让 config.py 中的设置项都生效

config.py

@@ -45,8 +45,6 @@ class ModelPaths:
     paraformer_path = Path() / 'models' / 'paraformer-offline-zh' / 'model.int8.onnx'
     tokens_path = Path() / 'models' / 'paraformer-offline-zh' / 'tokens.txt'
     punc_model_dir = Path() / 'models' / 'punc_ct-transformer_cn-en'
-
-
 
 
 class ParaformerArgs:

util/client_recv_result.py

@@ -3,6 +3,7 @@
 
 import keyboard
 import websockets
+from config import ClientConfig as Config
 from util.client_cosmic import Cosmic, console
 from util.client_check_websocket import check_websocket
 from util.client_hot_sub import hot_sub
@@ -37,11 +38,12 @@ async def recv_result():
             # 打字
             await type_result(text)
 
-            # 重命名录音文件
-            file_audio = rename_audio(message['task_id'], text, message['time_start'])
+            if Config.save_audio:
+                # 重命名录音文件
+                file_audio = rename_audio(message['task_id'], text, message['time_start'])
 
-            # 记录写入 md 文件
-            write_md(text, message['time_start'], file_audio)
+                # 记录写入 md 文件
+                write_md(text, message['time_start'], file_audio)
 
             # 控制台输出
             console.print(f'    转录时延：{delay:.2f}s')

util/client_rename_audio.py

@@ -1,7 +1,8 @@
-from util.client_cosmic import Cosmic, console
 from pathlib import Path
 from typing import Union
 import time
+from util.client_cosmic import Cosmic, console
+from config import ClientConfig as Config
 from os import makedirs
 
 
@@ -19,7 +20,7 @@ def rename_audio(task_id, text, time_start) -> Union[Path, None]:
     time_year = time.strftime('%Y', time.localtime(time_start))
     time_month = time.strftime('%m', time.localtime(time_start))
     time_ymdhms = time.strftime("%Y%m%d-%H%M%S", time.localtime(time_start))
-    file_stem = f'({time_ymdhms}){text[:20]}'
+    file_stem = f'({time_ymdhms}){text[:Config.audio_name_len]}'
 
     # 重命名
     file_path_new = file_path.with_name(file_stem + file_path.suffix)

util/client_send_audio.py

@@ -52,14 +52,14 @@ async def send_audio():
             Cosmic.queue_in.task_done()
             if task['type'] == 'begin':
                 time_start = task['time']
-            elif task['type'] ==  'data':
+            elif task['type'] == 'data':
                 # 在阈值之前积攒音频数据
                 if task['time'] - time_start < Config.threshold:
                     cache.append(task['data'])
                     continue
 
                 # 创建音频文件
-                if not file_path:
+                if Config.save_audio and not file_path:
                     file_path, file = create_file(task['data'].shape[1], time_start)
                     Cosmic.audio_files[task_id] = file_path
 
@@ -72,7 +72,8 @@ async def send_audio():
 
                 # 保存音频至本地文件
                 duration += len(data) / 48000
-                write_file(file, data)
+                if Config.save_audio:
+                    write_file(file, data)
 
                 # 发送音频数据用于识别
                 message = {
@@ -90,7 +91,8 @@ async def send_audio():
                 task = asyncio.create_task(send_message(message))
             elif task['type'] ==  'finish':
                 # 完成写入本地文件
-                finish_file(file)
+                if Config.save_audio:
+                    finish_file(file)
 
                 console.print(f'任务标识：{task_id}')
                 console.print(f'    录音时长：{duration:.2f}s')
@@ -109,4 +111,4 @@ async def send_audio():
                 task = asyncio.create_task(send_message(message))
                 break
     except Exception as e:
-        print(e)
+        print(e)

util/hot_kwds.py

@@ -1,4 +1,4 @@
-
+from config import ClientConfig as Config
 
 kwd_list = []
 
@@ -9,6 +9,11 @@ def do_updata_kwd(kwd_text: str):
     kwd_list.clear()
     kwd_list.append('')
 
+    # 如果不启用关键词功能，直接返回
+    if not Config.hot_kwd:
+        return len(kwd_list)
+
+    # 更新关键词
     for kwd in kwd_text.splitlines():
         kwd = kwd.strip()
         if not kwd or kwd.startswith('#'):

util/server_recognize.py

@@ -4,6 +4,7 @@
 import numpy as np 
 
 from util.server_cosmic import console
+from config import ServerConfig as Config
 from util.server_classes import Task, Result
 from util.chinese_itn import chinese_to_num
 from util.format_tools import adjust_space
@@ -13,6 +14,18 @@
 results = {}
 
 
+def format_text(text, punc_model):
+    if Config.format_spell:
+        text = adjust_space(text)       # 调空格
+    if Config.format_punc and punc_model and text:
+        text = punc_model(text)[0]  # 加标点
+    if Config.format_num:
+        text = chinese_to_num(text)     # 转数字
+    if Config.format_spell:
+        text = adjust_space(text)       # 调空格
+    return text
+
+
 def recognize(recognizer, punc_model, task: Task):
 
     # inspect({key:value for key, value in task.__dict__.items() if not key.startswith('_') and key != 'data'})
@@ -42,7 +55,7 @@ def recognize(recognizer, punc_model, task: Task):
     result.time_submit = task.time_submit
     result.time_complete = time.time()
 
-    # 先粗去重
+    # 先粗去重，依据：字级时间戳
     m = n = len(stream.result.timestamps)
     for i, timestamp in enumerate(stream.result.timestamps, start=0):
         if timestamp > task.overlap / 2: 
@@ -57,7 +70,7 @@ def recognize(recognizer, punc_model, task: Task):
     if task.is_final:
         n = len(stream.result.timestamps)
 
-    # 再细去重
+    # 再细去重，依据：在端点是否有重复的字
     if result.tokens and result.tokens[-2:] == stream.result.tokens[m:n][:2]:
         m += 2
     elif result.tokens and result.tokens[-1:] == stream.result.tokens[m:n][:1]:
@@ -67,7 +80,6 @@ def recognize(recognizer, punc_model, task: Task):
     result.timestamps += [t + task.offset for t in stream.result.timestamps[m:n]]
     result.tokens += [token for token in stream.result.tokens[m:n]]
 
-
     # token 合并为文本
     text = ' '.join(result.tokens).replace('@@ ', '')
     text = re.sub('([^a-zA-Z0-9]) (?![a-zA-Z0-9])', r'\1', text)
@@ -78,14 +90,7 @@ def recognize(recognizer, punc_model, task: Task):
         return result
 
     # 调整文本格式
-    text = adjust_space(text)       # 调空格
-    if punc_model and text:
-        text = punc_model(text)[0]  # 加标点
-    text = chinese_to_num(text)     # 转数字
-    text = adjust_space(text)       # 调空格
-
-    result.text = text
-
+    result.text = format_text(text, punc_model)
 
     # 若最后一个片段完成识别，从字典摘取任务
     result = results.pop(task.task_id)