from typing import Dict, Type
from .providers.mistralstt import MistralSTT
from .providers.groqstt import GroqSTT
from .providers.deepgramstt import DeepgramSTT
from .providers.aaistt import AssemblyAISTT
from .providers.customstt import CustomSTT

class NoSpeechDetected(Exception):
    pass
class ProviderUnavailable(Exception):
    pass


class STTFactory:
    _providers: Dict[str, Type] = {
        "mistral": MistralSTT,
        "groq": GroqSTT,
        "deepgram": DeepgramSTT, 
        "assemblyai": AssemblyAISTT,
        "custom": CustomSTT,
    }

    @classmethod
    def get_provider(cls, provider_name: str):
        provider_class = cls._providers.get(provider_name.lower())
        if not provider_class:
            raise ValueError(f"STT provider '{provider_name}' not found")
        return provider_class()

class STTRemoteManager:
    def __init__(self, default_provider: str = "mistral"):
        self.default_provider = default_provider

    async def transcribe_remote(self, audio_path: str, provider_name: str = None):
        provider_name = (provider_name or self.default_provider).lower()
        provider = STTFactory.get_provider(provider_name)

        try:
            text, language = await provider.transcribe(audio_path)
        except Exception as e:
            # Real provider error — not silence
            raise ProviderUnavailable(
                f"Provider '{provider_name}' failed: {e}"
            ) from e

        if not text or not text.strip():
            raise NoSpeechDetected(
                f"Provider '{provider_name}' returned no speech for this chunk."
            )

        return text, language