Source code for langchain_community.embeddings.dashscope

from __future__ import annotations

import logging
from typing import (
    Any,
    Callable,
    Dict,
    List,
    Optional,
)

from langchain_core.embeddings import Embeddings
from langchain_core.pydantic_v1 import BaseModel, Extra, root_validator
from langchain_core.utils import get_from_dict_or_env
from requests.exceptions import HTTPError
from tenacity import (
    before_sleep_log,
    retry,
    retry_if_exception_type,
    stop_after_attempt,
    wait_exponential,
)

logger = logging.getLogger(__name__)


def _create_retry_decorator(embeddings: DashScopeEmbeddings) -> Callable[[Any], Any]:
    multiplier = 1
    min_seconds = 1
    max_seconds = 4
    # Wait 2^x * 1 second between each retry starting with
    # 1 seconds, then up to 4 seconds, then 4 seconds afterwards
    return retry(
        reraise=True,
        stop=stop_after_attempt(embeddings.max_retries),
        wait=wait_exponential(multiplier, min=min_seconds, max=max_seconds),
        retry=(retry_if_exception_type(HTTPError)),
        before_sleep=before_sleep_log(logger, logging.WARNING),
    )


def embed_with_retry(embeddings: DashScopeEmbeddings, **kwargs: Any) -> Any:
    """使用tenacity来重试嵌入调用。"""
    retry_decorator = _create_retry_decorator(embeddings)

    @retry_decorator
    def _embed_with_retry(**kwargs: Any) -> Any:
        result = []
        i = 0
        input_data = kwargs["input"]
        while i < len(input_data):
            kwargs["input"] = input_data[i : i + 25]
            resp = embeddings.client.call(**kwargs)
            if resp.status_code == 200:
                result += resp.output["embeddings"]
            elif resp.status_code in [400, 401]:
                raise ValueError(
                    f"status_code: {resp.status_code} \n "
                    f"code: {resp.code} \n message: {resp.message}"
                )
            else:
                raise HTTPError(
                    f"HTTP error occurred: status_code: {resp.status_code} \n "
                    f"code: {resp.code} \n message: {resp.message}",
                    response=resp,
                )
            i += 25
        return result

    return _embed_with_retry(**kwargs)


class DashScopeEmbeddings(BaseModel, Embeddings):
    """DashScope嵌入模型。

    要使用，您应该已安装``dashscope`` python包，并且
    环境变量``DASHSCOPE_API_KEY``设置为您的API密钥，或将其传递
    作为构造函数的命名参数。

    示例:
        .. code-block:: python

            from langchain_community.embeddings import DashScopeEmbeddings
            embeddings = DashScopeEmbeddings(dashscope_api_key="my-api-key")

    示例:
        .. code-block:: python

            import os
            os.environ["DASHSCOPE_API_KEY"] = "your DashScope API KEY"

            from langchain_community.embeddings.dashscope import DashScopeEmbeddings
            embeddings = DashScopeEmbeddings(
                model="text-embedding-v1",
            )
            text = "This is a test query."
            query_result = embeddings.embed_query(text)"""

    client: Any  #: :meta private:
    """DashScope客户端。"""
    model: str = "text-embedding-v1"
    dashscope_api_key: Optional[str] = None
    max_retries: int = 5
    """生成时最大的重试次数。"""

    class Config:
        """此pydantic对象的配置。"""

        extra = Extra.forbid

    @root_validator()
    def validate_environment(cls, values: Dict) -> Dict:
        import dashscope

        """Validate that api key and python package exists in environment."""
        values["dashscope_api_key"] = get_from_dict_or_env(
            values, "dashscope_api_key", "DASHSCOPE_API_KEY"
        )
        dashscope.api_key = values["dashscope_api_key"]
        try:
            import dashscope

            values["client"] = dashscope.TextEmbedding
        except ImportError:
            raise ImportError(
                "Could not import dashscope python package. "
                "Please install it with `pip install dashscope`."
            )
        return values

    def embed_documents(self, texts: List[str]) -> List[List[float]]:
        """调用DashScope的嵌入端点以嵌入搜索文档。

参数：
    texts: 要嵌入的文本列表。
    chunk_size: 嵌入的块大小。如果为None，则使用类指定的块大小。

返回：
    每个文本的嵌入列表。
"""
        embeddings = embed_with_retry(
            self, input=texts, text_type="document", model=self.model
        )
        embedding_list = [item["embedding"] for item in embeddings]
        return embedding_list

    def embed_query(self, text: str) -> List[float]:
        """调用DashScope的嵌入端点以嵌入查询文本。

参数：
    text：要嵌入的文本。

返回：
    文本的嵌入。
"""
        embedding = embed_with_retry(
            self, input=text, text_type="query", model=self.model
        )[0]["embedding"]
        return embedding