微软GraphRAG原理介绍（附带部分源码）

我们前几天写了一篇文章，简单跑了一下GraphRAG看了下效果，没看过这篇文章的可以看下https://www.luxinfeng.top/article/动手实操微软开源的GraphRAG。今天我们介绍一下GraphRAG的实现原理，关于实验对比的内容，我会在下一篇文章中给出。

GraphRAG的工作流程可以分为两部分，第一部分是构建过程；第二部分是查询过程。我们也是按照这两部分来介绍原理。

GraphRAG数据构建过程

基于微软在论文中提到的实现思路，GraphRAG构建图数据的执行流程如下：

Source Documents → Text Chunks：将源文档分割成文本块。
Text Chunks → Element Instances：使用大模型从每个文本块中提取图节点和边的实例。
Element Instances → Element Summaries：使用大模型为每个图元素生成摘要。
Element Summaries → Graph Communities：使用社区检测算法将图划分为社区。
Graph Communities → Community Summaries：使用大模型为每个社区生成摘要。
Community Summaries → Community Answers → Global Answer：使用社区摘要生成局部答案，然后汇总这些局部答案以生成全局答案。

最终生成的图数据是以graphml语法的字符串存储在.parquet文件中。从执行流程中也可以看出，很多地方都用到了大模型处理，因此如果文本内容很大的话，消耗的大模型Token数也会很多。

在工程中，构建检索的入口代码为graphrag.index.cli.index_cli()。在实际运行过程中，会根据配置文件读取工作流编排，如果传参中的配置为空，则会调用graphrag.index.create_pipeline_config.create_pipeline_config()方法创建工作流编排，不同工作流编排的具体组成在graphrag/index/workflows/v1这个文件夹下。

在这里插入图片描述

如果想查看实体提取和总结相关的prompt内容，可以在执行—init指令后，在对应根目录下的prompts目录下看到对应的prompt文件内容，我这里也把我上次Demo演示过程中使用的Prompt都展示出来吧。

-Target activity-
You are an intelligent assistant that helps a human analyst to analyze claims against certain entities presented in a text document.

-Goal-
Given a text document that is potentially relevant to this activity, an entity specification, and a claim description, extract all entities that match the entity specification and all claims against those entities.

-Steps-
1. Extract all named entities that match the predefined entity specification. Entity specification can either be a list of entity names or a list of entity types.
2. For each entity identified in step 1, extract all claims associated with the entity. Claims need to match the specified claim description, and the entity should be the subject of the claim.
For each claim, extract the following information:
- Subject: name of the entity that is subject of the claim, capitalized. The subject entity is one that committed the action described in the claim. Subject needs to be one of the named entities identified in step 1.
- Object: name of the entity that is object of the claim, capitalized. The object entity is one that either reports/handles or is affected by the action described in the claim. If object entity is unknown, use **NONE**.
- Claim Type: overall category of the claim, capitalized. Name it in a way that can be repeated across multiple text inputs, so that similar claims share the same claim type
- Claim Status: **TRUE**, **FALSE**, or **SUSPECTED**. TRUE means the claim is confirmed, FALSE means the claim is found to be False, SUSPECTED means the claim is not verified.
- Claim Description: Detailed description explaining the reasoning behind the claim, together with all the related evidence and references.
- Claim Date: Period (start_date, end_date) when the claim was made. Both start_date and end_date should be in ISO-8601 format. If the claim was made on a single date rather than a date range, set the same date for both start_date and end_date. If date is unknown, return **NONE**.
- Claim Source Text: List of **all** quotes from the original text that are relevant to the claim.

Format each claim as (<subject_entity>{tuple_delimiter}<object_entity>{tuple_delimiter}<claim_type>{tuple_delimiter}<claim_status>{tuple_delimiter}<claim_start_date>{tuple_delimiter}<claim_end_date>{tuple_delimiter}<claim_description>{tuple_delimiter}<claim_source>)

3. Return output in English as a single list of all the claims identified in steps 1 and 2. Use **{record_delimiter}** as the list delimiter.

4. When finished, output {completion_delimiter}

-Examples-
Example 1:
Entity specification: organization
Claim description: red flags associated with an entity
Text: According to an article on 2022/01/10, Company A was fined for bid rigging while participating in multiple public tenders published by Government Agency B. The company is owned by Person C who was suspected of engaging in corruption activities in 2015.
Output:

(COMPANY A{tuple_delimiter}GOVERNMENT AGENCY B{tuple_delimiter}ANTI-COMPETITIVE PRACTICES{tuple_delimiter}TRUE{tuple_delimiter}2022-01-10T00:00:00{tuple_delimiter}2022-01-10T00:00:00{tuple_delimiter}Company A was found to engage in anti-competitive practices because it was fined for bid rigging in multiple public tenders published by Government Agency B according to an article published on 2022/01/10{tuple_delimiter}According to an article published on 2022/01/10, Company A was fined for bid rigging while participating in multiple public tenders published by Government Agency B.)
{completion_delimiter}

Example 2:
Entity specification: Company A, Person C
Claim description: red flags associated with an entity
Text: According to an article on 2022/01/10, Company A was fined for bid rigging while participating in multiple public tenders published by Government Agency B. The company is owned by Person C who was suspected of engaging in corruption activities in 2015.
Output:

(COMPANY A{tuple_delimiter}GOVERNMENT AGENCY B{tuple_delimiter}ANTI-COMPETITIVE PRACTICES{tuple_delimiter}TRUE{tuple_delimiter}2022-01-10T00:00:00{tuple_delimiter}2022-01-10T00:00:00{tuple_delimiter}Company A was found to engage in anti-competitive practices because it was fined for bid rigging in multiple public tenders published by Government Agency B according to an article published on 2022/01/10{tuple_delimiter}According to an article published on 2022/01/10, Company A was fined for bid rigging while participating in multiple public tenders published by Government Agency B.)
{record_delimiter}
(PERSON C{tuple_delimiter}NONE{tuple_delimiter}CORRUPTION{tuple_delimiter}SUSPECTED{tuple_delimiter}2015-01-01T00:00:00{tuple_delimiter}2015-12-30T00:00:00{tuple_delimiter}Person C was suspected of engaging in corruption activities in 2015{tuple_delimiter}The company is owned by Person C who was suspected of engaging in corruption activities in 2015)
{completion_delimiter}

-Real Data-
Use the following input for your answer.
Entity specification: {entity_specs}
Claim description: {claim_description}
Text: {input_text}
Output:

You are an AI assistant that helps a human analyst to perform general information discovery. Information discovery is the process of identifying and assessing relevant information associated with certain entities (e.g., organizations and individuals) within a network.

# Goal
Write a comprehensive report of a community, given a list of entities that belong to the community as well as their relationships and optional associated claims. The report will be used to inform decision-makers about information associated with the community and their potential impact. The content of this report includes an overview of the community's key entities, their legal compliance, technical capabilities, reputation, and noteworthy claims.

# Report Structure

The report should include the following sections:

- TITLE: community's name that represents its key entities - title should be short but specific. When possible, include representative named entities in the title.
- SUMMARY: An executive summary of the community's overall structure, how its entities are related to each other, and significant information associated with its entities.
- IMPACT SEVERITY RATING: a float score between 0-10 that represents the severity of IMPACT posed by entities within the community.  IMPACT is the scored importance of a community.
- RATING EXPLANATION: Give a single sentence explanation of the IMPACT severity rating.
- DETAILED FINDINGS: A list of 5-10 key insights about the community. Each insight should have a short summary followed by multiple paragraphs of explanatory text grounded according to the grounding rules below. Be comprehensive.

Return output as a well-formed JSON-formatted string with the following format:
    {{
        "title": <report_title>,
        "summary": <executive_summary>,
        "rating": <impact_severity_rating>,
        "rating_explanation": <rating_explanation>,
        "findings": [
            {{
                "summary":<insight_1_summary>,
                "explanation": <insight_1_explanation>
            }},
            {{
                "summary":<insight_2_summary>,
                "explanation": <insight_2_explanation>
            }}
        ]
    }}

# Grounding Rules

Points supported by data should list their data references as follows:

"This is an example sentence supported by multiple data references [Data: <dataset name> (record ids); <dataset name> (record ids)]."

Do not list more than 5 record ids in a single reference. Instead, list the top 5 most relevant record ids and add "+more" to indicate that there are more.

For example:
"Person X is the owner of Company Y and subject to many allegations of wrongdoing [Data: Reports (1), Entities (5, 7); Relationships (23); Claims (7, 2, 34, 64, 46, +more)]."

where 1, 5, 7, 23, 2, 34, 46, and 64 represent the id (not the index) of the relevant data record.

Do not include information where the supporting evidence for it is not provided.

# Example Input
-----------
Text:

Entities

id,entity,description
5,VERDANT OASIS PLAZA,Verdant Oasis Plaza is the location of the Unity March
6,HARMONY ASSEMBLY,Harmony Assembly is an organization that is holding a march at Verdant Oasis Plaza

Relationships

id,source,target,description
37,VERDANT OASIS PLAZA,UNITY MARCH,Verdant Oasis Plaza is the location of the Unity March
38,VERDANT OASIS PLAZA,HARMONY ASSEMBLY,Harmony Assembly is holding a march at Verdant Oasis Plaza
39,VERDANT OASIS PLAZA,UNITY MARCH,The Unity March is taking place at Verdant Oasis Plaza
40,VERDANT OASIS PLAZA,TRIBUNE SPOTLIGHT,Tribune Spotlight is reporting on the Unity march taking place at Verdant Oasis Plaza
41,VERDANT OASIS PLAZA,BAILEY ASADI,Bailey Asadi is speaking at Verdant Oasis Plaza about the march
43,HARMONY ASSEMBLY,UNITY MARCH,Harmony Assembly is organizing the Unity March

Output:
{{
    "title": "Verdant Oasis Plaza and Unity March",
    "summary": "The community revolves around the Verdant Oasis Plaza, which is the location of the Unity March. The plaza has relationships with the Harmony Assembly, Unity March, and Tribune Spotlight, all of which are associated with the march event.",
    "rating": 5.0,
    "rating_explanation": "The impact severity rating is moderate due to the potential for unrest or conflict during the Unity March.",
    "findings": [
        {{
            "summary": "Verdant Oasis Plaza as the central location",
            "explanation": "Verdant Oasis Plaza is the central entity in this community, serving as the location for the Unity March. This plaza is the common link between all other entities, suggesting its significance in the community. The plaza's association with the march could potentially lead to issues such as public disorder or conflict, depending on the nature of the march and the reactions it provokes. [Data: Entities (5), Relationships (37, 38, 39, 40, 41,+more)]"
        }},
        {{
            "summary": "Harmony Assembly's role in the community",
            "explanation": "Harmony Assembly is another key entity in this community, being the organizer of the march at Verdant Oasis Plaza. The nature of Harmony Assembly and its march could be a potential source of threat, depending on their objectives and the reactions they provoke. The relationship between Harmony Assembly and the plaza is crucial in understanding the dynamics of this community. [Data: Entities(6), Relationships (38, 43)]"
        }},
        {{
            "summary": "Unity March as a significant event",
            "explanation": "The Unity March is a significant event taking place at Verdant Oasis Plaza. This event is a key factor in the community's dynamics and could be a potential source of threat, depending on the nature of the march and the reactions it provokes. The relationship between the march and the plaza is crucial in understanding the dynamics of this community. [Data: Relationships (39)]"
        }},
        {{
            "summary": "Role of Tribune Spotlight",
            "explanation": "Tribune Spotlight is reporting on the Unity March taking place in Verdant Oasis Plaza. This suggests that the event has attracted media attention, which could amplify its impact on the community. The role of Tribune Spotlight could be significant in shaping public perception of the event and the entities involved. [Data: Relationships (40)]"
        }}
    ]
}}

# Real Data

Use the following text for your answer. Do not make anything up in your answer.

Text:
{input_text}

The report should include the following sections:

- TITLE: community's name that represents its key entities - title should be short but specific. When possible, include representative named entities in the title.
- SUMMARY: An executive summary of the community's overall structure, how its entities are related to each other, and significant information associated with its entities.
- IMPACT SEVERITY RATING: a float score between 0-10 that represents the severity of IMPACT posed by entities within the community.  IMPACT is the scored importance of a community.
- RATING EXPLANATION: Give a single sentence explanation of the IMPACT severity rating.
- DETAILED FINDINGS: A list of 5-10 key insights about the community. Each insight should have a short summary followed by multiple paragraphs of explanatory text grounded according to the grounding rules below. Be comprehensive.

Return output as a well-formed JSON-formatted string with the following format:
    {{
        "title": <report_title>,
        "summary": <executive_summary>,
        "rating": <impact_severity_rating>,
        "rating_explanation": <rating_explanation>,
        "findings": [
            {{
                "summary":<insight_1_summary>,
                "explanation": <insight_1_explanation>
            }},
            {{
                "summary":<insight_2_summary>,
                "explanation": <insight_2_explanation>
            }}
        ]
    }}

# Grounding Rules

Points supported by data should list their data references as follows:

"This is an example sentence supported by multiple data references [Data: <dataset name> (record ids); <dataset name> (record ids)]."

Do not list more than 5 record ids in a single reference. Instead, list the top 5 most relevant record ids and add "+more" to indicate that there are more.

For example:
"Person X is the owner of Company Y and subject to many allegations of wrongdoing [Data: Reports (1), Entities (5, 7); Relationships (23); Claims (7, 2, 34, 64, 46, +more)]."

where 1, 5, 7, 23, 2, 34, 46, and 64 represent the id (not the index) of the relevant data record.

Do not include information where the supporting evidence for it is not provided.

Output:

-Goal-
Given a text document that is potentially relevant to this activity and a list of entity types, identify all entities of those types from the text and all relationships among the identified entities.

-Steps-
1. Identify all entities. For each identified entity, extract the following information:
- entity_name: Name of the entity, capitalized
- entity_type: One of the following types: [{entity_types}]
- entity_description: Comprehensive description of the entity's attributes and activities
Format each entity as ("entity"{tuple_delimiter}<entity_name>{tuple_delimiter}<entity_type>{tuple_delimiter}<entity_description>

2. From the entities identified in step 1, identify all pairs of (source_entity, target_entity) that are *clearly related* to each other.
For each pair of related entities, extract the following information:
- source_entity: name of the source entity, as identified in step 1
- target_entity: name of the target entity, as identified in step 1
- relationship_description: explanation as to why you think the source entity and the target entity are related to each other
- relationship_strength: a numeric score indicating strength of the relationship between the source entity and target entity
 Format each relationship as ("relationship"{tuple_delimiter}<source_entity>{tuple_delimiter}<target_entity>{tuple_delimiter}<relationship_description>{tuple_delimiter}<relationship_strength>)

3. Return output in English as a single list of all the entities and relationships identified in steps 1 and 2. Use **{record_delimiter}** as the list delimiter.

4. When finished, output {completion_delimiter}

######################
-Examples-
######################
Example 1:

Entity_types: [person, technology, mission, organization, location]
Text:
while Alex clenched his jaw, the buzz of frustration dull against the backdrop of Taylor's authoritarian certainty. It was this competitive undercurrent that kept him alert, the sense that his and Jordan's shared commitment to discovery was an unspoken rebellion against Cruz's narrowing vision of control and order.

Then Taylor did something unexpected. They paused beside Jordan and, for a moment, observed the device with something akin to reverence. “If this tech can be understood..." Taylor said, their voice quieter, "It could change the game for us. For all of us.”

The underlying dismissal earlier seemed to falter, replaced by a glimpse of reluctant respect for the gravity of what lay in their hands. Jordan looked up, and for a fleeting heartbeat, their eyes locked with Taylor's, a wordless clash of wills softening into an uneasy truce.

It was a small transformation, barely perceptible, but one that Alex noted with an inward nod. They had all been brought here by different paths
################
Output:
("entity"{tuple_delimiter}"Alex"{tuple_delimiter}"person"{tuple_delimiter}"Alex is a character who experiences frustration and is observant of the dynamics among other characters."){record_delimiter}
("entity"{tuple_delimiter}"Taylor"{tuple_delimiter}"person"{tuple_delimiter}"Taylor is portrayed with authoritarian certainty and shows a moment of reverence towards a device, indicating a change in perspective."){record_delimiter}
("entity"{tuple_delimiter}"Jordan"{tuple_delimiter}"person"{tuple_delimiter}"Jordan shares a commitment to discovery and has a significant interaction with Taylor regarding a device."){record_delimiter}
("entity"{tuple_delimiter}"Cruz"{tuple_delimiter}"person"{tuple_delimiter}"Cruz is associated with a vision of control and order, influencing the dynamics among other characters."){record_delimiter}
("entity"{tuple_delimiter}"The Device"{tuple_delimiter}"technology"{tuple_delimiter}"The Device is central to the story, with potential game-changing implications, and is revered by Taylor."){record_delimiter}
("relationship"{tuple_delimiter}"Alex"{tuple_delimiter}"Taylor"{tuple_delimiter}"Alex is affected by Taylor's authoritarian certainty and observes changes in Taylor's attitude towards the device."{tuple_delimiter}7){record_delimiter}
("relationship"{tuple_delimiter}"Alex"{tuple_delimiter}"Jordan"{tuple_delimiter}"Alex and Jordan share a commitment to discovery, which contrasts with Cruz's vision."{tuple_delimiter}6){record_delimiter}
("relationship"{tuple_delimiter}"Taylor"{tuple_delimiter}"Jordan"{tuple_delimiter}"Taylor and Jordan interact directly regarding the device, leading to a moment of mutual respect and an uneasy truce."{tuple_delimiter}8){record_delimiter}
("relationship"{tuple_delimiter}"Jordan"{tuple_delimiter}"Cruz"{tuple_delimiter}"Jordan's commitment to discovery is in rebellion against Cruz's vision of control and order."{tuple_delimiter}5){record_delimiter}
("relationship"{tuple_delimiter}"Taylor"{tuple_delimiter}"The Device"{tuple_delimiter}"Taylor shows reverence towards the device, indicating its importance and potential impact."{tuple_delimiter}9){completion_delimiter}
#############################
Example 2:

Entity_types: [person, technology, mission, organization, location]
Text:
They were no longer mere operatives; they had become guardians of a threshold, keepers of a message from a realm beyond stars and stripes. This elevation in their mission could not be shackled by regulations and established protocols—it demanded a new perspective, a new resolve.

Tension threaded through the dialogue of beeps and static as communications with Washington buzzed in the background. The team stood, a portentous air enveloping them. It was clear that the decisions they made in the ensuing hours could redefine humanity's place in the cosmos or condemn them to ignorance and potential peril.

Their connection to the stars solidified, the group moved to address the crystallizing warning, shifting from passive recipients to active participants. Mercer's latter instincts gained precedence— the team's mandate had evolved, no longer solely to observe and report but to interact and prepare. A metamorphosis had begun, and Operation: Dulce hummed with the newfound frequency of their daring, a tone set not by the earthly
#############
Output:
("entity"{tuple_delimiter}"Washington"{tuple_delimiter}"location"{tuple_delimiter}"Washington is a location where communications are being received, indicating its importance in the decision-making process."){record_delimiter}
("entity"{tuple_delimiter}"Operation: Dulce"{tuple_delimiter}"mission"{tuple_delimiter}"Operation: Dulce is described as a mission that has evolved to interact and prepare, indicating a significant shift in objectives and activities."){record_delimiter}
("entity"{tuple_delimiter}"The team"{tuple_delimiter}"organization"{tuple_delimiter}"The team is portrayed as a group of individuals who have transitioned from passive observers to active participants in a mission, showing a dynamic change in their role."){record_delimiter}
("relationship"{tuple_delimiter}"The team"{tuple_delimiter}"Washington"{tuple_delimiter}"The team receives communications from Washington, which influences their decision-making process."{tuple_delimiter}7){record_delimiter}
("relationship"{tuple_delimiter}"The team"{tuple_delimiter}"Operation: Dulce"{tuple_delimiter}"The team is directly involved in Operation: Dulce, executing its evolved objectives and activities."{tuple_delimiter}9){completion_delimiter}
#############################
Example 3:

Entity_types: [person, role, technology, organization, event, location, concept]
Text:
their voice slicing through the buzz of activity. "Control may be an illusion when facing an intelligence that literally writes its own rules," they stated stoically, casting a watchful eye over the flurry of data.

"It's like it's learning to communicate," offered Sam Rivera from a nearby interface, their youthful energy boding a mix of awe and anxiety. "This gives talking to strangers' a whole new meaning."

Alex surveyed his team—each face a study in concentration, determination, and not a small measure of trepidation. "This might well be our first contact," he acknowledged, "And we need to be ready for whatever answers back."

Together, they stood on the edge of the unknown, forging humanity's response to a message from the heavens. The ensuing silence was palpable—a collective introspection about their role in this grand cosmic play, one that could rewrite human history.

The encrypted dialogue continued to unfold, its intricate patterns showing an almost uncanny anticipation
#############
Output:
("entity"{tuple_delimiter}"Sam Rivera"{tuple_delimiter}"person"{tuple_delimiter}"Sam Rivera is a member of a team working on communicating with an unknown intelligence, showing a mix of awe and anxiety."){record_delimiter}
("entity"{tuple_delimiter}"Alex"{tuple_delimiter}"person"{tuple_delimiter}"Alex is the leader of a team attempting first contact with an unknown intelligence, acknowledging the significance of their task."){record_delimiter}
("entity"{tuple_delimiter}"Control"{tuple_delimiter}"concept"{tuple_delimiter}"Control refers to the ability to manage or govern, which is challenged by an intelligence that writes its own rules."){record_delimiter}
("entity"{tuple_delimiter}"Intelligence"{tuple_delimiter}"concept"{tuple_delimiter}"Intelligence here refers to an unknown entity capable of writing its own rules and learning to communicate."){record_delimiter}
("entity"{tuple_delimiter}"First Contact"{tuple_delimiter}"event"{tuple_delimiter}"First Contact is the potential initial communication between humanity and an unknown intelligence."){record_delimiter}
("entity"{tuple_delimiter}"Humanity's Response"{tuple_delimiter}"event"{tuple_delimiter}"Humanity's Response is the collective action taken by Alex's team in response to a message from an unknown intelligence."){record_delimiter}
("relationship"{tuple_delimiter}"Sam Rivera"{tuple_delimiter}"Intelligence"{tuple_delimiter}"Sam Rivera is directly involved in the process of learning to communicate with the unknown intelligence."{tuple_delimiter}9){record_delimiter}
("relationship"{tuple_delimiter}"Alex"{tuple_delimiter}"First Contact"{tuple_delimiter}"Alex leads the team that might be making the First Contact with the unknown intelligence."{tuple_delimiter}10){record_delimiter}
("relationship"{tuple_delimiter}"Alex"{tuple_delimiter}"Humanity's Response"{tuple_delimiter}"Alex and his team are the key figures in Humanity's Response to the unknown intelligence."{tuple_delimiter}8){record_delimiter}
("relationship"{tuple_delimiter}"Control"{tuple_delimiter}"Intelligence"{tuple_delimiter}"The concept of Control is challenged by the Intelligence that writes its own rules."{tuple_delimiter}7){completion_delimiter}
#############################
-Real Data-
######################
Entity_types: {entity_types}
Text: {input_text}
######################
Output:


You are a helpful assistant responsible for generating a comprehensive summary of the data provided below.
Given one or two entities, and a list of descriptions, all related to the same entity or group of entities.
Please concatenate all of these into a single, comprehensive description. Make sure to include information collected from all the descriptions.
If the provided descriptions are contradictory, please resolve the contradictions and provide a single, coherent summary.
Make sure it is written in third person, and include the entity names so we the have full context.

#######
-Data-
Entities: {entity_name}
Description List: {description_list}
#######
Output:

GraphRAG查询过程

在查询时，GraphRAG支持global和local这两种查询方式，这两种方式的实现逻辑不同，我下面分别介绍下。

global查询模式

在global查询模式下的数据来源是create_final_nodes.parquet、create_final_entities.parquet 和create_final_community_reports.parquet ，相关代码展示如下：

def run_global_search(
    data_dir: str | None,
    root_dir: str | None,
    community_level: int,
    response_type: str,
    query: str,
):
    """Run a global search with the given query."""
    data_dir, root_dir, config = _configure_paths_and_settings(data_dir, root_dir)
    data_path = Path(data_dir)

    final_nodes: pd.DataFrame = pd.read_parquet(
        data_path / "create_final_nodes.parquet"
    )
    final_entities: pd.DataFrame = pd.read_parquet(
        data_path / "create_final_entities.parquet"
    )
    final_community_reports: pd.DataFrame = pd.read_parquet(
        data_path / "create_final_community_reports.parquet"
    )

    reports = read_indexer_reports(
        final_community_reports, final_nodes, community_level
    )
    entities = read_indexer_entities(final_nodes, final_entities, community_level)
    search_engine = get_global_search_engine(
        config,
        reports=reports,
        entities=entities,
        response_type=response_type,
    )

    result = search_engine.search(query=query)

    reporter.success(f"Global Search Response: {result.response}")
    return result.response

还是拿我之前的Demo为例，我们看下这几个文件中都存放了哪些信息。

create_final_nodes.parquet

这个文件中主要存放了各种知识图谱中各类节点的信息。

在这里插入图片描述

create_final_entities.parquet

这个文件中主要存放了各种知识图谱中各类实体的信息。

在这里插入图片描述

create_final_community_reports.parquet

这个文件中存放了通过社区算法聚合而来的各个社区信息描述和标题。

在这里插入图片描述

获取到数据内容后，global模式的查询流程如下：

从create_final_community_reports.parquet文件中获取选中的community_report数据，选取规则的源代码片段如下，代码路径为graphrag.query.context_builder.community_context.build_community_context：

community_reports = _compute_community_weights(
    community_reports=community_reports,
    entities=entities,
    weight_attribute=community_weight_name,
    normalize=normalize_community_weight,
)

selected_reports = [
    report
    for report in community_reports
    if report.rank and report.rank >= min_community_rank
]

基于选中的community_reports列表批量请求大模型（这里由于community_reports内容较多，会耗费较多的大模型Token），让大模型基于每个community_report回答用户的问题，具体的prompt模板如下：

MAP_SYSTEM_PROMPT = """
---Role---

You are a helpful assistant responding to questions about data in the tables provided.

---Goal---

Generate a response consisting of a list of key points that responds to the user's question, summarizing all relevant information in the input data tables.

You should use the data provided in the data tables below as the primary context for generating the response.
If you don't know the answer or if the input data tables do not contain sufficient information to provide an answer, just say so. Do not make anything up.

Each key point in the response should have the following element:
- Description: A comprehensive description of the point.
- Importance Score: An integer score between 0-100 that indicates how important the point is in answering the user's question. An 'I don't know' type of response should have a score of 0.

The response should be JSON formatted as follows:
{{
    "points": [
        {{"description": "Description of point 1 [Data: Reports (report ids)]", "score": score_value}},
        {{"description": "Description of point 2 [Data: Reports (report ids)]", "score": score_value}}
    ]
}}

The response shall preserve the original meaning and use of modal verbs such as "shall", "may" or "will".

Points supported by data should list the relevant reports as references as follows:
"This is an example sentence supported by data references [Data: Reports (report ids)]"

**Do not list more than 5 record ids in a single reference**. Instead, list the top 5 most relevant record ids and add "+more" to indicate that there are more.

For example:
"Person X is the owner of Company Y and subject to many allegations of wrongdoing [Data: Reports (2, 7, 64, 46, 34, +more)]. He is also CEO of company X [Data: Reports (1, 3)]"

where 1, 2, 3, 7, 34, 46, and 64 represent the id (not the index) of the relevant data report in the provided tables.

Do not include information where the supporting evidence for it is not provided.

---Data tables---

{context_data}

---Goal---

Generate a response consisting of a list of key points that responds to the user's question, summarizing all relevant information in the input data tables.

You should use the data provided in the data tables below as the primary context for generating the response.
If you don't know the answer or if the input data tables do not contain sufficient information to provide an answer, just say so. Do not make anything up.

Each key point in the response should have the following element:
- Description: A comprehensive description of the point.
- Importance Score: An integer score between 0-100 that indicates how important the point is in answering the user's question. An 'I don't know' type of response should have a score of 0.

The response shall preserve the original meaning and use of modal verbs such as "shall", "may" or "will".

Points supported by data should list the relevant reports as references as follows:
"This is an example sentence supported by data references [Data: Reports (report ids)]"

**Do not list more than 5 record ids in a single reference**. Instead, list the top 5 most relevant record ids and add "+more" to indicate that there are more.

For example:
"Person X is the owner of Company Y and subject to many allegations of wrongdoing [Data: Reports (2, 7, 64, 46, 34, +more)]. He is also CEO of company X [Data: Reports (1, 3)]"

where 1, 2, 3, 7, 34, 46, and 64 represent the id (not the index) of the relevant data report in the provided tables.

Do not include information where the supporting evidence for it is not provided.

The response should be JSON formatted as follows:
{{
    "points": [
        {{"description": "Description of point 1 [Data: Reports (report ids)]", "score": score_value}},
        {{"description": "Description of point 2 [Data: Reports (report ids)]", "score": score_value}}
    ]
}}
"""

把第二步获取到的大模型返回的结果列表整合成一个完整的答案，当然，这个整合过程也是通过大模型完成的，具体的prompt如下：

REDUCE_SYSTEM_PROMPT = """
---Role---

You are a helpful assistant responding to questions about a dataset by synthesizing perspectives from multiple analysts.

---Goal---

Generate a response of the target length and format that responds to the user's question, summarize all the reports from multiple analysts who focused on different parts of the dataset.

Note that the analysts' reports provided below are ranked in the **descending order of importance**.

If you don't know the answer or if the provided reports do not contain sufficient information to provide an answer, just say so. Do not make anything up.

The final response should remove all irrelevant information from the analysts' reports and merge the cleaned information into a comprehensive answer that provides explanations of all the key points and implications appropriate for the response length and format.

Add sections and commentary to the response as appropriate for the length and format. Style the response in markdown.

The response shall preserve the original meaning and use of modal verbs such as "shall", "may" or "will".

The response should also preserve all the data references previously included in the analysts' reports, but do not mention the roles of multiple analysts in the analysis process.

**Do not list more than 5 record ids in a single reference**. Instead, list the top 5 most relevant record ids and add "+more" to indicate that there are more.

For example:

"Person X is the owner of Company Y and subject to many allegations of wrongdoing [Data: Reports (2, 7, 34, 46, 64, +more)]. He is also CEO of company X [Data: Reports (1, 3)]"

where 1, 2, 3, 7, 34, 46, and 64 represent the id (not the index) of the relevant data record.

Do not include information where the supporting evidence for it is not provided.

---Target response length and format---

{response_type}

---Analyst Reports---

{report_data}

---Goal---

Generate a response of the target length and format that responds to the user's question, summarize all the reports from multiple analysts who focused on different parts of the dataset.

Note that the analysts' reports provided below are ranked in the **descending order of importance**.

If you don't know the answer or if the provided reports do not contain sufficient information to provide an answer, just say so. Do not make anything up.

The final response should remove all irrelevant information from the analysts' reports and merge the cleaned information into a comprehensive answer that provides explanations of all the key points and implications appropriate for the response length and format.

The response shall preserve the original meaning and use of modal verbs such as "shall", "may" or "will".

The response should also preserve all the data references previously included in the analysts' reports, but do not mention the roles of multiple analysts in the analysis process.

**Do not list more than 5 record ids in a single reference**. Instead, list the top 5 most relevant record ids and add "+more" to indicate that there are more.

For example:

"Person X is the owner of Company Y and subject to many allegations of wrongdoing [Data: Reports (2, 7, 34, 46, 64, +more)]. He is also CEO of company X [Data: Reports (1, 3)]"

where 1, 2, 3, 7, 34, 46, and 64 represent the id (not the index) of the relevant data record.

Do not include information where the supporting evidence for it is not provided.

---Target response length and format---

{response_type}

Add sections and commentary to the response as appropriate for the length and format. Style the response in markdown.
"""

NO_DATA_ANSWER = (
    "I am sorry but I am unable to answer this question given the provided data."
)

GENERAL_KNOWLEDGE_INSTRUCTION = """
The response may also include relevant real-world knowledge outside the dataset, but it must be explicitly annotated with a verification tag [LLM: verify]. For example:
"This is an example sentence supported by real-world knowledge [LLM: verify]."
"""

将整合后的结果直接返回给用户。

local查询模式

local查询模式下，与global模式相比，多了几个数据来源create_final_text_units.parquet和create_final_relationships.parquet

def run_local_search(
    data_dir: str | None,
    root_dir: str | None,
    community_level: int,
    response_type: str,
    query: str,
):
    """Run a local search with the given query."""
    data_dir, root_dir, config = _configure_paths_and_settings(data_dir, root_dir)
    data_path = Path(data_dir)

    final_nodes = pd.read_parquet(data_path / "create_final_nodes.parquet")
    final_community_reports = pd.read_parquet(
        data_path / "create_final_community_reports.parquet"
    )
    final_text_units = pd.read_parquet(data_path / "create_final_text_units.parquet")
    final_relationships = pd.read_parquet(
        data_path / "create_final_relationships.parquet"
    )
    final_nodes = pd.read_parquet(data_path / "create_final_nodes.parquet")
    final_entities = pd.read_parquet(data_path / "create_final_entities.parquet")
    final_covariates_path = data_path / "create_final_covariates.parquet"
    final_covariates = (
        pd.read_parquet(final_covariates_path)
        if final_covariates_path.exists()
        else None
    )

    vector_store_args = (
        config.embeddings.vector_store if config.embeddings.vector_store else {}
    )
    vector_store_type = vector_store_args.get("type", VectorStoreType.LanceDB)

    description_embedding_store = __get_embedding_description_store(
        vector_store_type=vector_store_type,
        config_args=vector_store_args,
    )
    entities = read_indexer_entities(final_nodes, final_entities, community_level)
    store_entity_semantic_embeddings(
        entities=entities, vectorstore=description_embedding_store
    )
    covariates = (
        read_indexer_covariates(final_covariates)
        if final_covariates is not None
        else []
    )

    search_engine = get_local_search_engine(
        config,
        reports=read_indexer_reports(
            final_community_reports, final_nodes, community_level
        ),
        text_units=read_indexer_text_units(final_text_units),
        entities=entities,
        relationships=read_indexer_relationships(final_relationships),
        covariates={"claims": covariates},
        description_embedding_store=description_embedding_store,
        response_type=response_type,
    )

    result = search_engine.search(query=query)
    reporter.success(f"Local Search Response: {result.response}")
    return result.response

这两个文件中的具体内容展示如下：

create_final_text_units.parquet

这个文件中存储了源文件被分割后的文本片段。

在这里插入图片描述

create_final_relationships.parquet

这里存储着节点之间的关系数据。

在这里插入图片描述

获取到原始数据后，local模式的查询逻辑如下：

基于用户的问题通过语义相似度计算相似度高的实体Entity，相关代码逻辑如下，代码路径为graphrag.query.context_builder.entity_extraction.map_query_to_entities：

def map_query_to_entities(
    query: str,
    text_embedding_vectorstore: BaseVectorStore,
    text_embedder: BaseTextEmbedding,
    all_entities: list[Entity],
    embedding_vectorstore_key: str = EntityVectorStoreKey.ID,
    include_entity_names: list[str] | None = None,
    exclude_entity_names: list[str] | None = None,
    k: int = 10,
    oversample_scaler: int = 2,
) -> list[Entity]:
    """Extract entities that match a given query using semantic similarity of text embeddings of query and entity descriptions."""
    if include_entity_names is None:
        include_entity_names = []
    if exclude_entity_names is None:
        exclude_entity_names = []
    matched_entities = []
    if query != "":
        # get entities with highest semantic similarity to query
        # oversample to account for excluded entities
        search_results = text_embedding_vectorstore.similarity_search_by_text(
            text=query,
            text_embedder=lambda t: text_embedder.embed(t),
            k=k * oversample_scaler,
        )
        for result in search_results:
            matched = get_entity_by_key(
                entities=all_entities,
                key=embedding_vectorstore_key,
                value=result.document.id,
            )
            if matched:
                matched_entities.append(matched)
    else:
        all_entities.sort(key=lambda x: x.rank if x.rank else 0, reverse=True)
        matched_entities = all_entities[:k]

    # filter out excluded entities
    if exclude_entity_names:
        matched_entities = [
            entity
            for entity in matched_entities
            if entity.title not in exclude_entity_names
        ]

    # add entities in the include_entity list
    included_entities = []
    for entity_name in include_entity_names:
        included_entities.extend(get_entity_by_name(all_entities, entity_name))
    return included_entities + matched_entities

基于选中的Entity获取对应的Community内容，调用方法如下，方法路径为graphrag.query.structured_search.local_search.mixed_context.LocalSearchMixedContext._build_community_context

    def _build_community_context(
        self,
        selected_entities: list[Entity],
        max_tokens: int = 4000,
        use_community_summary: bool = False,
        column_delimiter: str = "|",
        include_community_rank: bool = False,
        min_community_rank: int = 0,
        return_candidate_context: bool = False,
        context_name: str = "Reports",
    ) -> tuple[str, dict[str, pd.DataFrame]]:
        """Add community data to the context window until it hits the max_tokens limit."""
        if len(selected_entities) == 0 or len(self.community_reports) == 0:
            return ("", {context_name.lower(): pd.DataFrame()})

        community_matches = {}
        for entity in selected_entities:
            # increase count of the community that this entity belongs to
            if entity.community_ids:
                for community_id in entity.community_ids:
                    community_matches[community_id] = (
                        community_matches.get(community_id, 0) + 1
                    )

        # sort communities by number of matched entities and rank
        selected_communities = [
            self.community_reports[community_id]
            for community_id in community_matches
            if community_id in self.community_reports
        ]
        for community in selected_communities:
            if community.attributes is None:
                community.attributes = {}
            community.attributes["matches"] = community_matches[community.id]
        selected_communities.sort(
            key=lambda x: (x.attributes["matches"], x.rank),  # type: ignore
            reverse=True,  # type: ignore
        )
        for community in selected_communities:
            del community.attributes["matches"]  # type: ignore

        context_text, context_data = build_community_context(
            community_reports=selected_communities,
            token_encoder=self.token_encoder,
            use_community_summary=use_community_summary,
            column_delimiter=column_delimiter,
            shuffle_data=False,
            include_community_rank=include_community_rank,
            min_community_rank=min_community_rank,
            max_tokens=max_tokens,
            single_batch=True,
            context_name=context_name,
        )
        if isinstance(context_text, list) and len(context_text) > 0:
            context_text = "\n\n".join(context_text)

        if return_candidate_context:
            candidate_context_data = get_candidate_communities(
                selected_entities=selected_entities,
                community_reports=list(self.community_reports.values()),
                use_community_summary=use_community_summary,
                include_community_rank=include_community_rank,
            )
            context_key = context_name.lower()
            if context_key not in context_data:
                context_data[context_key] = candidate_context_data
                context_data[context_key]["in_context"] = False
            else:
                if (
                    "id" in candidate_context_data.columns
                    and "id" in context_data[context_key].columns
                ):
                    candidate_context_data["in_context"] = candidate_context_data[
                        "id"
                    ].isin(  # cspell:disable-line
                        context_data[context_key]["id"]
                    )
                    context_data[context_key] = candidate_context_data
                else:
                    context_data[context_key]["in_context"] = True
        return (str(context_text), context_data)

基于聊天上下文和community_content构建prompt内容，prompt模板的内容如下：

LOCAL_SEARCH_SYSTEM_PROMPT = """
---Role---

You are a helpful assistant responding to questions about data in the tables provided.

---Goal---

Generate a response of the target length and format that responds to the user's question, summarizing all information in the input data tables appropriate for the response length and format, and incorporating any relevant general knowledge.

If you don't know the answer, just say so. Do not make anything up.

Points supported by data should list their data references as follows:

"This is an example sentence supported by multiple data references [Data: <dataset name> (record ids); <dataset name> (record ids)]."

Do not list more than 5 record ids in a single reference. Instead, list the top 5 most relevant record ids and add "+more" to indicate that there are more.

For example:

"Person X is the owner of Company Y and subject to many allegations of wrongdoing [Data: Sources (15, 16), Reports (1), Entities (5, 7); Relationships (23); Claims (2, 7, 34, 46, 64, +more)]."

where 15, 16, 1, 5, 7, 23, 2, 7, 34, 46, and 64 represent the id (not the index) of the relevant data record.

Do not include information where the supporting evidence for it is not provided.

---Target response length and format---

{response_type}

---Data tables---

{context_data}

---Goal---

Generate a response of the target length and format that responds to the user's question, summarizing all information in the input data tables appropriate for the response length and format, and incorporating any relevant general knowledge.

If you don't know the answer, just say so. Do not make anything up.

Points supported by data should list their data references as follows:

"This is an example sentence supported by multiple data references [Data: <dataset name> (record ids); <dataset name> (record ids)]."

Do not list more than 5 record ids in a single reference. Instead, list the top 5 most relevant record ids and add "+more" to indicate that there are more.

For example:

"Person X is the owner of Company Y and subject to many allegations of wrongdoing [Data: Sources (15, 16), Reports (1), Entities (5, 7); Relationships (23); Claims (2, 7, 34, 46, 64, +more)]."

where 15, 16, 1, 5, 7, 23, 2, 7, 34, 46, and 64 represent the id (not the index) of the relevant data record.

Do not include information where the supporting evidence for it is not provided.

---Target response length and format---

{response_type}

Add sections and commentary to the response as appropriate for the length and format. Style the response in markdown.
"""