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