from typing import List
import numpy as np
from loguru import logger
import httpx
from tenacity import retry, stop_after_attempt, wait_exponential

from .utils import timing_decorator

class EmbeddingClient:
    def __init__(self):
        """
        Khởi tạo EmbeddingClient.
        Input: None
        Output: EmbeddingClient instance.
        """
        self._client = httpx.AsyncClient()

    @timing_decorator
    @retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=4, max=10))
    async def create_embedding(self, text: str) -> List[float]:
        """
        Tạo embedding vector từ text bằng dịch vụ embedding (ví dụ OpenAI).
        Input: text (str)
        Output: list[float] embedding vector.
        """
        try:
            # This is a placeholder for your actual embedding service
            # You should replace this with your preferred embedding service (e.g., OpenAI, Cohere, etc.)
            # For example, using OpenAI's embedding API:
            
            # response = await self._client.post(
            #     "https://api.openai.com/v1/embeddings",
            #     headers={"Authorization": f"Bearer {api_key}"},
            #     json={
            #         "input": text,
            #         "model": "text-embedding-ada-002"
            #     }
            # )
            # embedding = response.json()["data"][0]["embedding"]
            
            # For now, we'll return a random vector as a placeholder
            embedding = np.random.normal(0, 1, 1536).tolist()  # 1536 is OpenAI's embedding dimension
            return embedding
        except Exception as e:
            logger.error(f"Error creating embedding: {e}")
            raise

    def cosine_similarity(self, embedding1: List[float], embedding2: List[float]) -> float:
        """
        Tính cosine similarity giữa hai embedding.
        Input: embedding1 (list[float]), embedding2 (list[float])
        Output: float (giá trị similarity)
        """
        try:
            a = np.array(embedding1)
            b = np.array(embedding2)
            return float(np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b)))
        except Exception as e:
            logger.error(f"Error calculating similarity: {e}")
            return 0.0