openGauss AGEGraph + 大模型实现 GraphRAG，助力更强 RAG

CONTAINER ID        IMAGE                                                  COMMAND                  CREATED             STATUS                          PORTS                                              NAMES
253714c9c869        opengauss:7.0.0-rc1                                    "entrypoint.sh gauss…"   8 minutes ago       Up 8 minutes                    0.0.0.0:6543->5432/tcp                             opengauss-age
复制

#进入openGauss容器中 253714c9c869为 CONTAINER ID
docker exec -it 253714c9c869 bash


#切换到omm用户
su omm


#omm用户，连接数据库，默认使用omm数据库
gsql -r


#连接数据库后，omm数据库安装age插件
create extension age;


#退出数据库连接
\q
复制

pip install -U langchain_community langchain langgraph  langchain-ollama langchain-experimental  langchain-openai
复制

git clone https://gitee.com/lin-qiang123/langchain.git
复制

pip show langchain_community
复制

Name: langchain-community
Version: 0.3.20
Summary: Community contributed LangChain integrations.
Home-page:
Author:
Author-email:
License: MIT
Location: c:\users\test\appdata\local\programs\python\python310\lib\site-packages  #依赖库安装的位置
Requires: numpy, PyYAML, aiohttp, tenacity, dataclasses-json, langchain, httpx-sse, SQLAlchemy, pydantic-settings, langchain-core, langsmith, requests
Required-by: langchain-experimental
复制

from langchain_community.llms import Tongyi
import os


os.environ["DASHSCOPE_API_KEY"] = "sk-**"
graph_llm =Tongyi(model="qwen-plus", temperature=0, base_url="https://dashscope.aliyuncs.com/compatible-mode/v1")
复制

from langchain_core.documents import Document
from langchain_experimental.graph_transformers import LLMGraphTransformer


#设置需要提取的实体类型和关系
llm_transformer = LLMGraphTransformer(
    llm=graph_llm,
    allowed_nodes=["Person", "Organization", "Location", "Award", "ResearchField"],
    allowed_relationships = ["SPOUSE", "AWARD", "FIELD_OF_RESEARCH", "WORKS_AT", "IN_LOCATION"],
)


text = """
Marie Curie, 7 November 1867 – 4 July 1934, was a Polish and naturalised-French physicist and chemist who conducted pioneering research on radioactivity.
She was the first woman to win a Nobel Prize, the first person to win a Nobel Prize twice, and the only person to win a Nobel Prize in two scientific fields.
Her husband, Pierre Curie, was a co-winner of her first Nobel Prize, making them the first-ever married couple to win the Nobel Prize and launching the Curie family legacy of five Nobel Prizes.
She was, in 1906, the first woman to become a professor at the University of Paris.
Also, Robin Williams.
"""


# 将文本转化为图
documents = [Document(page_content=text)]
graph_documents = llm_transformer.convert_to_graph_documents(documents)




print(f"Nodes from graph doc:{graph_documents[0].nodes}")
print(f"Relationships from graph doc:{graph_documents[0].relationships}")
复制

Nodes from graph doc:[Node(id='Nobel Prize', type='Award', properties={}), Node(id='University of Paris', type='Organization', properties={}), Node(id='Marie Curie', type='Person', properties={}), Node(id='radioactivity', type='ResearchField', properties={}), Node(id='Pierre Curie', type='Person', properties={})]
Relationships from graph doc:[Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='radioactivity', type='ResearchField', properties={}), type='FIELD_OF_RESEARCH', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Nobel Prize', type='Award', properties={}), type='AWARD', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Nobel Prize', type='Award', properties={}), type='AWARD', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Pierre Curie', type='Person', properties={}), type='SPOUSE', properties={}), Relationship(source=Node(id='Pierre Curie', type='Person', properties={}), target=Node(id='Nobel Prize', type='Award', properties={}), type='AWARD', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='University of Paris', type='Organization', properties={}), type='WORKS_AT', properties={})]
复制

from langchain_community.graphs.age_graph import AGEGraph


conf = {
    "database": "omm",
    "user": "gaussdb",
    "password": "YourPassoword",
    "host": "Your IP",
    "port": 8888,
    "sslmode": "disable"
}
graph=AGEGraph(graph_name='graph_test1',conf=conf,create=True)
graph.add_graph_documents(graph_documents)
graph.refresh_schema()
复制

from langchain_core.prompts import PromptTemplate
from langchain.chains import GraphCypherQAChain


cypher_prompt = PromptTemplate(
    template="""你是 AGE Cypher 查询生成的专家。
    使用以下架构生成一个 Cypher 查询，以回答给定问题。
    不要出现name和properties和cypher


    架构：
    {schema}


    问题：{question}


    Cypher 查询：""",
    input_variables=["schema", "question"],
)


chain = GraphCypherQAChain.from_llm(
    graph_llm, graph=graph, verbose=True, allow_dangerous_requests=True, cypher_validation=True, return_intermediate_steps=True,cypher_prompt=cypher_prompt
)


question = "Who get Nobel Prize ?"
result = chain.invoke({"query": question})
复制

prompt = PromptTemplate(
    template="""You are an assistant for question-answering tasks. 
    Use the following pieces of retrieved context from a graph database to answer the question. If you don't know the answer, just say that you don't know. 
    Use two sentences maximum and keep the answer concise:
    Question: {question} 
    Graph Context: {graph_context}
    Answer: 
    """,
    input_variables=["question", "graph_context"],
)
复制

from langchain_core.output_parsers import StrOutputParser


composite_chain = prompt | graph_llm |StrOutputParser()


answer = composite_chain.invoke(
    {"question": question, "graph_context": result}
)
print(answer)
复制

Marie Curie and Pierre Curie received the Nobel Prize. They were recognized for their groundbreaking work in radioactivaity.
复制