使用GraphRAG和RAG进行LLM检索

2024年10月11日 由 alex 发表 44 0

RAG 是生成式 AI 领域的一大创新,它使人们能够连接他们的个人文档(如文本文件、PDF、视频等)并与 LLM 进行交流。最近推出了 RAG 的升级版 GraphRAG,它使用知识图谱通过 LLM 执行 RAG 检索。


RAG 和 GraphRAG 都有各自的优点和局限性。虽然 RAG 带来了向量相似度的强大功能,但 GraphRAG 使用图形分析和知识图谱来获得更清晰的答案。


如果两者可以一起用于检索会怎么样?


混合RAG

HybridRAG 是一个融合了 RAG 和 GraphRAG 的高级框架。这种集成旨在提高信息检索的准确性和上下文相关性。简而言之,HybridRAG 使用来自两个检索系统(RAG 和 GraphRAG)的上下文,最终输出是两个系统的混合。


HybridRAG 的优势

  • 提高准确性:通过利用结构化推理和灵活检索,与单独使用 VectorRAG 或 GraphRAG 相比,HybridRAG 可以提供更精确的答案。
  • 增强情境理解:这种结合可以更深入地理解实体之间的关系以及它们出现的情境。
  • 动态推理:知识图谱可以动态更新,使系统能够随时适应新的信息。


那么,让我们尝试使用 LangChain 构建一个 HybridRAG 系统


我将使用一个文件 Moon.txt 来演示这个超级英雄的故事。检查下面的内容


In the bustling city of Lunaris, where the streets sparkled with neon lights and the moon hung low in the sky, lived an unassuming young man named Max. By day, he was a mild-mannered astronomer, spending his hours studying the stars and dreaming of adventures beyond Earth. But as the sun dipped below the horizon, Max transformed into something extraordinary—Moon Man, the guardian of the night sky.
Max’s transformation began with a mysterious encounter. One fateful evening, while gazing through his telescope, a brilliant flash of light erupted from the moon. A celestial being, shimmering with silver light, descended and bestowed upon him a magical amulet. “With this, you shall harness the power of the moon,” the being declared. “Use it wisely, for the night sky needs a hero.”
With the amulet around his neck, Max felt energy coursing through him. He could leap great distances, manipulate moonlight, and even communicate with nocturnal creatures. He vowed to protect his city from the shadows that lurked in the night.
As Moon Man, Max donned a sleek, silver suit adorned with celestial patterns that glimmered like the stars. With his newfound abilities, he patrolled the city, rescuing lost pets, helping stranded motorists, and even thwarting petty criminals. The citizens of Lunaris began to whisper tales of their mysterious hero, who appeared under the glow of the moon.
One night, as he soared through the sky, he encountered a gang of thieves attempting to steal a priceless artifact from the Lunaris Museum. With a flick of his wrist, he summoned a beam of moonlight that blinded the thieves, allowing him to swoop in and apprehend them. The city erupted in cheers, and Moon Man became a beloved figure.
However, peace in Lunaris was short-lived. A dark force emerged from the depths of the cosmos—an evil sorceress named Umbra, who sought to extinguish the moon’s light and plunge the world into eternal darkness. With her army of shadow creatures, she began to wreak havoc, stealing the moon’s energy and spreading fear among the citizens.
Moon Man knew he had to confront this new threat. He gathered his courage and sought the wisdom of the celestial being who had granted him his powers. “To defeat Umbra, you must harness the full power of the moon,” the being advised. “Only then can you restore balance to the night sky.”
With determination in his heart, Moon Man prepared for the ultimate battle. He climbed to the highest peak in Lunaris, where the moon shone brightest, and focused on channeling its energy. As Umbra and her shadow creatures descended upon the city, Moon Man unleashed a magnificent wave of moonlight, illuminating the darkness.
The battle raged on, with Umbra conjuring storms of shadows and Moon Man countering with beams of silver light. The clash of powers lit up the night sky, creating a dazzling display that captivated the citizens below. In a final, desperate move, Moon Man summoned all his strength and unleashed a powerful blast of moonlight that enveloped Umbra, banishing her to the farthest reaches of the cosmos.
With Umbra defeated, the moon’s light returned to its full glory, and the city of Lunaris rejoiced. Max, still in his Moon Man guise, stood atop the highest building, watching as the citizens celebrated their hero. They had learned the importance of hope and courage, even in the darkest of times.
From that day forward, Moon Man became a symbol of resilience and bravery. Max continued to protect Lunaris, knowing that as long as the moon shone brightly, he would always be there to guard the night sky. And so, the legend of Moon Man lived on, inspiring generations to look up at the stars and believe in the extraordinary.
As the years passed, stories of Moon Man spread beyond Lunaris, becoming a beacon of hope for those who felt lost in the darkness. Children would gaze at the moon, dreaming of adventures, and Max would smile, knowing that he had made a difference. For in the heart of every dreamer, the spirit of Moon Man lived on, reminding them that even the smallest light can shine brightly against the shadows.


1. 导入包并设置 LLM 端嵌入模型(用于标准 RAG)


import os
from langchain_experimental.graph_transformers import LLMGraphTransformer
from langchain_core.documents import Document
from langchain_community.graphs.networkx_graph import NetworkxEntityGraph
from langchain.chains import GraphQAChain
from langchain.text_splitter import CharacterTextSplitter
from langchain.document_loaders import TextLoader
from langchain.chains import RetrievalQA
from langchain.vectorstores import Chroma
from langchain_google_genai import GoogleGenerativeAI,GoogleGenerativeAIEmbeddings
GOOGLE_API_KEY=''
embeddings = GoogleGenerativeAIEmbeddings(model="models/embedding-001",google_api_key=GOOGLE_API_KEY)
llm = GoogleGenerativeAI(model="gemini-pro",google_api_key=GOOGLE_API_KEY)


2. 接下来,我们将在文件“Moon.txt”上创建一个用于 GraphRAG 实现的函数(链对象)


def graphrag():
        with open('Moon.txt', 'r') as file:
            content = file.read()
        documents = [Document(page_content=content)]
        llm_transformer = LLMGraphTransformer(llm=llm)
        graph_documents = llm_transformer.convert_to_graph_documents(documents)
        graph = NetworkxEntityGraph()
        # Add nodes to the graph
        for node in graph_documents[0].nodes:
            graph.add_node(node.id)
        # Add edges to the graph
        for edge in graph_documents[0].relationships:
            graph._graph.add_edge(
                    edge.source.id,
                    edge.target.id,
                    relation=edge.type,
                )
            graph._graph.add_edge(
                    edge.target.id,
                    edge.source.id,
                    relation=edge.type+" by",
                )
        chain = GraphQAChain.from_llm(
            llm=llm, 
            graph=graph, 
            verbose=True
        )
        
        return chain


3. 同样,我们将创建一个函数来在同一个文件上实现标准 RAG


def rag():
        #Document Loader
        loader = TextLoader('Moon.txt')
        data = loader.load()
        #Document Transformer
        text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
        texts = text_splitter.split_documents(data)
        #Vector DB
        docsearch = Chroma.from_documents(texts, embeddings)
        #Hyperparameters to know
        retriever = docsearch.as_retriever(search_type='similarity_score_threshold',search_kwargs={"k": 7,"score_threshold":0.3})
        qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=retriever)
        
        return qa


4. 为两种类型的 RAG 创建对象


standard_rag = rag()
graph_rag = graphrag()


5. 实施 HybridRAG 的时间


def hybrid_rag(query,standard_rag,graph_rag):hybrid_rag(query,standard_rag,graph_rag):
    result1 = standard_rag.run(query)
    
    print("Standard RAG:",result1)
    result2 = graph_rag.run(query)
    
    
    print("Graph RAG:",result2)
    prompt = "Generate a final answer using the two context given : Context 1: {} \n Context 2: {} \n Question: {}".format(result1,result2,query)
    return llm(prompt)
query = "Some characteristics of Moon Man"
hybrid = hybrid_rag(query,standard_rag,graph_rag)
print("Hybrid:",hybrid)


如你所见:

  1. 我们针对给定的提示分别运行了标准 RAG 和 GraphRAG。
  2. 一旦检索到答案,就会以这两个答案为上下文生成最终答案。


说到输出,最终的 HybridRAG 确实从两个检索中获取了上下文,并产生了更好的结果。两个 RAG 系统都遗漏了一些要点,但最终 HybridRAG 结合起来给出了完美的答案。


STANDARD RAG: 
 Here are some characteristics of Moon Man, based on the story:
* **Brave:** He confronts danger and fights villains like Umbra.
* **Powerful:** He has superhuman abilities granted by the amulet.
* **Protective:** He safeguards Lunaris and its citizens.
* **Determined:** He doesn't give up, even when facing powerful enemies.
* **Compassionate:** He helps those in need, like rescuing lost pets.
* **Humble:** Despite his powers, he remains grounded and dedicated to his city. 


> Entering new GraphQAChain chain...
Entities Extracted:
Moon Man
Full Context:
Moon Man PROTECTS night sky
Moon Man WEARS silver suit
Moon Man PROTECTED Lunaris
Moon Man CAPTURED thieves
Moon Man DEFEATED Umbra
Moon Man INSPIRES hope
Moon Man INSPIRES courage
> Finished chain.
 @@ 
 GRAPH RAG: 
 Helpful Answer: 
* Protective (protects night sky, protected Lunaris)
* Courageous and Inspiring (inspires hope, inspires courage)
* Strong (captured thieves, defeated Umbra) 

 @@ 
 HYBRID RAG: 
 Moon Man is the **protective** champion of Lunaris, using his **strength** and **courage** to defend its citizens and the night sky. He is **powerful** and **determined**, facing down villains like Umbra without giving up. Yet, despite his abilities, he remains **humble** and **compassionate**, always willing to help those in need. Moon Man is a true inspiration, reminding everyone that even in darkness, hope and heroism can shine through. 


希望这篇文章对你有用,你也可以在你的数据集上试试!

文章来源:https://medium.com/data-science-in-your-pocket/hybrid-rag-graphrag-rag-combined-for-retrieval-using-llms-1011fb84cdbb
欢迎关注ATYUN官方公众号
商务合作及内容投稿请联系邮箱:bd@atyun.com
评论 登录
热门职位
Maluuba
20000~40000/月
Cisco
25000~30000/月 深圳市
PilotAILabs
30000~60000/年 深圳市
写评论取消
回复取消