跳到主要内容
Open In ColabOpen on GitHub

Nebius 文本嵌入

Nebius AI Studio 通过统一接口提供对高质量嵌入模型的 API 访问。Nebius 嵌入模型将文本转换为捕获语义含义的数值向量，这使得它们可用于语义搜索、聚类和推荐等各种应用。

概述​

NebiusEmbeddings 类通过 LangChain 提供对 Nebius AI Studio 嵌入模型的访问。这些嵌入可用于语义搜索、文档相似性以及需要文本向量表示的其他自然语言处理 (NLP) 任务。

集成详情​

  • 提供商：Nebius AI Studio
  • 模型类型：文本嵌入模型
  • 主要用例：生成文本的向量表示，用于语义相似性匹配和检索
  • 可用模型：包括 BAAI/bge-en-icl 在内的各种嵌入模型及其他
  • 维度：因模型而异（通常为 1024-4096 维）

设置​

安装​

Nebius 集成可以通过 pip 安装

%pip install --upgrade langchain-nebius

凭证​

Nebius 需要一个 API 密钥，该密钥可以作为初始化参数 api_key 传递，或者设置为环境变量 NEBIUS_API_KEY。您可以通过在 Nebius AI Studio 上创建账户来获取 API 密钥。

import getpass
import os

# Make sure you've set your API key as an environment variable
if "NEBIUS_API_KEY" not in os.environ:
    os.environ["NEBIUS_API_KEY"] = getpass.getpass("Enter your Nebius API key: ")

实例化​

NebiusEmbeddings 类可以使用可选参数（API 密钥和模型名称）进行实例化

from langchain_nebius import NebiusEmbeddings

# Initialize the embeddings model
embeddings = NebiusEmbeddings(
    # api_key="YOUR_API_KEY",  # You can pass the API key directly
    model="BAAI/bge-en-icl"  # The default embedding model
)

可用模型​

支持的模型列表可在 https://studio.nebius.com/?modality=embedding 查看

索引和检索​

嵌入模型常用于检索增强生成 (RAG) 流程中，既用于数据索引，也用于后续检索。以下示例演示了如何将 NebiusEmbeddings 与向量存储结合使用以进行文档检索。

from langchain_community.vectorstores import FAISS
from langchain_core.documents import Document

# Prepare documents
docs = [
    Document(
        page_content="Machine learning algorithms build mathematical models based on sample data"
    ),
    Document(page_content="Deep learning uses neural networks with many layers"),
    Document(page_content="Climate change is a major global environmental challenge"),
    Document(
        page_content="Neural networks are inspired by the human brain's structure"
    ),
]

# Create vector store
vector_store = FAISS.from_documents(docs, embeddings)

# Perform similarity search
query = "How does the brain influence AI?"
results = vector_store.similarity_search(query, k=2)

print("Search results for query:", query)
for i, doc in enumerate(results):
    print(f"Result {i + 1}: {doc.page_content}")
API 参考：FAISS | Document
Search results for query: How does the brain influence AI?
Result 1: Neural networks are inspired by the human brain's structure
Result 2: Deep learning uses neural networks with many layers

与 InMemoryVectorStore 配合使用​

您也可以将 InMemoryVectorStore 用于轻量级应用

from langchain_core.vectorstores import InMemoryVectorStore

# Create a sample text
text = "LangChain is a framework for developing applications powered by language models"

# Create a vector store
vectorstore = InMemoryVectorStore.from_texts(
    [text],
    embedding=embeddings,
)

# Use as a retriever
retriever = vectorstore.as_retriever()

# Retrieve similar documents
docs = retriever.invoke("What is LangChain?")
print(f"Retrieved document: {docs[0].page_content}")
API 参考：InMemoryVectorStore
Retrieved document: LangChain is a framework for developing applications powered by language models

直接用法​

您可以直接使用 NebiusEmbeddings 类为文本生成嵌入，而无需使用向量存储。

嵌入单个文本​

您可以使用 embed_query 方法嵌入单个文本

query = "What is machine learning?"
query_embedding = embeddings.embed_query(query)

# Check the embedding dimension
print(f"Embedding dimension: {len(query_embedding)}")
print(f"First few values: {query_embedding[:5]}")
Embedding dimension: 4096
First few values: [0.007419586181640625, 0.002246856689453125, 0.00193023681640625, -0.0066070556640625, -0.0179901123046875]

嵌入多个文本​

您可以使用 embed_documents 方法一次性嵌入多个文本

documents = [
    "Machine learning is a branch of artificial intelligence",
    "Deep learning is a subfield of machine learning",
    "Natural language processing deals with interactions between computers and human language",
]

document_embeddings = embeddings.embed_documents(documents)

# Check the results
print(f"Number of document embeddings: {len(document_embeddings)}")
print(f"Each embedding has {len(document_embeddings[0])} dimensions")
Number of document embeddings: 3
Each embedding has 4096 dimensions

异步支持​

NebiusEmbeddings 支持异步操作

import asyncio


async def generate_embeddings_async():
    # Embed a single query
    query_result = await embeddings.aembed_query("What is the capital of France?")
    print(f"Async query embedding dimension: {len(query_result)}")

    # Embed multiple documents
    docs = [
        "Paris is the capital of France",
        "Berlin is the capital of Germany",
        "Rome is the capital of Italy",
    ]
    docs_result = await embeddings.aembed_documents(docs)
    print(f"Async document embeddings count: {len(docs_result)}")


await generate_embeddings_async()
Async query embedding dimension: 4096
Async document embeddings count: 3

文档相似度示例​

import numpy as np
from scipy.spatial.distance import cosine

# Create some documents
documents = [
    "Machine learning algorithms build mathematical models based on sample data",
    "Deep learning uses neural networks with many layers",
    "Climate change is a major global environmental challenge",
    "Neural networks are inspired by the human brain's structure",
]

# Embed the documents
embeddings_list = embeddings.embed_documents(documents)


# Function to calculate similarity
def calculate_similarity(embedding1, embedding2):
    return 1 - cosine(embedding1, embedding2)


# Print similarity matrix
print("Document Similarity Matrix:")
for i, emb_i in enumerate(embeddings_list):
    similarities = []
    for j, emb_j in enumerate(embeddings_list):
        similarity = calculate_similarity(emb_i, emb_j)
        similarities.append(f"{similarity:.4f}")
    print(f"Document {i + 1}: {similarities}")
Document Similarity Matrix:
Document 1: ['1.0000', '0.8282', '0.5811', '0.7985']
Document 2: ['0.8282', '1.0000', '0.5897', '0.8315']
Document 3: ['0.5811', '0.5897', '1.0000', '0.5918']
Document 4: ['0.7985', '0.8315', '0.5918', '1.0000']

API 参考​

有关 Nebius AI Studio API 的更多详细信息，请访问 Nebius AI Studio 文档。