from __future__ import annotations
import logging
from typing import Any, Callable, Dict, List, Optional
import requests
from langchain_core.embeddings import Embeddings
from langchain_core.pydantic_v1 import BaseModel, Extra, SecretStr, root_validator
from langchain_core.utils import convert_to_secret_str, get_from_dict_or_env
from tenacity import (
before_sleep_log,
retry,
stop_after_attempt,
wait_exponential,
)
logger = logging.getLogger(__name__)
def _create_retry_decorator() -> Callable[[Any], Any]:
"""返回一个用于重试的坚韧性装饰器。"""
multiplier = 1
min_seconds = 1
max_seconds = 4
max_retries = 6
return retry(
reraise=True,
stop=stop_after_attempt(max_retries),
wait=wait_exponential(multiplier=multiplier, min=min_seconds, max=max_seconds),
before_sleep=before_sleep_log(logger, logging.WARNING),
)
[docs]def embed_with_retry(embeddings: MiniMaxEmbeddings, *args: Any, **kwargs: Any) -> Any:
"""使用tenacity来重试完成调用。"""
retry_decorator = _create_retry_decorator()
@retry_decorator
def _embed_with_retry(*args: Any, **kwargs: Any) -> Any:
return embeddings.embed(*args, **kwargs)
return _embed_with_retry(*args, **kwargs)
[docs]class MiniMaxEmbeddings(BaseModel, Embeddings):
"""MiniMax的嵌入式服务。
要使用，您应该设置环境变量``MINIMAX_GROUP_ID``和``MINIMAX_API_KEY``，并将其作为命名参数传递给构造函数。
示例：
.. code-block:: python
from langchain_community.embeddings import MiniMaxEmbeddings
embeddings = MiniMaxEmbeddings()
query_text = "This is a test query."
query_result = embeddings.embed_query(query_text)
document_text = "This is a test document."
document_result = embeddings.embed_documents([document_text])"""
endpoint_url: str = "https://api.minimax.chat/v1/embeddings"
"""要使用的端点URL。"""
model: str = "embo-01"
"""嵌入模型的名称。"""
embed_type_db: str = "db"
"""对于embed_documents"""
embed_type_query: str = "query"
"""对于embed_query"""
minimax_group_id: Optional[str] = None
"""MiniMax API的群组ID。"""
minimax_api_key: Optional[SecretStr] = None
"""MiniMax API 的 API 密钥。"""
class Config:
"""此pydantic对象的配置。"""
extra = Extra.forbid
@root_validator()
def validate_environment(cls, values: Dict) -> Dict:
"""验证环境中是否存在group id和api key。"""
minimax_group_id = get_from_dict_or_env(
values, "minimax_group_id", "MINIMAX_GROUP_ID"
)
minimax_api_key = convert_to_secret_str(
get_from_dict_or_env(values, "minimax_api_key", "MINIMAX_API_KEY")
)
values["minimax_group_id"] = minimax_group_id
values["minimax_api_key"] = minimax_api_key
return values
[docs] def embed(
self,
texts: List[str],
embed_type: str,
) -> List[List[float]]:
payload = {
"model": self.model,
"type": embed_type,
"texts": texts,
}
# HTTP headers for authorization
headers = {
"Authorization": f"Bearer {self.minimax_api_key.get_secret_value()}", # type: ignore[union-attr]
"Content-Type": "application/json",
}
params = {
"GroupId": self.minimax_group_id,
}
# send request
response = requests.post(
self.endpoint_url, params=params, headers=headers, json=payload
)
parsed_response = response.json()
# check for errors
if parsed_response["base_resp"]["status_code"] != 0:
raise ValueError(
f"MiniMax API returned an error: {parsed_response['base_resp']}"
)
embeddings = parsed_response["vectors"]
return embeddings
[docs] def embed_documents(self, texts: List[str]) -> List[List[float]]:
"""使用MiniMax嵌入端点嵌入文档。
参数:
texts: 要嵌入的文本列表。
返回:
每个文本的嵌入列表。
"""
embeddings = embed_with_retry(self, texts=texts, embed_type=self.embed_type_db)
return embeddings
[docs] def embed_query(self, text: str) -> List[float]:
"""使用MiniMax嵌入端点嵌入一个查询。
参数：
text: 要嵌入的文本。
返回：
文本的嵌入。
"""
embeddings = embed_with_retry(
self, texts=[text], embed_type=self.embed_type_query
)
return embeddings[0]