smallestai 2.1.0__tar.gz → 2.2.0__tar.gz

{smallestai-2.1.0 → smallestai-2.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: smallestai
-Version: 2.1.0
+Version: 2.2.0
 Summary: Official Python client for the Smallest AI API
 Author-email: Smallest <support@smallest.ai>
 License: MIT
@@ -15,7 +15,6 @@ License-File: LICENSE
 Requires-Dist: aiohttp
 Requires-Dist: aiofiles
 Requires-Dist: requests
-Requires-Dist: sacremoses
 Requires-Dist: pydub
 Provides-Extra: test
 Requires-Dist: jiwer; extra == "test"

{smallestai-2.1.0 → smallestai-2.2.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "smallestai"
-version = "2.1.0"
+version = "2.2.0"
 description = "Official Python client for the Smallest AI API"
 authors = [
     {name = "Smallest", email = "support@smallest.ai"},
@@ -18,7 +18,6 @@ dependencies = [
     "aiohttp",
     "aiofiles",
     "requests",
-    "sacremoses",
     "pydub"
 ]
 

{smallestai-2.1.0 → smallestai-2.2.0}/smallest/async_tts.py

@@ -4,7 +4,7 @@ import json
 import aiohttp
 import aiofiles
 import requests
-from typing import Optional, Union, List
+from typing import Optional, Union, List, AsyncIterator
 
 from smallest.exceptions import TTSError, APIError
 from smallest.utils import (TTSOptions, validate_input, preprocess_text, add_wav_header, chunk_text,
@@ -19,6 +19,9 @@ class AsyncSmallest:
         sample_rate: Optional[int] = 24000,
         voice_id: Optional[str] = "emily",
         speed: Optional[float] = 1.0,
+        consistency: Optional[float] = 0.5,
+        similarity: Optional[float] = 0.0,
+        enhancement: Optional[int] = 1,
         add_wav_header: Optional[bool] = True
     ) -> None:
         """
@@ -34,6 +37,9 @@ class AsyncSmallest:
         - sample_rate (int): The sample rate for the audio output.
         - voice_id (TTSVoices): The voice to be used for synthesis.
         - speed (float): The speed of the speech synthesis.
+        - consistency (float): This parameter controls word repetition and skipping. Decrease it to prevent skipped words, and increase it to prevent repetition. Only supported in `lightning-large` model. Range - [0, 1]
+        - similarity (float): This parameter controls the similarity between the synthesized audio and the reference audio. Increase it to make the speech more similar to the reference audio. Only supported in `lightning-large` model. Range - [0, 1]
+        - enhancement (int): Enhances speech quality at the cost of increased latency. Only supported in `lightning-large` model. Range - [0, 2].
         - add_wav_header (bool): Whether to add a WAV header to the output audio.
 
         Methods:
@@ -45,7 +51,7 @@ class AsyncSmallest:
         self.api_key = api_key or os.environ.get("SMALLEST_API_KEY")
         if not self.api_key:
             raise TTSError()
-        if model == "lightning-large":
+        if model == "lightning-large" and voice_id is None:
             voice_id = "lakshya"
 
         self.chunk_size = 250
@@ -56,7 +62,10 @@ class AsyncSmallest:
             voice_id=voice_id,
             api_key=self.api_key,
             add_wav_header=add_wav_header,
-            speed=speed
+            speed=speed,
+            consistency=consistency,
+            similarity=similarity,
+            enhancement=enhancement
         )
         self.session = None
 
@@ -121,27 +130,25 @@ class AsyncSmallest:
     async def synthesize(
             self,
             text: str,
-            consistency: Optional[float] = 0.5,
-            similarity: Optional[float] = 0,
-            enhancement: Optional[bool] = False,
+            stream: Optional[bool] = False,
             save_as: Optional[str] = None,
             **kwargs
-        ) -> Union[bytes, None]:
+        ) -> Union[bytes, None, AsyncIterator[bytes]]:
         """
         Asynchronously synthesize speech from the provided text.
 
         Args:
         - text (str): The text to be converted to speech.
+        - stream (Optional[bool]): If True, returns an iterator yielding audio chunks instead of a full byte array.
         - save_as (Optional[str]): If provided, the synthesized audio will be saved to this file path.
                                    The file must have a .wav extension.
-        - consistency (Optional[float]): This parameter controls word repetition and skipping. Decrease it to prevent skipped words, and increase it to prevent repetition. Only supported in `lightning-large` model.
-        - similarity (Optional[float]): This parameter controls the similarity between the synthesized audio and the reference audio. Increase it to make the speech more similar to the reference audio. Only supported in `lightning-large` model.
-        - enhancement (Optional[bool]): Enhances speech quality at the cost of increased latency. Only supported in `lightning-large` model.
         - kwargs: Additional optional parameters to override `__init__` options for this call.
 
         Returns:
-        - Union[bytes, None]: The synthesized audio content in bytes if `save_as` is not specified;
-                              otherwise, returns None after saving the audio to the specified file.
+        - Union[bytes, None, Iterator[bytes]]: 
+            - If `stream=True`, returns an iterator yielding audio chunks.
+            - If `save_as` is provided, saves the file and returns None.
+            - Otherwise, returns the synthesized audio content as bytes.
 
         Raises:
         - TTSError: If the provided file name does not have a .wav extension when `save_as` is specified.
@@ -165,44 +172,50 @@ class AsyncSmallest:
             for key, value in kwargs.items():
                 setattr(opts, key, value)
 
-            validate_input(preprocess_text(text), opts.model, opts.sample_rate, opts.speed, consistency, similarity, enhancement)
+            text = preprocess_text(text)
+            validate_input(text, opts.model, opts.sample_rate, opts.speed, opts.consistency, opts.similarity, opts.enhancement)
 
             self.chunk_size = 250
             if opts.model == 'lightning-large':
                 self.chunk_size = 140
 
             chunks = chunk_text(text, self.chunk_size)
-            audio_content = b""
-
-            for chunk in chunks:
-                payload = {
-                    "text": preprocess_text(chunk),
-                    "sample_rate": opts.sample_rate,
-                    "voice_id": opts.voice_id,
-                    "add_wav_header": False,
-                    "speed": opts.speed,
-                    "model": opts.model
-                }
-                
-                if opts.model == "lightning-large":
-                    if consistency:
-                        payload["consistency"] = consistency
-                    if similarity:
-                        payload["similarity"] = similarity
-                    if enhancement:
-                        payload["enhancement"] = enhancement
-
-
-                headers = {
-                    "Authorization": f"Bearer {self.api_key}",
-                    "Content-Type": "application/json",
-                }
-
-                async with self.session.post(f"{API_BASE_URL}/{opts.model}/get_speech", json=payload, headers=headers) as res:
-                    if res.status != 200:
-                        raise APIError(f"Failed to synthesize speech: {await res.text()}. For more information, visit https://waves.smallest.ai/")
-
-                    audio_content += await res.read()
+
+            async def audio_stream():
+                for chunk in chunks:
+                    payload = {
+                        "text": chunk,
+                        "sample_rate": opts.sample_rate,
+                        "voice_id": opts.voice_id,
+                        "add_wav_header": False,
+                        "speed": opts.speed,
+                        "model": opts.model
+                    }
+                    
+                    if opts.model == "lightning-large":
+                        if opts.consistency is not None:
+                            payload["consistency"] = opts.consistency
+                        if opts.similarity is not None:
+                            payload["similarity"] = opts.similarity
+                        if opts.enhancement is not None:
+                            payload["enhancement"] = opts.enhancement
+
+
+                    headers = {
+                        "Authorization": f"Bearer {self.api_key}",
+                        "Content-Type": "application/json",
+                    }
+
+                    async with self.session.post(f"{API_BASE_URL}/{opts.model}/get_speech", json=payload, headers=headers) as res:
+                        if res.status != 200:
+                            raise APIError(f"Failed to synthesize speech: {await res.text()}. For more information, visit https://waves.smallest.ai/")
+
+                        yield await res.read()
+            
+            if stream:
+                return audio_stream()
+
+            audio_content = b"".join([chunk async for chunk in audio_stream()])
 
             if save_as:
                 if not save_as.endswith(".wav"):

{smallestai-2.1.0 → smallestai-2.2.0}/smallest/stream_tts.py

@@ -1,7 +1,8 @@
 import asyncio
+import time
 from threading import Thread
 from queue import Queue, Empty
-from typing import AsyncGenerator, Optional, Union
+from typing import AsyncGenerator, Optional, Union, List, Dict, Any
 
 from smallest.tts import Smallest
 from smallest.exceptions import APIError
@@ -13,7 +14,8 @@ class TextToAudioStream:
         self,
         tts_instance: Union[Smallest, AsyncSmallest],
         queue_timeout: Optional[float] = 5.0,
-        max_retries: Optional[int] = 3
+        max_retries: Optional[int] = 3,
+        verbose: bool = False
     ):
         """
         A real-time text-to-speech processor that converts streaming text into audio output.
@@ -30,8 +32,9 @@ class TextToAudioStream:
 
         Args:
             tts_instance: The text-to-speech engine to use (Smallest or AsyncSmallest)
-            queue_timeout: How long to wait for new text (seconds, default: 5.0)
+            queue_timeout: How long to wait for new text (seconds, default: 1.0)
             max_retries: Number of retry attempts for failed synthesis (default: 3)
+            verbose: Whether to log detailed metrics about TTS requests (default: False)
         """
         self.tts_instance = tts_instance
         self.tts_instance.opts.add_wav_header = False
@@ -41,6 +44,14 @@ class TextToAudioStream:
         self.queue = Queue()
         self.buffer_size = 250
         self.stop_flag = False
+        self.verbose = verbose
+        
+        # Metrics tracking
+        self.request_count = 0
+        self.request_logs: List[Dict[str, Any]] = []
+        self.start_time = 0
+        self.first_api_response_time = None
+        self.end_time = 0
 
         if self.tts_instance.opts.model == 'lightning-large':
             self.buffer_size = 140
@@ -90,24 +101,76 @@ class TextToAudioStream:
 
     def _synthesize_sync(self, sentence: str, retries: int = 0) -> Optional[bytes]:
         """Synchronously synthesizes a given sentence."""
+        request_start_time = time.time()
+        request_id = self.request_count + 1
+        
         try:
-            return self.tts_instance.synthesize(sentence)
+            audio_content = self.tts_instance.synthesize(sentence)
+            self.request_count += 1
+            request_end_time = time.time()
+            
+            if self.verbose:
+                request_duration = request_end_time - request_start_time
+                if self.first_api_response_time is None:
+                    self.first_api_response_time = time.time() - self.start_time
+                
+                self.request_logs.append({
+                    "id": request_id,
+                    "text": sentence,
+                    "start_time": request_start_time - self.start_time,
+                    "end_time": request_end_time - self.start_time,
+                    "duration": request_duration,
+                    "char_count": len(sentence),
+                    "retries": retries
+                })
+                
+            return audio_content
         except APIError as e:
             if retries < self.max_retries:
+                if self.verbose:
+                    print(f"Retry {retries + 1}/{self.max_retries} for request: '{sentence[:30]}...'")
                 return self._synthesize_sync(sentence, retries + 1)
             else:
-                raise APIError(f"Error: {e}. Retries Exhausted, for more information, visit https://waves.smallest.ai/")
+                if self.verbose:
+                    print(f"Synthesis failed for sentence: {sentence} - Error: {e}. Retries Exhausted, for more information, visit https://waves.smallest.ai/")
+                return None
             
 
     async def _synthesize_async(self, sentence: str, retries: int = 0) -> Optional[bytes]:
         """Asynchronously synthesizes a given sentence."""
+        request_start_time = time.time()
+        request_id = self.request_count + 1
+        
         try:
-            return await self.tts_instance.synthesize(sentence)
+            audio_content = await self.tts_instance.synthesize(sentence)
+            self.request_count += 1
+            request_end_time = time.time()
+            
+            if self.verbose:
+                request_duration = request_end_time - request_start_time
+                if self.first_api_response_time is None:
+                    self.first_api_response_time = time.time() - self.start_time
+                
+                self.request_logs.append({
+                    "id": request_id,
+                    "text": sentence,
+                    "start_time": request_start_time - self.start_time,
+                    "end_time": request_end_time - self.start_time,
+                    "duration": request_duration,
+                    "char_count": len(sentence),
+                    "retries": retries
+                })
+                
+            return audio_content
         except APIError as e:
             if retries < self.max_retries:
+                if self.verbose:
+                    print(f"Retry {retries + 1}/{self.max_retries} for request: '{sentence[:30]}...'")
                 return await self._synthesize_async(sentence, retries + 1)
             else:
-                raise APIError(f"Error: {e}. Retries Exhausted, for more information, visit https://waves.smallest.ai/")
+                if self.verbose:
+                    print(f"Synthesis failed for sentence: {sentence} - Error: {e}. Retries Exhausted, for more information, visit https://waves.smallest.ai/")
+                return None
 
 
     async def _run_synthesis(self) -> AsyncGenerator[bytes, None]:
@@ -117,7 +180,8 @@ class TextToAudioStream:
         """
         while not self.stop_flag or not self.queue.empty():
             try:
-                sentence = self.queue.get(timeout=self.queue_timeout)
+                sentence = self.queue.get_nowait()
+                
                 if isinstance(self.tts_instance, AsyncSmallest):
                     audio_content = await self._synthesize_async(sentence)
                 else:
@@ -126,10 +190,55 @@ class TextToAudioStream:
                 
                 if audio_content:
                     yield audio_content
+                    
             except Empty:
-                if self.stop_flag:
+                # Quick check if we should exit
+                if self.stop_flag and self.queue.empty():
                     break
-                await asyncio.sleep(0.1)  # avoid busy waiting if the queue is empty
+                    
+                # Short sleep to avoid busy-waiting
+                await asyncio.sleep(0.01)  # Much shorter sleep time (10ms)
+
+
+    def _print_verbose_summary(self) -> None:
+        """Print a summary of all metrics if verbose mode is enabled."""
+        if not self.verbose:
+            return
+            
+        total_duration = self.end_time - self.start_time
+        
+        print("\n" + "="*100)
+        print(f"TEXT-TO-AUDIO STREAM METRICS")
+        print("="*100)
+        
+        print(f"\nOVERALL STATISTICS:")
+        print(f"  Total requests made: {self.request_count}")
+        print(f"  Time to first API response: {self.first_api_response_time:.3f}s")
+        print(f"  Total processing time: {total_duration:.3f}s")
+        
+        # Print table header
+        print("\nREQUEST DETAILS:")
+        header = f"{'#':4} {'Start (s)':10} {'End (s)':10} {'Duration (s)':12} {'Characters':15} {'Text'}"
+        print("\n" + header)
+        print("-" * 100)
+        
+        # Print table rows
+        for log in self.request_logs:
+            row = (
+                f"{log['id']:4} "
+                f"{log['start_time']:10.3f} "
+                f"{log['end_time']:10.3f} "
+                f"{log['duration']:12.3f} "
+                f"{log['char_count']:15} "
+                f"{log['text'][:50]}{'...' if len(log['text']) > 50 else ''}"
+            )
+            print(row)
+            
+            # Print retry information if any
+            if log['retries'] > 0:
+                print(f"{'':4} {'':10} {'':10} {'':12} {'':15} Retries: {log['retries']}")
+                
+        print("\n" + "="*100)
 
 
     async def process(self, llm_output: AsyncGenerator[str, None]) -> AsyncGenerator[bytes, None]:
@@ -149,13 +258,15 @@ class TextToAudioStream:
             - Streamed over a network
             - Further processed as needed
         """
-        stream_task = asyncio.create_task(self._stream_llm_output(llm_output))
+        self.start_time = time.time()
+        
+        llm_thread = Thread(target=asyncio.run, args=(self._stream_llm_output(llm_output),))
+        llm_thread.start()
 
-        try:
-            async for audio_content in self._run_synthesis():
-                yield audio_content
-        except Exception as e:
-            raise APIError(f"Error during synthesis processing: {e}")
+        async for audio_content in self._run_synthesis():
+            yield audio_content
 
-        finally:
-            await stream_task
+        llm_thread.join()
+        
+        self.end_time = time.time()
+        self._print_verbose_summary()

{smallestai-2.1.0 → smallestai-2.2.0}/smallest/tts.py

@@ -3,7 +3,7 @@ import json
 import wave
 import copy
 import requests
-from typing import Optional, Union, List
+from typing import Optional, Union, List, Iterator
 
 from smallest.exceptions import TTSError, APIError
 from smallest.utils import (TTSOptions, validate_input, preprocess_text, add_wav_header, chunk_text,
@@ -17,6 +17,9 @@ class Smallest:
         sample_rate: Optional[int] = 24000,
         voice_id: Optional[str] = "emily",
         speed: Optional[float] = 1.0,
+        consistency: Optional[float] = 0.5,
+        similarity: Optional[float] = 0.0,
+        enhancement: Optional[int] = 1,
         add_wav_header: Optional[bool] = True
     ) -> None:
         """
@@ -31,6 +34,9 @@ class Smallest:
         - sample_rate (int): The sample rate for the audio output.
         - voice_id (TTSVoices): The voice to be used for synthesis.
         - speed (float): The speed of the speech synthesis.
+        - consistency (float): This parameter controls word repetition and skipping. Decrease it to prevent skipped words, and increase it to prevent repetition. Only supported in `lightning-large` model. Range - [0, 1]
+        - similarity (float): This parameter controls the similarity between the synthesized audio and the reference audio. Increase it to make the speech more similar to the reference audio. Only supported in `lightning-large` model. Range - [0, 1]
+        - enhancement (int): Enhances speech quality at the cost of increased latency. Only supported in `lightning-large` model. Range - [0, 2].
         - add_wav_header (bool): Whether to add a WAV header to the output audio.
 
         Methods:
@@ -42,7 +48,7 @@ class Smallest:
         self.api_key = api_key or os.environ.get("SMALLEST_API_KEY")
         if not self.api_key:
             raise TTSError()
-        if model == "lightning-large":
+        if model == "lightning-large" and voice_id is None:
             voice_id = "lakshya"
 
         self.chunk_size = 250
@@ -53,7 +59,10 @@ class Smallest:
             voice_id=voice_id,
             api_key=self.api_key,
             add_wav_header=add_wav_header,
-            speed=speed
+            speed=speed,
+            consistency=consistency,
+            similarity=similarity,
+            enhancement=enhancement
         )
     
         
@@ -98,27 +107,24 @@ class Smallest:
     def synthesize(
             self,
             text: str,
-            consistency: Optional[float] = 0.5,
-            similarity: Optional[float] = 0,
-            enhancement: Optional[bool] = False,
+            stream: Optional[bool] = False,
             save_as: Optional[str] = None,
             **kwargs
-        ) -> Union[bytes, None]:
+        ) -> Union[bytes, None, Iterator[bytes]]:
         """
         Synthesize speech from the provided text.
 
-        Args:
         - text (str): The text to be converted to speech.
-        - save_as (Optional[str]): If provided, the synthesized audio will be saved to this file path. 
+        - stream (Optional[bool]): If True, returns an iterator yielding audio chunks instead of a full byte array.
+        - save_as (Optional[str]): If provided, the synthesized audio will be saved to this file path.
                                    The file must have a .wav extension.
-        - consistency (Optional[float]): This parameter controls word repetition and skipping. Decrease it to prevent skipped words, and increase it to prevent repetition. Only supported in `lightning-large` model.
-        - similarity (Optional[float]): This parameter controls the similarity between the synthesized audio and the reference audio. Increase it to make the speech more similar to the reference audio. Only supported in `lightning-large` model.
-        - enhancement (Optional[bool]): Enhances speech quality at the cost of increased latency. Only supported in `lightning-large` model.
         - kwargs: Additional optional parameters to override `__init__` options for this call.
 
         Returns:
-        - Union[bytes, None]: The synthesized audio content in bytes if `save_as` is not specified; 
-                              otherwise, returns None after saving the audio to the specified file.
+        - Union[bytes, None, Iterator[bytes]]: 
+            - If `stream=True`, returns an iterator yielding audio chunks.
+            - If `save_as` is provided, saves the file and returns None.
+            - Otherwise, returns the synthesized audio content as bytes.
 
         Raises:
         - TTSError: If the provided file name does not have a .wav extension when `save_as` is specified.
@@ -134,42 +140,48 @@ class Smallest:
         for key, value in kwargs.items():
             setattr(opts, key, value)
 
-        validate_input(preprocess_text(text), opts.model, opts.sample_rate, opts.speed, consistency, similarity, enhancement)
+        text = preprocess_text(text)
+        validate_input(text, opts.model, opts.sample_rate, opts.speed, opts.consistency, opts.similarity, opts.enhancement)
 
         self.chunk_size = 250
         if opts.model == "lightning-large":
             self.chunk_size = 140
 
         chunks = chunk_text(text, self.chunk_size)
-        audio_content = b""
-
-        for chunk in chunks:
-            payload = {
-                "text": preprocess_text(chunk),
-                "sample_rate": opts.sample_rate,
-                "voice_id": opts.voice_id,
-                "add_wav_header": False,
-                "speed": opts.speed,
-            }
-
-            if opts.model == "lightning-large":
-                if consistency:
-                    payload["consistency"] = consistency
-                if similarity:
-                    payload["similarity"] = similarity
-                if enhancement:
-                    payload["enhancement"] = enhancement
-
-            headers = {
-                "Authorization": f"Bearer {self.api_key}",
-                "Content-Type": "application/json",
-            }
-
-            res = requests.post(f"{API_BASE_URL}/{opts.model}/get_speech", json=payload, headers=headers)
-            if res.status_code != 200:
-                raise APIError(f"Failed to synthesize speech: {res.text}. For more information, visit https://waves.smallest.ai/")
+
+        def audio_stream():
+            for chunk in chunks:
+                payload = {
+                    "text": chunk,
+                    "sample_rate": opts.sample_rate,
+                    "voice_id": opts.voice_id,
+                    "add_wav_header": False,
+                    "speed": opts.speed,
+                }
+
+                if opts.model == "lightning-large":
+                    if opts.consistency is not None:
+                        payload["consistency"] = opts.consistency
+                    if opts.similarity is not None:
+                        payload["similarity"] = opts.similarity
+                    if opts.enhancement is not None:
+                        payload["enhancement"] = opts.enhancement
+
+                headers = {
+                    "Authorization": f"Bearer {self.api_key}",
+                    "Content-Type": "application/json",
+                }
+
+                res = requests.post(f"{API_BASE_URL}/{opts.model}/get_speech", json=payload, headers=headers)
+                if res.status_code != 200:
+                    raise APIError(f"Failed to synthesize speech: {res.text}. Please check if you have set the correct API key. For more information, visit https://waves.smallest.ai/")
+
+                yield res.content
             
-            audio_content += res.content
+        if stream:
+            return audio_stream()
+        
+        audio_content = b"".join(audio_stream())
 
         if save_as:
             if not save_as.endswith(".wav"):

{smallestai-2.1.0 → smallestai-2.2.0}/smallest/utils.py

@@ -4,7 +4,6 @@ from typing import List
 from typing import Optional
 from pydub import AudioSegment
 from dataclasses import dataclass
-from sacremoses import MosesPunctNormalizer
 
 from smallest.exceptions import ValidationError
 from smallest.models import TTSModels, TTSLanguages
@@ -12,7 +11,6 @@ from smallest.models import TTSModels, TTSLanguages
 
 API_BASE_URL = "https://waves-api.smallest.ai/api/v1"
 SENTENCE_END_REGEX = re.compile(r'.*[-.—!?,;:…।|]$')
-mpn = MosesPunctNormalizer()
 SAMPLE_WIDTH = 2
 CHANNELS = 1
 ALLOWED_AUDIO_EXTENSIONS = ['.mp3', '.wav']
@@ -26,9 +24,12 @@ class TTSOptions:
     api_key: str
     add_wav_header: bool
     speed: float
+    consistency: float
+    similarity: float
+    enhancement: int
 
 
-def validate_input(text: str, model: str, sample_rate: int, speed: float, consistency: Optional[float] = None, similarity: Optional[float] = None, enhancement: Optional[bool] = None):
+def validate_input(text: str, model: str, sample_rate: int, speed: float, consistency: Optional[float] = None, similarity: Optional[float] = None, enhancement: Optional[int] = None):
     if not text:
         raise ValidationError("Text cannot be empty.")
     if model not in TTSModels:
@@ -41,8 +42,8 @@ def validate_input(text: str, model: str, sample_rate: int, speed: float, consis
         raise ValidationError(f"Invalid consistency: {consistency}. Must be between 0.0 and 1.0")
     if similarity is not None and not 0.0 <= similarity <= 1.0:
         raise ValidationError(f"Invalid similarity: {similarity}. Must be between 0.0 and 1.0")
-    if enhancement is not None and not isinstance(enhancement, bool):
-        raise ValidationError(f"Invalid enhancement: {enhancement}. Must be a boolean value.")
+    if enhancement is not None and not 0 <= enhancement <= 2:
+        raise ValidationError(f"Invalid enhancement: {enhancement}. Must be between 0 and 2.")
 
 
 def add_wav_header(frame_input: bytes, sample_rate: int = 24000, sample_width: int = 2, channels: int = 1) -> bytes:
@@ -56,7 +57,6 @@ def add_wav_header(frame_input: bytes, sample_rate: int = 24000, sample_width: i
 def preprocess_text(text: str) -> str:
     text = text.replace("\n", " ").replace("\t", " ").replace("—", " ").replace("-", " ").replace("–", " ")
     text = re.sub(r'\s+', ' ', text)
-    text = mpn.normalize(text)
     return text.strip()
 
 

{smallestai-2.1.0 → smallestai-2.2.0}/smallestai.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: smallestai
-Version: 2.1.0
+Version: 2.2.0
 Summary: Official Python client for the Smallest AI API
 Author-email: Smallest <support@smallest.ai>
 License: MIT
@@ -15,7 +15,6 @@ License-File: LICENSE
 Requires-Dist: aiohttp
 Requires-Dist: aiofiles
 Requires-Dist: requests
-Requires-Dist: sacremoses
 Requires-Dist: pydub
 Provides-Extra: test
 Requires-Dist: jiwer; extra == "test"

{smallestai-2.1.0 → smallestai-2.2.0}/smallestai.egg-info/requires.txt

@@ -1,7 +1,6 @@
 aiohttp
 aiofiles
 requests
-sacremoses
 pydub
 
 [test]