You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
langchain/libs/community/langchain_community/embeddings/bookend.py

91 lines
2.7 KiB
Python

"""Wrapper around Bookend AI embedding models."""
import json
from typing import Any, List
import requests
from langchain_core.embeddings import Embeddings
from langchain_core.pydantic_v1 import BaseModel, Field
API_URL = "https://api.bookend.ai/"
DEFAULT_TASK = "embeddings"
PATH = "/models/predict"
class BookendEmbeddings(BaseModel, Embeddings):
"""Bookend AI sentence_transformers embedding models.
Example:
.. code-block:: python
from langchain_community.embeddings import BookendEmbeddings
bookend = BookendEmbeddings(
domain={domain}
api_token={api_token}
model_id={model_id}
)
bookend.embed_documents([
"Please put on these earmuffs because I can't you hear.",
"Baby wipes are made of chocolate stardust.",
])
bookend.embed_query(
"She only paints with bold colors; she does not like pastels."
)
"""
domain: str
"""Request for a domain at https://bookend.ai/ to use this embeddings module."""
api_token: str
"""Request for an API token at https://bookend.ai/ to use this embeddings module."""
model_id: str
"""Embeddings model ID to use."""
auth_header: dict = Field(default_factory=dict)
def __init__(self, **kwargs: Any):
super().__init__(**kwargs)
self.auth_header = {"Authorization": "Basic {}".format(self.api_token)}
def embed_documents(self, texts: List[str]) -> List[List[float]]:
"""Embed documents using a Bookend deployed embeddings model.
Args:
texts: The list of texts to embed.
Returns:
List of embeddings, one for each text.
"""
result = []
headers = self.auth_header
headers["Content-Type"] = "application/json; charset=utf-8"
params = {
"model_id": self.model_id,
"task": DEFAULT_TASK,
}
for text in texts:
data = json.dumps(
{"text": text, "question": None, "context": None, "instruction": None}
)
r = requests.request(
"POST",
API_URL + self.domain + PATH,
headers=headers,
params=params,
data=data,
)
result.append(r.json()[0]["data"])
return result
def embed_query(self, text: str) -> List[float]:
"""Embed a query using a Bookend deployed embeddings model.
Args:
text: The text to embed.
Returns:
Embeddings for the text.
"""
return self.embed_documents([text])[0]