298 lines
12 KiB
Python
298 lines
12 KiB
Python
import json
|
||
import logging
|
||
import time
|
||
import uuid
|
||
import oss2
|
||
from aliyunsdkcore.client import AcsClient
|
||
from aliyunsdkcore.acs_exception.exceptions import ClientException, ServerException
|
||
# 尝试导入最新的 API 版本，如果有问题可能需要调整
|
||
try:
|
||
from aliyunsdktingwu.request.v20230930 import CreateTaskRequest, GetTaskInfoRequest
|
||
except ImportError:
|
||
# Fallback or error handling if version differs
|
||
pass
|
||
|
||
from django.conf import settings
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
class AliyunTingwuService:
|
||
def __init__(self):
|
||
self.access_key_id = settings.ALIYUN_ACCESS_KEY_ID
|
||
self.access_key_secret = settings.ALIYUN_ACCESS_KEY_SECRET
|
||
self.oss_bucket_name = settings.ALIYUN_OSS_BUCKET_NAME
|
||
self.oss_endpoint = settings.ALIYUN_OSS_ENDPOINT
|
||
self.tingwu_app_key = settings.ALIYUN_TINGWU_APP_KEY
|
||
self.region_id = "cn-shanghai" # 听悟服务区域，根据文档应与OSS区域一致，或者使用 'cn-beijing'
|
||
|
||
# 初始化 OSS Bucket
|
||
if self.access_key_id and self.access_key_secret and self.oss_endpoint:
|
||
auth = oss2.Auth(self.access_key_id, self.access_key_secret)
|
||
self.bucket = oss2.Bucket(auth, self.oss_endpoint, self.oss_bucket_name)
|
||
else:
|
||
self.bucket = None
|
||
logger.warning("Aliyun OSS configuration missing.")
|
||
|
||
# 初始化听悟 Client
|
||
if self.access_key_id and self.access_key_secret:
|
||
self.client = AcsClient(
|
||
self.access_key_id,
|
||
self.access_key_secret,
|
||
self.region_id
|
||
)
|
||
# 显式添加听悟服务的 Endpoint 映射，解决 EndpointResolvingError
|
||
# 听悟 API 的服务接入点通常是 tingwu.cn-beijing.aliyuncs.com
|
||
# 但新版听悟 API (tingwu.aliyuncs.com) 可能不同，需根据实际情况添加
|
||
# 这里添加一个通用的 Endpoint 映射
|
||
try:
|
||
# 尝试为 tingwu 产品设置 Endpoint
|
||
# 注意：听悟服务主要部署在北京，Endpoint 通常为 tingwu.cn-beijing.aliyuncs.com
|
||
# 如果您的服务在上海，也可能需要连接到北京的接入点
|
||
self.client.add_endpoint(self.region_id, "tingwu", "tingwu.cn-beijing.aliyuncs.com")
|
||
except Exception as e:
|
||
logger.warning(f"Failed to add endpoint: {e}")
|
||
|
||
else:
|
||
self.client = None
|
||
logger.warning("Aliyun AccessKey configuration missing.")
|
||
|
||
def upload_to_oss(self, file_obj, file_name):
|
||
"""
|
||
上传文件到 OSS 并返回带签名的 URL (有效期 3 小时)
|
||
"""
|
||
if not self.bucket:
|
||
raise Exception("OSS Client not initialized")
|
||
|
||
try:
|
||
# 上传文件
|
||
# file_obj 应该是打开的文件对象或字节流
|
||
self.bucket.put_object(file_name, file_obj)
|
||
|
||
# 生成签名 URL，有效期 3 小时 (3600 * 3)
|
||
url = self.bucket.sign_url('GET', file_name, 3600 * 3)
|
||
return url
|
||
except Exception as e:
|
||
logger.error(f"OSS Upload failed: {e}")
|
||
raise e
|
||
|
||
def create_transcription_task(self, file_url, language="cn"):
|
||
"""
|
||
创建听悟转写任务
|
||
"""
|
||
if not self.client:
|
||
raise Exception("Tingwu Client not initialized")
|
||
|
||
request = CreateTaskRequest.CreateTaskRequest()
|
||
|
||
# 针对阿里云 SDK 不同版本的兼容性处理
|
||
# "type" 参数是听悟 API (ROA 风格) 的必填项，用于指定任务类型
|
||
# 根据官方文档，离线任务的 type 通常就是 'offline'
|
||
request.add_query_param('type', 'offline')
|
||
|
||
# 构造请求体 (Body)
|
||
# 根据听悟 API 文档，AppKey, Input, Parameters 应位于 JSON Body 中
|
||
# 而不是 Query Parameter
|
||
body = {
|
||
"AppKey": self.tingwu_app_key,
|
||
"Input": {
|
||
"FileUrl": file_url,
|
||
"SourceLanguage": language,
|
||
"TaskKey": str(uuid.uuid4())
|
||
},
|
||
"Parameters": {
|
||
"Transcoding": {
|
||
"TargetAudioFormat": "mp3"
|
||
},
|
||
"Transcription": {
|
||
"DiarizationEnabled": True,
|
||
"ChannelId": 0
|
||
},
|
||
"TranscriptionEnabled": True,
|
||
"AutoChaptersEnabled": True,
|
||
"SummarizationEnabled": True,
|
||
"Summarization": {
|
||
"Types": ["Paragraph", "Conversational", "QuestionsAnswering", "MindMap"]
|
||
}
|
||
}
|
||
}
|
||
|
||
# 设置 Body 内容
|
||
request.set_content(json.dumps(body))
|
||
request.add_header('Content-Type', 'application/json')
|
||
|
||
# 强制设置 Endpoint，避免 SDK.EndpointResolvingError
|
||
# 听悟目前主要服务点在北京
|
||
request.set_endpoint("tingwu.cn-beijing.aliyuncs.com")
|
||
|
||
# 显式设置 Method 为 PUT
|
||
request.set_method('PUT')
|
||
|
||
try:
|
||
response = self.client.do_action_with_exception(request)
|
||
return json.loads(response)
|
||
except (ClientException, ServerException) as e:
|
||
logger.error(f"Tingwu CreateTask failed: {e}")
|
||
raise e
|
||
|
||
def get_task_info(self, task_id):
|
||
"""
|
||
查询任务状态和结果
|
||
"""
|
||
if not self.client:
|
||
raise Exception("Tingwu Client not initialized")
|
||
|
||
request = GetTaskInfoRequest.GetTaskInfoRequest()
|
||
request.set_TaskId(task_id)
|
||
|
||
try:
|
||
response = self.client.do_action_with_exception(request)
|
||
return json.loads(response)
|
||
except (ClientException, ServerException) as e:
|
||
logger.error(f"Tingwu GetTaskInfo failed: {e}")
|
||
raise e
|
||
|
||
def parse_and_update_task(self, task, result):
|
||
"""
|
||
解析听悟结果并更新任务
|
||
:param task: TranscriptionTask 实例
|
||
:param result: get_task_info 返回的完整 JSON (或 Data 部分)
|
||
"""
|
||
# 1. 提取 Data 对象
|
||
if isinstance(result, dict):
|
||
data_obj = result.get('Data', result)
|
||
else:
|
||
data_obj = result
|
||
|
||
if not isinstance(data_obj, dict):
|
||
logger.error(f"Unexpected data format: {type(data_obj)}")
|
||
return
|
||
|
||
# 2. 更新状态
|
||
task_status = data_obj.get('TaskStatus') or data_obj.get('Status')
|
||
if task_status in ['COMPLETE', 'COMPLETED', 'SUCCEEDED']:
|
||
task.status = 'SUCCEEDED' # 使用字符串引用，避免导入模型循环引用
|
||
elif task_status == 'FAILED':
|
||
task.status = 'FAILED'
|
||
task.error_message = data_obj.get('TaskStatusText', data_obj.get('Message', 'Unknown error'))
|
||
task.save()
|
||
return
|
||
else:
|
||
# 仍在处理中，不更新内容
|
||
return
|
||
|
||
# 3. 解析结果
|
||
task_result = data_obj.get('Result', {})
|
||
|
||
# --- A. 处理逐字稿 (Transcription) ---
|
||
transcription_data = task_result.get('Transcription', {})
|
||
|
||
# 处理 URL 下载
|
||
if isinstance(transcription_data, str) and transcription_data.startswith('http'):
|
||
try:
|
||
import requests
|
||
t_resp = requests.get(transcription_data)
|
||
if t_resp.status_code == 200:
|
||
transcription_data = t_resp.json()
|
||
except Exception as e:
|
||
logger.error(f"Download transcription failed: {e}")
|
||
transcription_data = {}
|
||
elif isinstance(transcription_data, dict) and 'TranscriptionUrl' in transcription_data:
|
||
try:
|
||
import requests
|
||
t_resp = requests.get(transcription_data['TranscriptionUrl'])
|
||
if t_resp.status_code == 200:
|
||
transcription_data = t_resp.json()
|
||
except Exception as e:
|
||
logger.error(f"Download transcription url failed: {e}")
|
||
|
||
# 保存原始数据
|
||
task.transcription_data = transcription_data
|
||
|
||
# 提取文本
|
||
# 结构: {"Transcription": {"Paragraphs": [{"Words": [{"Text": "..."}]}]}}
|
||
# 或直接 {"Paragraphs": ...}
|
||
content_source = transcription_data
|
||
if 'Transcription' in content_source and isinstance(content_source['Transcription'], dict):
|
||
content_source = content_source['Transcription']
|
||
|
||
paragraphs = content_source.get('Paragraphs', [])
|
||
full_text_lines = []
|
||
|
||
if paragraphs and isinstance(paragraphs, list):
|
||
for p in paragraphs:
|
||
# 尝试从 Words 中提取
|
||
words = p.get('Words', [])
|
||
if words:
|
||
line_text = "".join([str(w.get('Text', '')) for w in words])
|
||
full_text_lines.append(line_text)
|
||
# 兼容旧结构或直接 Text
|
||
elif 'Text' in p:
|
||
full_text_lines.append(p['Text'])
|
||
|
||
if full_text_lines:
|
||
task.transcription = "\n".join(full_text_lines)
|
||
|
||
# --- B. 处理 AI 总结 (Summarization) ---
|
||
summarization = task_result.get('Summarization', {})
|
||
|
||
# 处理 URL 下载
|
||
if isinstance(summarization, str) and summarization.startswith('http'):
|
||
try:
|
||
import requests
|
||
s_resp = requests.get(summarization)
|
||
if s_resp.status_code == 200:
|
||
summarization = s_resp.json()
|
||
except Exception as e:
|
||
logger.error(f"Download summarization failed: {e}")
|
||
summarization = {}
|
||
|
||
# 保存原始数据
|
||
task.summary_data = summarization
|
||
|
||
# 提取文本 (MindMapSummary)
|
||
# 结构: {"MindMapSummary": [{"Title": "...", "Topic": [...]}]}
|
||
summary_text = []
|
||
|
||
def parse_mindmap_topic(topic_list, level=0):
|
||
indent = " " * level
|
||
for topic in topic_list:
|
||
title = topic.get('Title', '')
|
||
if title:
|
||
summary_text.append(f"{indent}- {title}")
|
||
|
||
sub_topics = topic.get('Topic', [])
|
||
if sub_topics:
|
||
parse_mindmap_topic(sub_topics, level + 1)
|
||
|
||
if 'MindMapSummary' in summarization:
|
||
parse_mindmap_topic(summarization['MindMapSummary'])
|
||
elif 'Text' in summarization:
|
||
summary_text.append(summarization['Text'])
|
||
elif 'Headline' in summarization:
|
||
summary_text.append(summarization['Headline'])
|
||
|
||
if summary_text:
|
||
task.summary = "\n".join(summary_text)
|
||
|
||
# --- C. 处理章节 (AutoChapters) ---
|
||
auto_chapters = task_result.get('AutoChapters', [])
|
||
|
||
# 处理 URL 下载
|
||
if isinstance(auto_chapters, str) and auto_chapters.startswith('http'):
|
||
try:
|
||
import requests
|
||
ac_resp = requests.get(auto_chapters)
|
||
if ac_resp.status_code == 200:
|
||
auto_chapters = ac_resp.json()
|
||
except Exception as e:
|
||
logger.error(f"Download auto chapters failed: {e}")
|
||
auto_chapters = []
|
||
|
||
# 保存原始数据
|
||
task.auto_chapters_data = auto_chapters
|
||
|
||
# (可选) 将章节信息追加到 summary 或 evaluation 中，或者仅保存 raw data
|
||
# 根据用户需求，这里主要保存到 model 的 auto_chapters_data 字段 (已在 models.py 定义)
|
||
|
||
task.save()
|