonic() + min(max(args.wait, 0), 3600)
    while time.monotonic() < end and job['status'] in ('queued', 'recognizing', 'submitting'):
        time.sleep(5)
        # 每次推进是独立请求；结果未知时停止，不自动重新调用供应商识别。
        job = request('advance', {'jobId': job['id']}, str(uuid.uuid4()))['job']
        print(json.dumps({'job': job}, ensure_ascii=False))
    if job['status'] in ('failed', 'uncertain'):
        print('请先核对任务与供应商费用，脚本不会自动重提。', file=sys.stderr)
    return 0


if __name__ == '__main__':
    try:
        raise SystemExit(main())
    except (RuntimeError, OSError, ValueError) as error:
        print(str(error), file=sys.stderr)
        raise SystemExit(1)
