基于neo4j存储知识树-mac

1、安装jdk21 for mac(jdk-21_macos-aarch64_bin.dmg)

2、安装neo4j for mac(neo4j-community-5.26.0-unix.tar.gz)

3、使用默认neo4j/neo4j登录http://localhost:7474

修改登录密码，可以使用生成按钮生成密码，连接数据库，默认设置为neo4j://localhost:7687。

3、具体代码如下：

import networkx as nx
import json
import requests
import matplotlib.pyplot as plt
from matplotlib.font_manager import FontProperties
from py2neo import Graph, Node, Relationship

# 设置字体路径
font_path = "/System/Library/Fonts/STHeiti Light.ttc"  # macOS系统自带的黑体字体路径
font_prop = FontProperties(fname=font_path)

# 连接到Neo4j数据库
def connect_to_neo4j(uri, user, password):
    graph = Graph(uri, auth=(user, password))
    return graph

# 将知识树保存到Neo4j数据库
def save_knowledge_tree_to_neo4j(graph, tree):
    # 清空数据库
    graph.delete_all()
    # 添加节点和关系
    for node in tree.nodes():
        neo4j_node = Node("Concept", name=node)
        graph.create(neo4j_node)
    for u, v in tree.edges():
        u_node = graph.nodes.match("Concept", name=u).first()
        v_node = graph.nodes.match("Concept", name=v).first()
        if u_node and v_node:
            rel = Relationship(u_node, "ANSWERED_BY", v_node)
            graph.create(rel)

# 可视化知识树
def visualize_tree(tree):
    pos = nx.spring_layout(tree, k=0.5, iterations=20)  # 节点布局
    nx.draw(tree, pos, with_labels=True, node_size=500, node_color="skyblue", font_size=8, font_weight='bold', font_family=font_prop.get_name())
    plt.title("知识树", fontproperties=font_prop)
    plt.axis('off')  # 关闭坐标轴
    plt.show()

# 示例对话内容
dialogue = """
客户：你好，我想了解一下产品的保修政策。
坐席：您好！我们的产品保修期为一年，从购买之日起计算。
客户：如果产品在保修期内出现故障，我该怎么办？
坐席：您可以在保修期内将产品送到我们指定的维修点进行免费维修。
客户：维修点在哪里？
坐席：您可以在我们的官方网站上查询最近的维修点地址。
客户：好的，谢谢！
坐席：不客气，祝您生活愉快！
"""

# 手动解析对话内容
def parse_dialogue(dialogue):
    lines = dialogue.strip().split("\n")
    questions = []
    answers = []
    for i in range(0, len(lines), 2):
        question = lines[i].replace("客户：", "").strip()
        answer = lines[i + 1].replace("坐席：", "").strip()
        questions.append(question)
        answers.append(answer)
    return questions, answers

questions, answers = parse_dialogue(dialogue)
print("问题：", questions)
print("答案：", answers)

def build_knowledge_tree(questions, answers):
    tree = nx.DiGraph()
    root = "根据对话初始化的知识树"
    tree.add_node(root)
    
    for q, a in zip(questions, answers):
        tree.add_node(q)
        tree.add_edge(root, q)
        tree.add_node(a)
        tree.add_edge(q, a)
    return tree

knowledge_tree = build_knowledge_tree(questions, answers)

# 可视化知识树
visualize_tree(knowledge_tree)

# 生成需要补充的问题
def generate_supplement_questions(questions, answers):
    prompt = "基于以上对话内容，您认为还需要补充哪些问题？请以问题的形式列出。\n"
    for q, a in zip(questions, answers):
        prompt += f"问题：{q}\n答案：{a}\n"
    prompt += "需要补充的问题："
    print('user:', prompt)
    # 调用 Ollama API 生成补充问题
    data = {
        "model": "qwen2.5:14b",
        "prompt": prompt,
        "stream": False,
        "temperature": 0.7,
        "max_tokens": 200
    }
    try:
        response = requests.post("http://127.0.0.1:11434/api/generate", json=data)
        if response.status_code == 200:
            return response.json().get("response", "")
        else:
            return f"API 请求失败，状态码：{response.status_code}"
    except Exception as e:
        return f"API 请求失败，错误信息：{e}"

# 生成需要补充的问题
supplement_questions = generate_supplement_questions(questions, answers)
print("assistant:需要补充的问题：", supplement_questions)

# 专家回答
def expert_answers(supplement_questions):
    print("专家，请回答以下问题：")
    print(supplement_questions)
    answers = []
    for q in supplement_questions.split("\n"):
        if q.strip():
            print(f"问题：{q}")
            answer = input("专家回答输入：")
            answers.append(answer)
    return answers

# 专家回答
expert_answers_list = expert_answers(supplement_questions)
print("专家的回答：", expert_answers_list)

# 构建知识树
def build_knowledge_tree(questions, answers, supplement_questions, expert_answers):
    tree = nx.DiGraph()
    root = "专家补充回答后的知识树"
    tree.add_node(root)
    
    for q, a in zip(questions, answers):
        tree.add_node(q)
        tree.add_edge(root, q)
        tree.add_node(a)
        tree.add_edge(q, a)
    
    for q, a in zip(supplement_questions.split("\n"), expert_answers):
        if q.strip():
            tree.add_node(q)
            tree.add_edge(root, q)
            tree.add_node(a)
            tree.add_edge(q, a)
    
    return tree

# 构建知识树
knowledge_tree = build_knowledge_tree(questions, answers, supplement_questions, expert_answers_list)

# 连接到Neo4j数据库
neo4j_uri = "neo4j://localhost:7687"
neo4j_user = "neo4j"
neo4j_password = "密码"  # 替换为你的Neo4j密码（生成或设置的密码）
graph = connect_to_neo4j(neo4j_uri, neo4j_user, neo4j_password)

# 将知识树保存到Neo4j数据库
save_knowledge_tree_to_neo4j(graph, knowledge_tree)

# 查询功能
def query_knowledge_tree(graph, question):
    query = """
    MATCH (q:Concept {name: $question})-[:ANSWERED_BY]->(a:Concept)
    RETURN a.name AS answer
    """
    result = graph.run(query, question=question)
    return [record["answer"] for record in result]

# 测试查询功能
for question in questions:
    result = query_knowledge_tree(graph, question)
    print("原始问题：", question)
    if result:
        print(f"查询结果：")
        for node in result:
            print(node)
    else:
        print("未找到相关问题的答案。")

for question in supplement_questions.split("\n"):
    result = query_knowledge_tree(graph, question)
    print("补充问题：", question)
    if result:
        print(f"查询结果：")
        for node in result:
            print(node)
    else:
        print("未找到相关问题的答案。")

4、显示效果：