add new logic of check text length

transcribe/strategy.py

@@ -160,7 +160,7 @@ class TranscriptBuffer:
 
     def __init__(self, source_lang:str, separator:str):
         self._segments: List[str] = collections.deque(maxlen=2)     # 确认的完整段落
-        self._sentences: List[str] = []    # 当前段落中的短句
+        self._sentences: List[str] = collections.deque()   # 当前段落中的短句
         self._buffer: str = ""             # 当前缓冲中的文本
         self._current_seg_id: int = 0
         self.source_language = source_lang
@@ -189,17 +189,27 @@ class TranscriptBuffer:
         Args:
             end_of_sentence: 是否为句子结尾（如检测到句号）
         """
-        if self._sentences:
-            self._segments.append("".join(self._sentences))
-            self._sentences.clear()
+        
+        count = 0
+        current_sentences = []
+        while len(self._sentences) and count <=20:
+            item = self._sentences.popleft()
+            current_sentences.append(item)
+            count += len(item.split(self._separator))
+        if current_sentences:
+            self._segments.append("".join(current_sentences))
+        logger.debug(f"=== count to paragraph ===")
+        logger.debug(f"push: {current_sentences}")
+        logger.debug(f"rest: {self._sentences}")
+        # if self._sentences:
+        #     self._segments.append("".join(self._sentences))
+        #     self._sentences.clear()
 
     def rebuild(self, text):
         output = self.split_and_join(
                     text.replace(
                         self._separator, ""))
-        if output != text:
-            logger.debug(f"Source string: {text.replace(self._separator, '')}")
-            logger.debug(f"Output string: {output}")
+
         return output
     
     @staticmethod
@@ -241,7 +251,7 @@ class TranscriptBuffer:
             current_text_len =  len(self.current_not_commit_text.split(self._separator)) if self._separator else len(self.current_not_commit_text)
             # current_text_len = len(self.current_not_commit_text.split(self._separator))
             self.update_pending_text(remaining_string)
-            if current_text_len >=20:
+            if current_text_len >= 20:
                 self.commit_paragraph()
                 self._current_seg_id += 1
                 return True
@@ -255,7 +265,7 @@ class TranscriptBuffer:
     @property
     def un_commit_paragraph(self) -> str:
         """当前短句组合"""
-        return "".join(self._sentences)
+        return "".join([i for i in self._sentences])
 
     @property
     def pending_text(self) -> str:
@@ -299,7 +309,7 @@ class TranscriptStabilityAnalyzer:
             return
 
         # yield from self._handle_short_buffer(current, prev)
-        if buffer_duration < 4:
+        if buffer_duration <= 4:
             yield from self._handle_short_buffer(current, prev)
         else:
             yield from self._handle_long_buffer(current)