util/Embeddings.py

import os
from copy import copy
from typing import Dict, List, Optional, Tuple, Union
import numpy as np
import time
from functools import wraps

os.environ['CURL_CA_BUNDLE'] = ''
from dotenv import load_dotenv, find_dotenv
_ = load_dotenv(find_dotenv())


class BaseEmbeddings:
    """
    Base class for embeddings
    """
    def __init__(self, path: str, is_api: bool) -> None:
        self.path = path
        self.is_api = is_api

    def get_embedding(self, text: str, model: str) -> List[float]:
        raise NotImplementedError

    @classmethod
    def cosine_similarity(cls, vector1: List[float], vector2: List[float]) -> float:
        """
        calculate cosine similarity between two vectors
        """
        dot_product = np.dot(vector1, vector2)
        magnitude = np.linalg.norm(vector1) * np.linalg.norm(vector2)
        if not magnitude:
            return 0
        return dot_product / magnitude


class OpenAIEmbedding(BaseEmbeddings):
    """
    class for OpenAI embeddings
    """
    def __init__(self, path: str = '', is_api: bool = True) -> None:
        super().__init__(path, is_api)
        if self.is_api:
            from openai import OpenAI
            self.client = OpenAI()
            self.client.api_key = os.getenv("OPENAI_API_KEY")
            self.client.base_url = os.getenv("OPENAI_BASE_URL")

    def get_embedding(self, text: str, model: str = "text-embedding-3-large") -> List[float]:
        if self.is_api:
            text = text.replace("\n", " ")
            return self.client.embeddings.create(input=[text], model=model).data[0].embedding
        else:
            raise NotImplementedError

class JinaEmbedding(BaseEmbeddings):
    """
    class for Jina embeddings
    """
    def __init__(self, path: str = 'jinaai/jina-embeddings-v2-base-zh', is_api: bool = False) -> None:
        super().__init__(path, is_api)
        self._model = self.load_model()

    def get_embedding(self, text: str) -> List[float]:
        return self._model.encode([text])[0].tolist()

    def load_model(self):
        import torch
        from transformers import AutoModel
        if torch.cuda.is_available():
            device = torch.device("cuda")
        else:
            device = torch.device("cpu")
        model = AutoModel.from_pretrained(self.path, trust_remote_code=True).to(device)
        return model

class ZhipuEmbedding(BaseEmbeddings):
    """
    class for Zhipu embeddings
    """ 
    def __init__(self, path: str = '', is_api: bool = True) -> None:
        super().__init__(path, is_api)
        if self.is_api:
            from zhipuai import ZhipuAI
            self.client = ZhipuAI(api_key=os.getenv("ZHIPUAI_API_KEY"))

    def get_embedding(self, text: str) -> List[float]:
        response = self.client.embeddings.create(
        model="embedding-2",
        input=text,
        )
        return response.data[0].embedding
    

class DashscopeEmbedding(BaseEmbeddings):
    """
    class for Dashscope embeddings
    """
    def __init__(self, path: str = '', is_api: bool = True) -> None:
        super().__init__(path, is_api)
        if self.is_api:
            import dashscope
            dashscope.api_key = os.getenv("DASHSCOPE_API_KEY")
            self.client = dashscope.TextEmbedding

    def get_embedding(self, text: str, model: str='text-embedding-v1') -> List[float]:
        response = self.client.call(
            model=model,
            input=text
        )
        return response.output['embeddings'][0]['embedding']


class BgeEmbedding(BaseEmbeddings):
    """
    class for BGE embeddings
    """

    def __init__(self, path: str = 'BAAI/bge-en-icl', is_api: bool = False) -> None:
        super().__init__(path, is_api)
        self._model, self._tokenizer = self.load_model(path)

    def get_embedding(self, text: str) -> List[float]:
        import torch
        encoded_input = self._tokenizer([text], padding=True, truncation=True, return_tensors='pt')
        encoded_input = {k: v.to(self._model.device) for k, v in encoded_input.items()}
        with torch.no_grad():
            model_output = self._model(**encoded_input)
            sentence_embeddings = model_output[0][:, 0]
        sentence_embeddings = torch.nn.functional.normalize(sentence_embeddings, p=2, dim=1)
        return sentence_embeddings[0].tolist()

    def load_model(self, path: str):
        import torch
        from transformers import AutoModel, AutoTokenizer
        if torch.cuda.is_available():
            device = torch.device("cuda")
        else:
            device = torch.device("cpu")
        tokenizer = AutoTokenizer.from_pretrained(path)
        model = AutoModel.from_pretrained(path).to(device)
        model.eval()
        return model, tokenizer


def rate_limiter():
    def rate_limiter_decorator(func):
        @wraps(func)
        def wrapper(self, *args, **kwargs):
            max_calls_per_minute = self.max_qpm 
            interval = 60 / max_calls_per_minute
            current_time = time.time()
            
            # Check if there's a record of the last call, if not set it to 0
            if not hasattr(self, '_last_called'):
                self._last_called = 0
            elapsed_time = current_time - self._last_called
            if elapsed_time < interval:
                time_to_wait = interval - elapsed_time
                if self.silent is False:
                    print(f"## Rate limit reached. Waiting for {time_to_wait:.2f} seconds.")
                time.sleep(time_to_wait)
            result = func(self, *args, **kwargs)
            self._last_called = time.time()  
            return result
        return wrapper

    return rate_limiter_decorator


class TextEmb3LargeEmbedding(BaseEmbeddings):
    """
    class for text-embedding-3-large embeddings
    """ 
    def __init__(self, max_qpm, is_silent=False):
        from langchain_openai import AzureOpenAIEmbeddings

        ## https://gpt.bytedance.net/gpt_openapi/
        base_url = "https://search-va.byteintl.net/gpt/openapi/online/v2/crawl"
        api_version = "2024-03-01-preview"
        ak = "5dXdIKxZc8JWVVgvX0DN92HWIYb9NfEb_GPT_AK"
        model_name = "text-embedding-3-large"
        api_type = "azure"       
        self.llm = AzureOpenAIEmbeddings(
            azure_endpoint=base_url,
            openai_api_version=api_version,
            deployment=model_name,
            openai_api_key=ak,
            openai_api_type=api_type,
        )
        self.max_qpm = max_qpm
        self.silent = is_silent

    @rate_limiter()
    def get_embedding(self, text: str):
        return self.llm.embed_query(text)