app/services/async_embeddings_service.py

"""
AsyncEmbeddingsService:
- Wraps the synchronous EmbeddingsService to provide async interface
- Uses asyncio.to_thread to offload blocking model operations to a thread pool
- Manages concurrency with asyncio.Lock
"""

import asyncio
from typing import List, Optional
from .embeddings_service import EmbeddingsService

class AsyncEmbeddingsService:
    def __init__(self, model_name: Optional[str] = None):
        # We compose the synchronous service
        self._sync_service = EmbeddingsService(model_name)
        self._lock = asyncio.Lock()

    @property
    def is_ready(self) -> bool:
        return self._sync_service.is_ready

    @property
    def device(self):
        return self._sync_service.device

    async def load_model(self) -> None:
        """Async wrapper for loading the model."""
        if self.is_ready:
            return
            
        async with self._lock:
            if self.is_ready:
                return
            # Run the heavy lifting in a thread
            await asyncio.to_thread(self._sync_service.load_model)

    async def generate_single_embedding(self, text: str) -> List[float]:
        """
        Generate a single embedding asynchronously.
        """
        # Ensure model is loaded first (non-blocking check mostly)
        await self.load_model()
        return await asyncio.to_thread(self._sync_service.generate_single_embedding, text)

    async def generate_batch_embeddings(self, texts: List[str]) -> List[List[float]]:
        """
        Efficient batch processing asynchronously.
        """
        await self.load_model()
        return await asyncio.to_thread(self._sync_service.generate_batch_embeddings, texts)