smallestai 1.3.0__py3-none-any.whl → 1.3.1__py3-none-any.whl

smallest/async_tts.py

@@ -6,7 +6,7 @@ from typing import Optional, Union, List
 
 from .models import TTSModels, TTSVoices
 from .exceptions import TTSError, APIError
-from .utils import (TTSOptions, validate_input, preprocess_text, add_wav_header,
+from .utils import (TTSOptions, validate_input, preprocess_text, add_wav_header, split_into_chunks,
                      get_smallest_languages, get_smallest_voices, get_smallest_models, SENTENCE_END_REGEX, API_BASE_URL)
 
 
@@ -47,7 +47,7 @@ class AsyncSmallest:
         """
         self.api_key = api_key or os.environ.get("SMALLEST_API_KEY")
         if not self.api_key:
-            raise TTSError("API key is required")
+            raise TTSError()
         self.chunk_size = 250
         
         self.opts = TTSOptions(
@@ -71,47 +71,6 @@ class AsyncSmallest:
         if self.session:
             await self.session.close()
 
-    def _split_into_chunks(self, text: str) -> List[str]:
-        """
-        Splits the input text into chunks based on sentence boundaries and the maximum chunk size.
-        """
-        chunks = []
-        current_chunk = ""
-        last_break_index = 0
-
-        i = 0
-        while i < len(text):
-            current_chunk += text[i]
-
-            if text[i] in ".,":
-                last_break_index = i
-
-            if len(current_chunk) >= self.chunk_size:
-                if last_break_index > 0:
-                    chunk = text[:last_break_index + 1].strip()
-                    chunk = chunk.replace("—", " ")
-                    chunks.append(chunk)
-                
-                    text = text[last_break_index + 1:]
-                    i = -1
-                    current_chunk = ""
-                    last_break_index = 0
-                else:
-                    # No break point found, split at max length
-                    current_chunk = current_chunk.replace("—", " ")
-                    chunks.append(current_chunk.strip())
-                    text = text[self.chunk_size:]
-                    i = -1
-                    current_chunk = ""
-
-            i += 1
-
-        if text:
-            text = text.replace("—", " ")
-            chunks.append(text.strip())
-
-        return chunks
-
 
     def get_languages(self) -> List[str]:
         """Returns a list of available languages."""
@@ -153,7 +112,7 @@ class AsyncSmallest:
             setattr(opts, key, value)
 
         validate_input(text, opts.voice, opts.model, opts.sample_rate, opts.speed)
-        chunks = self._split_into_chunks(text)
+        chunks = split_into_chunks(text)
         audio_content = b""
 
         for chunk in chunks:

smallest/exceptions.py

@@ -1,6 +1,9 @@
 class TTSError(Exception):
     """Base exception for TTS SDK"""
-    pass
+    default_message = "API key is required. Please set the `SMALLEST_API_KEY` environment variable or visit https://waves.smallest.ai/ to obtain your API key."
+    
+    def __init__(self, message=None):
+        super().__init__(message or self.default_message)
 
 class APIError(TTSError):
     """Raised when the API returns an error"""

smallest/stream_tts.py

@@ -46,20 +46,40 @@ class TextToAudioStream:
 
     async def _stream_llm_output(self, llm_output: AsyncGenerator[str, None]) -> None:
         """
-        Streams the LLM output, splitting it into sentences and adding each to the queue.
+        Streams the LLM output, splitting it into sentences based on the regex 
+        and chunk size, and adding each chunk to the queue.
 
         Parameters:
         - llm_output (AsyncGenerator[str, None]): An async generator yielding LLM output.
         """
         buffer = ""
+        last_break_index = 0
+
         async for chunk in llm_output:
             buffer += chunk
-            if self.sentence_end_regex.match(buffer) or len(buffer) > self.buffer_size:
-                self.queue.put(buffer)
-                buffer = ""
+            i = 0
+
+            while i < len(buffer):
+                current_chunk = buffer[:i + 1]
+                if self.sentence_end_regex.match(current_chunk):
+                    last_break_index = i
+
+                if len(current_chunk) >= self.buffer_size:
+                    if last_break_index > 0:
+                        self.queue.put(buffer[:last_break_index + 1].replace("—", " ").strip())
+                        buffer = buffer[last_break_index + 1:] 
+                    else:
+                        # No sentence boundary, split at max chunk size
+                        self.queue.put(buffer[:self.buffer_size].replace("—", " ").strip())
+                        buffer = buffer[self.buffer_size:] 
+
+                    last_break_index = 0
+                    i = -1 
+
+                i += 1
 
         if buffer:
-            self.queue.put(buffer)
+            self.queue.put(buffer.replace("—", " ").strip())
 
         self.stop_flag = True  # completion flag when LLM output ends
 

smallest/tts.py

@@ -6,7 +6,7 @@ from typing import Optional, Union, List
 
 from .models import TTSModels, TTSVoices
 from .exceptions import TTSError, APIError
-from .utils import (TTSOptions, validate_input, preprocess_text, add_wav_header, 
+from .utils import (TTSOptions, validate_input, preprocess_text, add_wav_header, split_into_chunks,
 get_smallest_languages, get_smallest_voices, get_smallest_models, SENTENCE_END_REGEX, API_BASE_URL)
 
 class Smallest:
@@ -45,7 +45,7 @@ class Smallest:
         """
         self.api_key = api_key or os.environ.get("SMALLEST_API_KEY")
         if not self.api_key:
-            raise TTSError("API key is required")
+            raise TTSError()
         
         self.chunk_size = 250
         
@@ -59,47 +59,6 @@ class Smallest:
             transliterate=transliterate,
             remove_extra_silence=remove_extra_silence
         )
-
-    def _split_into_chunks(self, text: str) -> List[str]:
-        """
-        Splits the input text into chunks based on sentence boundaries and the maximum chunk size.
-        """
-        chunks = []
-        current_chunk = ""
-        last_break_index = 0
-
-        i = 0
-        while i < len(text):
-            current_chunk += text[i]
-
-            if text[i] in ".,":
-                last_break_index = i
-
-            if len(current_chunk) >= self.chunk_size:
-                if last_break_index > 0:
-                    chunk = text[:last_break_index + 1].strip()
-                    chunk = chunk.replace("—", " ")
-                    chunks.append(chunk)
-                
-                    text = text[last_break_index + 1:]
-                    i = -1
-                    current_chunk = ""
-                    last_break_index = 0
-                else:
-                    # No break point found, split at max length
-                    current_chunk = current_chunk.replace("—", " ")
-                    chunks.append(current_chunk.strip())
-                    text = text[self.chunk_size:]
-                    i = -1
-                    current_chunk = ""
-
-            i += 1
-
-        if text:
-            text = text.replace("—", " ")
-            chunks.append(text.strip())
-
-        return chunks
     
         
     def get_languages(self) -> List[str]:
@@ -143,7 +102,7 @@ class Smallest:
 
         validate_input(text, opts.voice, opts.model, opts.sample_rate, opts.speed)
 
-        chunks = self._split_into_chunks(text)
+        chunks = split_into_chunks(text)
         audio_content = b""
 
         for chunk in chunks:

smallest/utils.py

@@ -59,6 +59,50 @@ def preprocess_text(text: str) -> str:
     text = mpn.normalize(text)
     return text.strip()
 
+def split_into_chunks(self, text: str) -> List[str]:
+        """
+        Splits the input text into chunks based on sentence boundaries 
+        defined by SENTENCE_END_REGEX and the maximum chunk size.
+        """
+        chunks = []
+        current_chunk = ""
+        last_break_index = 0
+
+        i = 0
+        while i < len(text):
+            current_chunk += text[i]
+
+            # Check for sentence boundary using regex
+            if SENTENCE_END_REGEX.match(current_chunk):
+                last_break_index = i
+
+            if len(current_chunk) >= self.chunk_size:
+                if last_break_index > 0:
+                    # Split at the last valid sentence boundary
+                    chunk = text[:last_break_index + 1].strip()
+                    chunk = chunk.replace("—", " ")
+                    chunks.append(chunk)
+
+                    text = text[last_break_index + 1:]
+                    i = -1  # Reset index to process the remaining text
+                    current_chunk = ""
+                    last_break_index = 0
+                else:
+                    # No sentence boundary found, split at max length
+                    current_chunk = current_chunk.replace("—", " ")
+                    chunks.append(current_chunk.strip())
+                    text = text[self.chunk_size:]
+                    i = -1  # Reset index to process the remaining text
+                    current_chunk = ""
+
+            i += 1
+
+        if text:
+            text = text.replace("—", " ")
+            chunks.append(text.strip())
+
+        return chunks
+
 
 def get_smallest_languages() -> List[str]:
     return list(TTSLanguages.__args__)

{smallestai-1.3.0.dist-info → smallestai-1.3.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: smallestai
-Version: 1.3.0
+Version: 1.3.1
 Summary: Official Python client for the Smallest AI API
 Author-email: Smallest <info@smallest.ai>
 License: MIT
@@ -104,7 +104,7 @@ if __name__ == "__main__":
 - `transliterate`: Enable text transliteration (default: False)
 - `remove_extra_silence`: Remove additional silence (default: True)  
 
-These parameters are part of the `Smallest` instance. They can be set when creating the instance (as shown above). However, the `synthesize` function also accepts kwargs, allowing you to override these parameters for a specific synthesis request.
+These parameters are part of the `Smallest` instance. They can be set when creating the instance (as shown above). However, the `synthesize` function also accepts `kwargs`, allowing you to override these parameters for a specific synthesis request.
 
 For example, you can modify the speech speed and sample rate just for a particular synthesis call:  
 ```py
@@ -149,7 +149,7 @@ if __name__ == "__main__":
 - `transliterate`: Enable text transliteration (default: False)
 - `remove_extra_silence`: Remove additional silence (default: True)  
 
-These parameters are part of the AsyncSmallest instance. They can be set when creating the instance (as shown above). However, the synthesize function also accepts kwargs, allowing you to override any of these parameters on a per-request basis.
+These parameters are part of the `AsyncSmallest` instance. They can be set when creating the instance (as shown above). However, the `synthesize` function also accepts `kwargs`, allowing you to override any of these parameters on a per-request basis.
 
 For example, you can modify the speech speed and sample rate just for a particular synthesis request:  
 ```py

smallestai-1.3.1.dist-info/RECORD

@@ -0,0 +1,12 @@
+smallest/__init__.py,sha256=vaoIBml_IobavpVvFazB86iikg2iEy4h3ddxqv_0Fy4,190
+smallest/async_tts.py,sha256=ReYTePzR0e3UcWxwnetGnwO4q_F7g2LRZPxIVfmgB4Y,6120
+smallest/exceptions.py,sha256=nY6I8fCXe2By54CytQ0-i3hFiYtt8TYAKj0g6OYsCjc,585
+smallest/models.py,sha256=R5UZZA9SibrJ2DsWPi_mkKI13WfyC-MLd-7kptfjns4,390
+smallest/stream_tts.py,sha256=0OypcUzgP7CN3VGcGJDnQ2FDw2JOzPaSQ1cXK69k5dY,6198
+smallest/tts.py,sha256=l8VHaOE8-Feg3Ey8C3osOrLs3ffYz0q_J1ACiEtZ8y0,5999
+smallest/utils.py,sha256=Xg4sYqhTgEgmT6H0qMNmlVlF1ilqY2BUCBTlk-_yONg,3795
+smallestai-1.3.1.dist-info/LICENSE,sha256=kK3HNKhN7luQhkjkNWIvy9_gizbEDUM4mSv_HWq9uuM,1068
+smallestai-1.3.1.dist-info/METADATA,sha256=anPPrjMmnWa_b1S65Wg23efwLUaoww7pbyU2qLhZTAk,9853
+smallestai-1.3.1.dist-info/WHEEL,sha256=PZUExdf71Ui_so67QXpySuHtCi3-J3wvF4ORK6k_S8U,91
+smallestai-1.3.1.dist-info/top_level.txt,sha256=i5ktbWkG-2aS28vrYTeuhKtA-tY2ZG7SHgLHi87QTLw,9
+smallestai-1.3.1.dist-info/RECORD,,

smallestai-1.3.0.dist-info/RECORD

@@ -1,12 +0,0 @@
-smallest/__init__.py,sha256=vaoIBml_IobavpVvFazB86iikg2iEy4h3ddxqv_0Fy4,190
-smallest/async_tts.py,sha256=zqZGuQUWaV2_if9WVdYutxb9G2UoUAxbyAbNlF1tv3U,7445
-smallest/exceptions.py,sha256=41GLVvNTfRQMQsPLGk0lHuhK2mak8_dVtiFLEtT23Dc,333
-smallest/models.py,sha256=R5UZZA9SibrJ2DsWPi_mkKI13WfyC-MLd-7kptfjns4,390
-smallest/stream_tts.py,sha256=4h_AktweZ386qgVIe8UeqO-ZxZO_x6Zj0uJQH09V1CE,5425
-smallest/tts.py,sha256=CHtZwcA2S4zfYfqhv5qikBKOME8XBjS_0R4HXpzXeAU,7325
-smallest/utils.py,sha256=WL71OByTxH8Y1gouP2K5YDDMwqhUdqMJ_bhqNryI3KQ,2222
-smallestai-1.3.0.dist-info/LICENSE,sha256=kK3HNKhN7luQhkjkNWIvy9_gizbEDUM4mSv_HWq9uuM,1068
-smallestai-1.3.0.dist-info/METADATA,sha256=0lqX-j9c0CkSeA6OeG5RUIdnaeWXWnvWPznCkO7vJCA,9845
-smallestai-1.3.0.dist-info/WHEEL,sha256=PZUExdf71Ui_so67QXpySuHtCi3-J3wvF4ORK6k_S8U,91
-smallestai-1.3.0.dist-info/top_level.txt,sha256=i5ktbWkG-2aS28vrYTeuhKtA-tY2ZG7SHgLHi87QTLw,9
-smallestai-1.3.0.dist-info/RECORD,,