ollama list

ollama run qwen2.5-coder:latest

ollama show qwen2.5-coder:latest

Model    architecture        qwen2    parameters          7.6B    context length      32768    embedding length    3584    quantization        Q4_K_M
  System    You are Qwen, created by Alibaba Cloud. You are a helpful assistant.
  License    Apache License    Version 2.0, January 2004

# 提供类型提示from typing import Any, Dict, List, Tuple
# 标准库import astimport loggingimport reimport warnings
# 第三方包 - 数据操作import pandas as pdfrom tqdm import tqdm
# 第三方包 - 环境配置和数据库from dotenv import load_dotenvfrom neo4j import GraphDatabase
# 第三方包 - 错误处理与重试逻辑from tenacity import retry, stop_after_attempt, wait_exponential
# Langchain - 核心库from langchain.chains import GraphCypherQAChainfrom langchain.prompts import PromptTemplatefrom langchain_core.documents import Document
# Langchain - 模型和连接器from langchain_google_genai import ChatGoogleGenerativeAI, GoogleGenerativeAIfrom langchain_ollama.llms import OllamaLLM
# Langchain - 图相关库与体验包from langchain_community.graphs import Neo4jGraphfrom langchain_experimental.graph_transformers import LLMGraphTransformer
# 抑制警告warnings.filterwarnings('ignore')
# 加载环境变量load_dotenv()

movies = pd.read_csv('data/wiki_movies.csv') # 按照你自己实际保存该数据集的路径进行修改movies.head()

 def clean_data(df: pd.DataFrame) -> pd.DataFrame:    """清理并预处理 DataFrame.        参数:        data: 输入 DataFrame            返回值:        清理后的 DataFrame    """    df.drop(["Wiki Page"], axis=1, inplace=True)
    # 删除重复项    df = df.drop_duplicates(subset='Title', keep='first')        # 获取对象类型的列    col_obj = df.select_dtypes(include=["object"]).columns        # 清理字符串列    for col in col_obj:        # Strip whitespace        df[col] = df[col].str.strip()                # 替换未知/空值        df[col] = df[col].apply(            lambda x: None if pd.isna(x) or x.lower() in ["", "unknown"]             else x.capitalize()        )        # 删除包含任何空值的行    df = df.dropna(how="any", axis=0)        return df
movies = clean_data(movies).head(1000)movies.head()

class Neo4jConnection:    def __init__(self, uri, user, password):        self.driver = GraphDatabase.driver(uri, auth=(user, password))
    def close(self):        self.driver.close()        print("连接已关闭。")
    def reset_database(self):        with self.driver.session() as session:            session.run("MATCH (n) DETACH DELETE n")            print("数据库重置成功！")
    def execute_query(self, query, parameters=None):        with self.driver.session() as session:            result = session.run(query, parameters or {})            return [record for record in result]

uri = "bolt://localhost:7687"user = "neo4j"password = "mysecret"conn = Neo4jConnection(uri, user, password)conn.reset_database()

def parse_number(value: Any, target_type: type) -> Optional[float]:    """将字符串解析为数字，并进行适当的错误处理。"""    if pd.isna(value):        return None    try:        cleaned = str(value).strip().replace(',', '')        return target_type(cleaned)    except (ValueError, TypeError):        return None
def clean_text(text: str) -> str:    """清理并规范化文本字段。"""    if pd.isna(text):        return ""    return str(text).strip().title()

def load_movies_to_neo4j(movies_df: pd.DataFrame, connection: GraphDatabase) -> None:    """将电影数据加载到 Neo4j 中，并进行进度跟踪和错误处理。"""        logger = logging.getLogger(__name__)    logger.setLevel(logging.INFO)        # 查询模板    MOVIE_QUERY = """        MERGE (movie:Movie {title: $title})        SET movie.year = $year,            movie.origin = $origin,            movie.genre = $genre,            movie.plot = $plot    """        DIRECTOR_QUERY = """        MATCH (movie:Movie {title: $title})        MERGE (director:Director {name: $name})        MERGE (director)-[:DIRECTED]->(movie)    """        ACTOR_QUERY = """        MATCH (movie:Movie {title: $title})        MERGE (actor:Actor {name: $name})        MERGE (actor)-[:ACTED_IN]->(movie)    """
    # 处理每部电影    for _, row in tqdm(movies_df.iterrows(), total=len(movies_df), desc="正在加载电影数据"):        try:            # 准备电影相关实体的参数            movie_params = {                "title": clean_text(row["Title"]),                "year": parse_number(row["Release Year"], int),                "origin": clean_text(row["Origin/Ethnicity"]),                "genre": clean_text(row["Genre"]),                "plot": str(row["Plot"]).strip()            }                        # 创建电影节点            connection.execute_query(MOVIE_QUERY, parameters=movie_params)                        # Process directors            for director in str(row["Director"]).split(" and "):                director_params = {                    "name": clean_text(director),                    "title": movie_params["title"]                }                connection.execute_query(DIRECTOR_QUERY, parameters=director_params)                        # 处理演员相关数据            if pd.notna(row["Cast"]):                for actor in row["Cast"].split(","):                    actor_params = {                        "name": clean_text(actor),                        "title": movie_params["title"]                    }                    connection.execute_query(ACTOR_QUERY, parameters=actor_params)                            except Exception as e:            logger.error(f"加载 {row['Title']} 错误: {str(e)}")            continue
    logger.info("加载电影数据到 Neo4j 已完成。")

load_movies_to_neo4j(movies, conn)

query = """MATCH (m:Movie)-[:ACTED_IN]-(a:Actor)RETURN m.title, a.nameLIMIT 10;"""results = conn.execute_query(query)for record in results:    print(record)

[<Record m.title='Daniel Boone' a.name='William Craven'>, <Record m.title='Daniel Boone' a.name='Florence Lawrence'>, <Record m.title='Laughing Gas' a.name='Bertha Regustus'>, <Record m.title='Laughing Gas' a.name='Edward Boulden'>, <Record m.title="A Drunkard'S Reformation" a.name='Arthur V. Johnson'>, <Record m.title='The Adventures Of Dollie' a.name='Arthur V. Johnson'>, <Record m.title='A Calamitous Elopement' a.name='Linda Arvidson'>, <Record m.title='The Adventures Of Dollie' a.name='Linda Arvidson'>, <Record m.title='The Black Viper' a.name='D. W. Griffith'>, <Record m.title='A Calamitous Elopement' a.name='Harry Solter'>]

MATCH p=(m:Movie)-[r]-(n)RETURN pLIMIT 100;