Spaces:
Paused
Paused
Delete stt_interface.py
Browse files- stt_interface.py +0 -80
stt_interface.py
DELETED
|
@@ -1,80 +0,0 @@
|
|
| 1 |
-
"""
|
| 2 |
-
STT (Speech-to-Text) Interface for Flare
|
| 3 |
-
"""
|
| 4 |
-
from abc import ABC, abstractmethod
|
| 5 |
-
from typing import Optional, Dict, Any, AsyncIterator, List
|
| 6 |
-
from dataclasses import dataclass
|
| 7 |
-
from enum import Enum
|
| 8 |
-
import json
|
| 9 |
-
|
| 10 |
-
class STTEngineType(Enum):
|
| 11 |
-
NO_STT = "no_stt"
|
| 12 |
-
GOOGLE = "google"
|
| 13 |
-
AZURE = "azure"
|
| 14 |
-
AMAZON = "amazon"
|
| 15 |
-
FLICKER = "flicker"
|
| 16 |
-
|
| 17 |
-
@dataclass
|
| 18 |
-
class STTConfig:
|
| 19 |
-
"""STT configuration parameters"""
|
| 20 |
-
language: str = "tr-TR"
|
| 21 |
-
sample_rate: int = 16000
|
| 22 |
-
encoding: str = "WEBM_OPUS"
|
| 23 |
-
enable_punctuation: bool = True
|
| 24 |
-
enable_word_timestamps: bool = False
|
| 25 |
-
model: str = "latest_long"
|
| 26 |
-
use_enhanced: bool = True
|
| 27 |
-
single_utterance: bool = False
|
| 28 |
-
interim_results: bool = True
|
| 29 |
-
|
| 30 |
-
# Voice Activity Detection
|
| 31 |
-
vad_enabled: bool = True
|
| 32 |
-
speech_timeout_ms: int = 2000
|
| 33 |
-
|
| 34 |
-
# Noise reduction
|
| 35 |
-
noise_reduction_enabled: bool = True
|
| 36 |
-
noise_reduction_level: int = 2
|
| 37 |
-
|
| 38 |
-
@dataclass
|
| 39 |
-
class TranscriptionResult:
|
| 40 |
-
"""Result from STT engine"""
|
| 41 |
-
text: str
|
| 42 |
-
is_final: bool
|
| 43 |
-
confidence: float
|
| 44 |
-
timestamp: float
|
| 45 |
-
word_timestamps: Optional[List[Dict]] = None
|
| 46 |
-
language: Optional[str] = None
|
| 47 |
-
is_interrupt: bool = False
|
| 48 |
-
|
| 49 |
-
class STTInterface(ABC):
|
| 50 |
-
"""Abstract base class for STT providers"""
|
| 51 |
-
|
| 52 |
-
@abstractmethod
|
| 53 |
-
async def start_streaming(self, config: STTConfig) -> None:
|
| 54 |
-
"""Start streaming session"""
|
| 55 |
-
pass
|
| 56 |
-
|
| 57 |
-
@abstractmethod
|
| 58 |
-
async def stream_audio(self, audio_chunk: bytes) -> AsyncIterator[TranscriptionResult]:
|
| 59 |
-
"""Stream audio chunk and get transcription results"""
|
| 60 |
-
pass
|
| 61 |
-
|
| 62 |
-
@abstractmethod
|
| 63 |
-
async def stop_streaming(self) -> Optional[TranscriptionResult]:
|
| 64 |
-
"""Stop streaming and get final result"""
|
| 65 |
-
pass
|
| 66 |
-
|
| 67 |
-
@abstractmethod
|
| 68 |
-
def supports_realtime(self) -> bool:
|
| 69 |
-
"""Check if provider supports real-time streaming"""
|
| 70 |
-
pass
|
| 71 |
-
|
| 72 |
-
@abstractmethod
|
| 73 |
-
def get_supported_languages(self) -> List[str]:
|
| 74 |
-
"""Get list of supported language codes"""
|
| 75 |
-
pass
|
| 76 |
-
|
| 77 |
-
@abstractmethod
|
| 78 |
-
def get_provider_name(self) -> str:
|
| 79 |
-
"""Get provider name for logging"""
|
| 80 |
-
pass
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|