[GraphRAG实现原理-10]进一步完善提取的实体关系图

原创于 2026-10-29 09:18:10 发布 · 312 阅读

0 ·

本内容遵循CC 4.0 BY-SA版权协议

GEO检测

收录于

34.GraphRAG实现原理

本文介绍GraphRAG中对初始实体关系图的增强方法，重点实现两个图论指标：实体度数（反映节点中心性）和关系组合度数（源+目标实体度数之和）。通过Workflow调用LLM提取结果后，在数据库表中注入这两个指标，并支持GraphML格式导出以供Gephi等工具可视化。该步骤提升了关系图的分析深度与可解释性，是构建高质量知识图谱的关键环节。

GraphRAG实现原理-09:调用LLM提取实体关系图介绍了如何利用extract_graph借助LLM的力量将所有的文档切片（TextUnit）转换成实体关系图，这张图具体体现在生成entities和relations表中。但是这张图目前仅仅描述了两个实体之间是否有关系，缺少必要的指标。在图论中，度数（Degree）的是一个非常重要的指标，它一个顶点所关联的边的数量。对于我们构建的实体关系图，如果某个实体度数越大，意味着它越接近于图的中心为止。实体的度数利用finalize_graph这个Workflow来计算并添加，这个Wokflow还会额外计算并添加关系组合度数。

1. 我们现在在哪里？

这是standard管道系列的第五篇文章，finalize_graph也是我们介绍的五个Workflow。经过前面的四道工具，我们已经拥有了documents和text_units着两个表，前者存储完整的文档，后者存储分割文档生成的切片，两者之间通过text_unit_ids和document_id保持双向引用。extract_graph这个Workflow进一步将提取出来的实体关系图保存在entities和relations表中。

load_input_documents
    └── create_base_text_units
            └── create_final_documents
                    └── extract_graph
                            └── **finalize_graph**
                                    └── extract_covariates
                                            └── create_communities
                                                    └── create_final_text_units
                                                            └──create_community_reports
                                                                    └── generate_text_embeddings

通过GraphRAG实现原理-09:调用LLM提取实体关系图提供的实例演示我们指导目前的entities和relations表具有如下的结果：

entities表：

 #   Column         Non-Null Count  Dtype 
---  ------         --------------  ----- 
 0   title          34 non-null     str   
 1   type           34 non-null     str   
 2   text_unit_ids  34 non-null     object
 3   frequency      34 non-null     int64 
 4   description    34 non-null     str

relations表：

  #   Column         Non-Null Count  Dtype  
---  ------         --------------  -----  
 0   source         33 non-null     str    
 1   target         33 non-null     str    
 2   text_unit_ids  33 non-null     object 
 3   weight         33 non-null     float64
 4   description    33 non-null     str

finalize_graph这个Workflow旨在为提取的实体关系图生成如下两个指标：

实体度数（Degree）：指一个实体节点在图中直接相连的关系（边）数量。一个实体的度数越高，说明它与越多其他实体存在关系，在图中处于越中心的连接位置；
关系组合度数（Combined Degree）：这是GraphRAG对关系边引入的一个派生指标，计算方式为该关系源实体的度数加上目标实体的度数。它用于衡量一条关系所连接的两个端点各自的活跃程度之和，帮助识别那些连接着高连接性实体的重要关系。

2. Workflow的实现

finalize_graph这个Workflow的WorkflowFunction对应如下这个run_workflow函数（所在模块目录为graphrag/workflows/finalize_graph）。它会利用PipelineRunContext上下文提供的TableProvider开启entities和relationships表，并将它们作为参数调用finalize_graph函数完成上述两个指标的计算和写入。如果GraphRagConfig提供的配置开启了记录基于GraphML的快照，还会进一步读取relationships表并将其转换成DataFrame，并将后者作为参数调用snapshot_graphml函数将最终构建的图以GraphML的形式保存下来。当我们需要可视化展现图的时候，存下来的这个快照会很有用。

async def run_workflow(
    config: GraphRagConfig,
    context: PipelineRunContext,
) -> WorkflowFunctionOutput:
    async with (
        context.output_table_provider.open(
            "entities",
            transformer=transform_entity_row,
        ) as entities_table,
        context.output_table_provider.open(
            "relationships",
            transformer=transform_relationship_row,
        ) as relationships_table,
    ):
        result = await finalize_graph(
            entities_table,
            relationships_table,
        )

    if config.snapshots.graphml:
        rels = await context.output_table_provider.read_dataframe("relationships")
        await snapshot_graphml(
            rels,
            name="graph",
            storage=context.output_storage,
        )

    return WorkflowFunctionOutput(result=result)

finalize_graph函数的处理逻辑很简单：就是读取每一条关系记录，然后调用_build_degree_map计算每个实体的度数（不考虑方向），然后基于这个度数分别调用finalize_entities将实体实体度数写入entities表，在调用finalize_relationships函数计算每个关系的关系组合度数(两个实体度数之和)并写入relationships表。最终返回前5条entities和relationships记录。

async def finalize_graph(
    entities_table: Table,
    relationships_table: Table,
) -> dict[str, list[dict[str, Any]]]:
    degree_map = await _build_degree_map(relationships_table)
    entity_samples = await finalize_entities(entities_table, degree_map)
    relationship_samples = await finalize_relationships(relationships_table, degree_map)
    return {
        "entities": entity_samples,
        "relationships": relationship_samples,
    }

async def _build_degree_map(
    relationships_table: Table,
) -> dict[str, int]:
    seen: set[tuple[str, str]] = set()
    degree: Counter[str] = Counter()
    async for row in relationships_table:
        lo, hi = sorted((row["source"], row["target"]))
        if (lo, hi) not in seen:
            seen.add((lo, hi))
            degree[lo] += 1
            degree[hi] += 1
    return dict(degree)

async def finalize_entities(
    entities_table: Table,
    degree_map: dict[str, int],
) -> list[dict[str, Any]]:
    sample_rows: list[dict[str, Any]] = []
    seen_titles: set[str] = set()
    human_readable_id = 0
    async for row in entities_table:
        title = row.get("title")
        if not title or title in seen_titles:
            continue
        seen_titles.add(title)
        row["degree"] = degree_map.get(title, 0)
        row["human_readable_id"] = human_readable_id
        row["id"] = str(uuid4())
        human_readable_id += 1
        out = {col: row.get(col) for col in ENTITIES_FINAL_COLUMNS}
        await entities_table.write(out)
        if len(sample_rows) < 5:
            sample_rows.append(out)

    return sample_rows

async def finalize_relationships(
    relationships_table: Table,
    degree_map: dict[str, int],
) -> list[dict[str, Any]]:
    sample_rows: list[dict[str, Any]] = []
    seen: set[tuple[str, str]] = set()
    human_readable_id = 0
    async for row in relationships_table:
        key = (row.get("source", ""), row.get("target", ""))
        if key in seen:
            continue
        seen.add(key)
        row["combined_degree"] = degree_map.get(key[0], 0) + degree_map.get(key[1], 0)
        row["human_readable_id"] = human_readable_id
        row["id"] = str(uuid4())
        human_readable_id += 1
        final = {col: row.get(col) for col in RELATIONSHIPS_FINAL_COLUMNS}
        await relationships_table.write(final)
        if len(sample_rows) < 5:
            sample_rows.append(final)

    return sample_rows

3. 查看最终的实体关系图

#[GraphRAG实现原理-09]调用LLM提取实体关系图()提供了演示程序完成了初始实体关系图的构建，现在我们在此基础上应用finalize_graph这个Workflow完成针对实体度数和关系组合度数的计算和写入。我们通过如下的程序执行一个由finalize_graph这个单一Workflow构建的管道。在完成管道执行后，我们将./output/entities.parquet和./output/relationships.parquet加载到一个DataFrame中，并将两个表的结构和数据输出来。从输出可以看出：实体度数通过degree字段附加到每个实体记录，关系组合度数通过combined_degree字段附加到每一条关系记录上。

import asyncio
import graphrag.index.workflows.finalize_graph as  finalize_graph
from graphrag.index.typing.pipeline import Pipeline
from graphrag.config.load_config import load_config
from graphrag.index.run.run_pipeline import run_pipeline
from graphrag.callbacks.noop_workflow_callbacks import NoopWorkflowCallbacks
import pandas as pd

async def main():   
    pipeline = Pipeline(workflows=[
        ("finalize_graph", finalize_graph.run_workflow)
    ])
    async for _ in  run_pipeline(
        pipeline = pipeline,
        config = load_config(root_dir="./"),
        callbacks= NoopWorkflowCallbacks(),
    ):
        pass

    print(f"----------entities table schema----------")
    df = pd.read_parquet('./output/entities.parquet')
    print(df.info())

    print(f"----------entities----------")
    for _, row in df.iterrows():
        for col in df.columns:
            print(f"{col}: {row[col]}, ")
        print() 

    print(f"----------relationships table schema----------")
    df = pd.read_parquet('./output/relationships.parquet')
    print(df.info())

    print(f"----------relationships----------")
    for _, row in df.iterrows():
        for col in df.columns:
            print(f"{col}: {row[col]}, ")
        print() 
asyncio.run(main())

entities表的结构：

<class 'pandas.DataFrame'>
RangeIndex: 34 entries, 0 to 33
Data columns (total 8 columns):
 #   Column             Non-Null Count  Dtype 
---  ------             --------------  ----- 
 0   id                 34 non-null     str   
 1   human_readable_id  34 non-null     int64 
 2   title              34 non-null     str   
 3   type               34 non-null     str   
 4   description        34 non-null     str   
 5   text_unit_ids      34 non-null     object
 6   frequency          34 non-null     int64 
 7   degree             34 non-null     int64 
dtypes: int64(3), object(1), str(4)
memory usage: 7.7+ KB
None

entities表的数据：

id: 9c867d16-5825-44a6-bbb1-c6d61d888a48, 
human_readable_id: 0, 
title: 李明, 
type: PERSON, 
description: 李明是一位在杭州经营咖啡店的个体经营者。他在2024年3月决定开设一家咖啡店，并在考察多个选址后，最终选择了一个写字楼商圈的商业区域，签下了为期两年的租约。开业后，李明将门店的核心产品定位为拿铁和美式咖啡，同时推出了季节限定的桂花拿铁，以丰富产品线并提升门店吸引力。

在门店运营方面，李明在开业初期通过小红书和抖音进行引流，借助社交平台扩大曝光和吸引顾客。他还建立了会员体系，以增强顾客复购和长期黏性，并与周边商家开展合作互推，借助社区联动拓展客源。为保障供应链稳定，李明后来又增加了一家供应商作为备选，以降低单一供应风险，提升门店运营的稳定性。, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'
 'a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 4, 
degree: 17, 

id: 3a823ac8-7f73-4bf2-9ff3-642a0fe1212e, 
human_readable_id: 1, 
title: 杭州, 
type: GEO, 
description: Hangzhou is the city where 李明 planned to open a coffee shop, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 
frequency: 1, 
degree: 1, 

id: 01a9d251-99e8-41fc-9750-7572448e35c1, 
human_readable_id: 2, 
title: 学校附近, 
type: GEO, 
description: The area near a school, evaluated by 李明 as a potential coffee shop location with high foot traffic but high rent, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 
frequency: 1, 
degree: 1, 

id: 53634c26-e8c8-48f4-9998-0b9fd7488ba1, 
human_readable_id: 3, 
title: 写字楼商圈, 
type: GEO, 
description: An office-district commercial area evaluated by 李明 as a potential coffee shop location with stable weekday traffic and quiet weekends; it was ultimately chosen for the coffee shop, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 
frequency: 1, 
degree: 2, 

id: 51225d37-3298-4650-944a-bae033f38527, 
human_readable_id: 4, 
title: 老城区街角, 
type: GEO, 
description: A street corner in the old城区 evaluated by 李明 as a potential coffee shop location with low rent but fewer young customers, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 
frequency: 1, 
degree: 1, 

id: 9c277686-5f60-4ee0-b55b-8fae694b5c82, 
human_readable_id: 5, 
title: 咖啡店, 
type: EVENT, 
description: The planned opening of a coffee shop by 李明 in Hangzhou in March 2024, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 
frequency: 1, 
degree: 2, 

id: 4d104a70-3605-4c87-81ef-e1300f3412e6, 
human_readable_id: 6, 
title: 王芳, 
type: PERSON, 
description: 王芳 is the store manager who suggested adding member discounts to improve repurchase rate, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 2, 

id: 958ed09e-c043-475a-a72d-f9bac6e7c4f5, 
human_readable_id: 7, 
title: 拿铁, 
type: EVENT, 
description: Latte is one of the core products chosen by 李明, priced at 28 yuan and described as suitable for大众口味, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 2, 

id: 4377333d-b67a-4765-aaab-29df431a9e9a, 
human_readable_id: 8, 
title: 美式, 
type: EVENT, 
description: Americano is one of the core products chosen by 李明, priced at 22 yuan and aimed at coffee lovers, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 2, 

id: 28492532-d148-49de-b33b-e6e52b39831c, 
human_readable_id: 9, 
title: 桂花拿铁, 
type: EVENT, 
description: Osmanthus latte is a seasonal limited product launched by 李明 in May 2024, with first-month sales accounting for 30% of the total, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 3, 

id: f4b15245-448c-4b7f-87ef-080a988ac0c3, 
human_readable_id: 10, 
title: 会员折扣, 
type: EVENT, 
description: A proposed member discount promotion suggested by 王芳 to improve repurchase rate, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 2, 

id: d9c520c1-7f41-4ccc-a527-96505d81cc0d, 
human_readable_id: 11, 
title: 复购率, 
type: EVENT, 
description: The repurchase rate metric that 王芳 wanted to improve through member discounts, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 1, 

id: 5810ddc3-916b-4c3c-b9f9-c286ef68ebca, 
human_readable_id: 12, 
title: 季节限定款, 
type: EVENT, 
description: A seasonal limited product launch category associated with 桂花拿铁 in 2024年5月, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 1, 

id: 17d517b8-1208-4f17-8837-517aa2328df7, 
human_readable_id: 13, 
title: 首月销量, 
type: EVENT, 
description: The first-month sales performance metric for 桂花拿铁, which accounted for 30% of total sales, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 2, 

id: 0b896a65-d80b-4abc-ac5c-9c1f58e4a6cc, 
human_readable_id: 14, 
title: 总额, 
type: EVENT, 
description: The total sales amount used as the basis for calculating 桂花拿铁's first-month share, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 1, 

id: 26b2de16-aca9-4dd9-bc1e-1a7e37be3f8b, 
human_readable_id: 15, 
title: 大众口味, 
type: EVENT, 
description: A broad consumer taste segment that 拿铁 is intended to suit, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 1, 

id: f951e814-7a29-4461-bef0-4dc2bdb07211, 
human_readable_id: 16, 
title: 咖啡爱好者, 
type: EVENT, 
description: A target customer group that 美式 is aimed at, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 
frequency: 1, 
degree: 1, 

id: bb1ec95e-8fa8-4828-b79c-ff063e7e1bbb, 
human_readable_id: 17, 
title: 小红书, 
type: ORGANIZATION, 
description: 小红书是李明用于开业初期引流的社交平台之一, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 
frequency: 1, 
degree: 1, 

id: 98412328-9620-4467-8b4a-334e8edfe300, 
human_readable_id: 18, 
title: 抖音, 
type: ORGANIZATION, 
description: 抖音是李明用于开业初期引流的社交平台之一, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 
frequency: 2, 
degree: 1, 

id: 25744a41-96ae-49aa-a1fe-26192cc27dc4, 
human_readable_id: 19, 
title: 周边书店, 
type: ORGANIZATION, 
description: 周边书店是与李明开展合作互推的本地商家之一, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 
frequency: 2, 
degree: 1, 

id: 76353321-1578-4788-98ce-de7b7c141b89, 
human_readable_id: 20, 
title: 健身房, 
type: ORGANIZATION, 
description: 健身房是与李明开展合作互推的本地商家之一, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 
frequency: 2, 
degree: 1, 

id: 16bfc6a9-195a-4b15-97b2-b112fc72d1a5, 
human_readable_id: 21, 
title: 会员体系, 
type: EVENT, 
description: 会员体系是李明于2024年6月建立的营销和客户留存活动，通过充值赠送优惠来提升会员数量, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 
frequency: 2, 
degree: 2, 

id: 09994328-862a-48dc-b852-4e72ac39603e, 
human_readable_id: 22, 
title: 开业初期, 
type: EVENT, 
description: 开业初期是李明进行引流并在首周新增粉丝五千人的时间阶段, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 
frequency: 2, 
degree: 1, 

id: 1b3f34e9-366d-40f8-b608-a1e25cf129e0, 
human_readable_id: 23, 
title: 2024年6月, 
type: EVENT, 
description: 2024年6月是李明建立会员体系的时间节点, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 
frequency: 2, 
degree: 1, 

id: c84d0f9c-5ac7-4e90-b7a1-a6fc947f2f75, 
human_readable_id: 24, 
title: 张伟, 
type: PERSON, 
description: 张伟 is a supplier in Pu'er, Yunnan, who supplies coffee beans once a month at a unit price of 120 yuan per kilogram), 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 2, 

id: 6fa972a6-e769-4c63-8cd7-0ea32ba41001, 
human_readable_id: 25, 
title: 云南普洱, 
type: GEO, 
description: Yunnan Pu'er is the geographic location of the coffee-bean supplier Zhang Wei), 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 1, 

id: f446b208-63cc-4a1f-b4da-9a6e75bf37fd, 
human_readable_id: 26, 
title: 本地乳企, 
type: ORGANIZATION, 
description: A local dairy enterprise that delivers milk to the store every day), 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 2, 

id: 9c47f7bd-2313-430f-a848-540699a7e06a, 
human_readable_id: 27, 
title: 第二家供应商, 
type: ORGANIZATION, 
description: An additional backup supplier added by Li Ming to reduce supply disruption risk), 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 1, 

id: fdaf7e37-c7bd-4fef-a61f-ac827ca1178b, 
human_readable_id: 28, 
title: 2024年7月, 
type: EVENT, 
description: A time period in July 2024 when rainy-season transportation delays caused the store to run out of stock for two days and lose about 4,000 yuan), 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 2, 

id: 0b9fc0a4-29b1-4940-9e3b-d88ef0963c33, 
human_readable_id: 29, 
title: 牛奶, 
type: ORGANIZATION, 
description: Milk is a supplied product delivered daily by a local dairy enterprise, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 1, 

id: 03add403-9061-4bb1-8e6a-0b1a04b1233d, 
human_readable_id: 30, 
title: 门店, 
type: ORGANIZATION, 
description: The store experienced a two-day stockout and financial loss due to rainy-season transportation delays, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 2, 

id: ada56761-6a06-42c4-9275-137d01dafa33, 
human_readable_id: 31, 
title: 雨季运输延误, 
type: EVENT, 
description: Rainy-season transportation delays in July 2024 caused the store to run out of stock for two days and lose about 4,000 yuan, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 2, 

id: a4c8eba2-8b33-4c1e-bb59-5ded8c2a76bc, 
human_readable_id: 32, 
title: 断货两天, 
type: EVENT, 
description: A two-day stockout event at the store caused by transportation delays during the rainy season, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 2, 

id: 26e053a6-5d80-489d-bc0b-0357cf2ad3f1, 
human_readable_id: 33, 
title: 损失约四千元, 
type: EVENT, 
description: A financial-loss event in which the store lost about 4,000 yuan due to the stockout), 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 
frequency: 1, 
degree: 1,

relationships表的结构：

<class 'pandas.DataFrame'>
RangeIndex: 33 entries, 0 to 32
Data columns (total 8 columns):
 #   Column             Non-Null Count  Dtype  
---  ------             --------------  -----  
 0   id                 33 non-null     str    
 1   human_readable_id  33 non-null     int64  
 2   source             33 non-null     str    
 3   target             33 non-null     str    
 4   description        33 non-null     str    
 5   weight             33 non-null     float64
 6   combined_degree    33 non-null     int64  
 7   text_unit_ids      33 non-null     object 
dtypes: float64(1), int64(2), object(1), str(4)
memory usage: 6.1+ KB
None

relationships表记录：

id: 735bc825-d971-47cf-90a4-f0fa0c3ef813, 
human_readable_id: 0, 
source: 李明, 
target: 杭州, 
description: 李明 decided to open a coffee shop in Hangzhou, 
weight: 9.0, 
combined_degree: 18, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 

id: 1c63f47b-b87b-41a8-b466-329409c61681, 
human_readable_id: 1, 
source: 李明, 
target: 咖啡店, 
description: 李明 planned and initiated the opening of the coffee shop, 
weight: 10.0, 
combined_degree: 19, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 

id: a040ce72-ed4c-48d0-a107-a0b20bc03bbe, 
human_readable_id: 2, 
source: 李明, 
target: 学校附近, 
description: 李明 evaluated the area near a school as one possible location for the coffee shop, 
weight: 8.0, 
combined_degree: 18, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 

id: a5d33262-6b05-4c5d-84a2-68d8e2a8d998, 
human_readable_id: 3, 
source: 李明, 
target: 写字楼商圈, 
description: 李明 evaluated the office-district commercial area and ultimately selected it for the coffee shop, 
weight: 10.0, 
combined_degree: 19, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 

id: 5d116773-3dd5-48d5-bb0a-33b75e5999f4, 
human_readable_id: 4, 
source: 李明, 
target: 老城区街角, 
description: 李明 evaluated the old-town street corner as one possible location for the coffee shop, 
weight: 8.0, 
combined_degree: 18, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 

id: 0320c25e-6265-48ad-80cb-f789d930c9e6, 
human_readable_id: 5, 
source: 咖啡店, 
target: 写字楼商圈, 
description: The coffee shop was ultimately planned to be located in the office-district commercial area, 
weight: 9.0, 
combined_degree: 4, 
text_unit_ids: ['301f5f4fbb4ea9f27ad8305acf10c48413f6e090bb00cdc9db5222a0a87373aaf005d1aeb3b918324e1849e02937990c8262b268b1fadafab4d960f34cd2a707'], 

id: 0ed642d8-fb07-4d1d-9db8-243aa9c387ee, 
human_readable_id: 6, 
source: 李明, 
target: 拿铁, 
description: 李明 designated 拿铁 as one of the core products, 
weight: 9.0, 
combined_degree: 19, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: e1515b8e-f9bd-4f8a-87b9-ec8cb7175e34, 
human_readable_id: 7, 
source: 李明, 
target: 美式, 
description: 李明 designated 美式 as one of the core products, 
weight: 9.0, 
combined_degree: 19, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: 40ca6bc6-6c05-4d6c-9640-5839d02ef30a, 
human_readable_id: 8, 
source: 李明, 
target: 桂花拿铁, 
description: 李明于2024年5月推出了季节限定的桂花拿铁，并在同一时间发布了桂花拿铁这一产品。该产品由李明主导上线，属于2024年5月推出的限定饮品，体现了其在该时期对季节性新品的开发与推广。, 
weight: 19.0, 
combined_degree: 20, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'
 'a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: baff7cc1-b90a-4f84-9d73-827220fdf74c, 
human_readable_id: 9, 
source: 王芳, 
target: 李明, 
description: 王芳 is the store manager giving a suggestion in the same business context as 李明's product strategy, 
weight: 1.0, 
combined_degree: 19, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: 906e6321-cc8c-49ba-8d21-da897af7c358, 
human_readable_id: 10, 
source: 拿铁, 
target: 大众口味, 
description: 拿铁 is described as suitable for 大众口味, 
weight: 8.0, 
combined_degree: 3, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: 2328ef49-dc98-43b3-9e71-2626a472feee, 
human_readable_id: 11, 
source: 美式, 
target: 咖啡爱好者, 
description: 美式 is described as aimed at 咖啡爱好者, 
weight: 8.0, 
combined_degree: 3, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: 94758fba-7bb9-4e86-baf2-6f53122cc1b4, 
human_readable_id: 12, 
source: 桂花拿铁, 
target: 季节限定款, 
description: 桂花拿铁 is a seasonal limited product, 
weight: 9.0, 
combined_degree: 4, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: f1b50cf1-e121-436e-a911-2598e8aaf01e, 
human_readable_id: 13, 
source: 桂花拿铁, 
target: 首月销量, 
description: 桂花拿铁 had first-month sales performance described by 首月销量, 
weight: 7.0, 
combined_degree: 5, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: f3fedcca-40ec-4ddf-a710-d9591065fd2e, 
human_readable_id: 14, 
source: 首月销量, 
target: 总额, 
description: 首月销量 is measured as a share of 总额, 
weight: 7.0, 
combined_degree: 3, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: bf1f251b-4ece-4872-abd2-5a1fcc79789d, 
human_readable_id: 15, 
source: 王芳, 
target: 会员折扣, 
description: 王芳 suggested adding 会员折扣, 
weight: 10.0, 
combined_degree: 4, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: 50faa652-88e2-45fa-8b8e-6205baf8ba6b, 
human_readable_id: 16, 
source: 会员折扣, 
target: 复购率, 
description: 会员折扣 was proposed to improve 复购率, 
weight: 1.0, 
combined_degree: 3, 
text_unit_ids: ['a63ee3ab5c7fb0edf8a8a5beeafc567d2e1e82e6b4d7d3db04b8f07a27c89954e7e18e8625471d029037d34a21d7148d822ca536dc39d4035ffab3421751acd4'], 

id: 947a9e5e-0df8-488d-b82c-de1e6a9900ee, 
human_readable_id: 17, 
source: 李明, 
target: 小红书, 
description: 李明通过小红书进行引流获客, 
weight: 18.0, 
combined_degree: 18, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 

id: a04a2cbd-a12c-4333-905d-7e6bafbdf9f1, 
human_readable_id: 18, 
source: 李明, 
target: 抖音, 
description: 李明通过抖音进行引流获客, 
weight: 18.0, 
combined_degree: 18, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 

id: 94894497-be39-4a72-8877-25f43e75056e, 
human_readable_id: 19, 
source: 李明, 
target: 周边书店, 
description: 李明与周边书店合作互推以降低获客成本, 
weight: 16.0, 
combined_degree: 18, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 

id: 9c814eb9-e70a-4a1e-8b9e-b7ff646429ee, 
human_readable_id: 20, 
source: 李明, 
target: 健身房, 
description: 李明与健身房合作互推以降低获客成本, 
weight: 16.0, 
combined_degree: 18, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 

id: fb05614d-4b48-4a1a-a403-df46f0aa645f, 
human_readable_id: 21, 
source: 李明, 
target: 会员体系, 
description: 李明在2024年6月建立了会员体系, 
weight: 18.0, 
combined_degree: 19, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 

id: 2cb41edf-b40e-4828-85c5-b7f665e85fbf, 
human_readable_id: 22, 
source: 会员体系, 
target: 2024年6月, 
description: 会员体系是在2024年6月建立的, 
weight: 20.0, 
combined_degree: 3, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 

id: 2a12d950-302a-4b9e-bf67-60276158eb58, 
human_readable_id: 23, 
source: 开业初期, 
target: 李明, 
description: 李明在开业初期通过社交平台引流并快速增加粉丝, 
weight: 8.0, 
combined_degree: 18, 
text_unit_ids: ['d1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'
 'd1da6221f377cc7af3f4622e1716480446bc837cf55742e9f5010ad4af342490837cf8d4fed3a1980fded46a75c6431b389a0097859c5884392d691ddb362fc2'], 

id: 38737c40-bcf2-44fb-b2d5-627bf38cf731, 
human_readable_id: 24, 
source: 张伟, 
target: 云南普洱, 
description: Zhang Wei is described as a supplier from Yunnan Pu'er, 
weight: 8.0, 
combined_degree: 3, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 

id: b935f8a7-bdcf-41a0-bff9-f6b6fce933d8, 
human_readable_id: 25, 
source: 李明, 
target: 第二家供应商, 
description: Li Ming added the second supplier as a backup, 
weight: 9.0, 
combined_degree: 18, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 

id: 9a3324bb-bd27-4fc8-a1d8-4f23ad0a090a, 
human_readable_id: 26, 
source: 本地乳企, 
target: 2024年7月, 
description: The local dairy enterprise is part of the store's supply situation affected by the July 2024 disruption context, 
weight: 3.0, 
combined_degree: 4, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 

id: 4cd68752-5b32-419e-b39f-8eeef5c25f3c, 
human_readable_id: 27, 
source: 张伟, 
target: 2024年7月, 
description: The coffee-bean supply from Zhang Wei was affected during the July 2024 rainy-season delay period, 
weight: 1.0, 
combined_degree: 4, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 

id: 6a8f36e3-20bd-460b-8818-19886205285b, 
human_readable_id: 28, 
source: 本地乳企, 
target: 牛奶, 
description: The local dairy enterprise delivers milk every day, 
weight: 10.0, 
combined_degree: 3, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 

id: a42c8b99-790b-4984-b246-8aca78e79926, 
human_readable_id: 29, 
source: 门店, 
target: 雨季运输延误, 
description: The store was affected by rainy-season transportation delays, 
weight: 9.0, 
combined_degree: 4, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 

id: 48985ff4-52a6-4c6a-9566-451b77c31dc5, 
human_readable_id: 30, 
source: 雨季运输延误, 
target: 断货两天, 
description: The transportation delay caused the two-day stockout, 
weight: 10.0, 
combined_degree: 4, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 

id: 3ab20504-5025-4455-9190-353e94c893e8, 
human_readable_id: 31, 
source: 断货两天, 
target: 损失约四千元, 
description: The two-day stockout resulted in a loss of about 4,000 yuan, 
weight: 9.0, 
combined_degree: 3, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'], 

id: 3dd5c112-2553-4a6c-a646-214229a8e730, 
human_readable_id: 32, 
source: 李明, 
target: 门店, 
description: Li Ming took action to improve the store's supply resilience by adding a backup supplier, 
weight: 1.0, 
combined_degree: 19, 
text_unit_ids: ['f2d54bee6965bb80537dbc03402d24b17195cd53b9bdb3f33f98581738eab7044b8e7f9c7a55ff91e485a2850f2526ad1fb907d9078880335bf0a5eec477457c'],

4. 以可视化形成查看实体关系图

我们在上面提到过，如果在执行finalize_graph这个Workflow时按照如下的方式将RagConfig的snapshots.graphml字段设置为True，它会将当前这个图以GraphML文件的形式存下来。

import asyncio
import graphrag.index.workflows.finalize_graph as  finalize_graph
from graphrag.index.typing.pipeline import Pipeline
from graphrag.config.load_config import load_config
from graphrag.index.run.run_pipeline import run_pipeline
from graphrag.callbacks.noop_workflow_callbacks import NoopWorkflowCallbacks
import pandas as pd

async def main():   
    pipeline = Pipeline(workflows=[
        ("finalize_graph", finalize_graph.run_workflow)
    ])
    config = load_config(root_dir="./")
    config.snapshots.graphml = True
    async for _ in  run_pipeline(
        pipeline = pipeline,
        config=config,
        callbacks= NoopWorkflowCallbacks(),
    ):
        pass
asyncio.run(main())