milvus-logo
LFAI
Home
  • Tutoriales

Graph RAG con Milvus

Open In Colab

La aplicación generalizada de grandes modelos lingüísticos pone de relieve la importancia de mejorar la precisión y pertinencia de sus respuestas. La Generación Mejorada por Recuperación (RAG) mejora los modelos con bases de conocimiento externas, proporcionando más información contextual y mitigando problemas como la alucinación y el conocimiento insuficiente. Sin embargo, basarse únicamente en paradigmas RAG sencillos tiene sus limitaciones, sobre todo cuando se trata de relaciones complejas entre entidades y preguntas con varios saltos, en las que el modelo suele tener dificultades para ofrecer respuestas precisas.

La introducción de grafos de conocimiento (KG) en el sistema GAR ofrece una nueva solución. Los KG presentan las entidades y sus relaciones de forma estructurada, proporcionando información más precisa y ayudando a RAG a gestionar mejor las tareas de respuesta a preguntas complejas. KG-RAG se encuentra todavía en sus primeras fases, y no hay consenso sobre cómo recuperar eficazmente entidades y relaciones a partir de KGs o cómo integrar la búsqueda de similitud vectorial con estructuras de grafos.

En este cuaderno, introducimos un enfoque sencillo pero potente para mejorar en gran medida el rendimiento de este escenario. Se trata de un simple paradigma RAG con recuperación multidireccional y posterior reordenación, pero implementa Graph RAG de forma lógica y consigue un rendimiento puntero en el manejo de preguntas multi-salto. Veamos cómo se implementa.

Requisitos previos

Antes de ejecutar este cuaderno, asegúrate de tener instaladas las siguientes dependencias:

$ pip install --upgrade --quiet pymilvus numpy scipy langchain langchain-core langchain-openai tqdm

Si utilizas Google Colab, para habilitar las dependencias que acabas de instalar, es posible que tengas que reiniciar el tiempo de ejecución (haz clic en el menú "Tiempo de ejecución" en la parte superior de la pantalla, y selecciona "Reiniciar sesión" en el menú desplegable).

Utilizaremos los modelos de OpenAI. Debes preparar la clave api OPENAI_API_KEY como variable de entorno.

import os

os.environ["OPENAI_API_KEY"] = "sk-***********"

Importe las bibliotecas y dependencias necesarias.

import numpy as np

from collections import defaultdict
from scipy.sparse import csr_matrix
from pymilvus import MilvusClient
from langchain_core.messages import AIMessage, HumanMessage
from langchain_core.prompts import ChatPromptTemplate, HumanMessagePromptTemplate
from langchain_core.output_parsers import StrOutputParser, JsonOutputParser
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from tqdm import tqdm

Inicialice la instancia del cliente Milvus, el LLM y el modelo de incrustación.

milvus_client = MilvusClient(uri="./milvus.db")

llm = ChatOpenAI(
    model="gpt-4o",
    temperature=0,
)
embedding_model = OpenAIEmbeddings(model="text-embedding-3-small")

Para los args en MilvusClient:

  • Establecer el uri como un archivo local, por ejemplo./milvus.db, es el método más conveniente, ya que utiliza automáticamente Milvus Lite para almacenar todos los datos en este archivo.
  • Si tiene una gran escala de datos, puede configurar un servidor Milvus más eficiente en docker o kubernetes. En esta configuración, por favor utilice la uri del servidor, por ejemplohttp://localhost:19530, como su uri.
  • Si desea utilizar Zilliz Cloud, el servicio en la nube totalmente gestionado para Milvus, ajuste uri y token, que corresponden al punto final público y a la clave Api en Zilliz Cloud.

Carga de datos sin conexión

Preparación de datos

Utilizaremos como ejemplo un nano conjunto de datos que introduce la relación entre la familia Bernoulli y Euler. El nano conjunto de datos contiene 4 pasajes y un conjunto de tripletas correspondientes, donde cada tripleta contiene un sujeto, un predicado y un objeto. En la práctica, puede utilizar cualquier enfoque para extraer las tripletas de su propio corpus personalizado.

nano_dataset = [
    {
        "passage": "Jakob Bernoulli (1654–1705): Jakob was one of the earliest members of the Bernoulli family to gain prominence in mathematics. He made significant contributions to calculus, particularly in the development of the theory of probability. He is known for the Bernoulli numbers and the Bernoulli theorem, a precursor to the law of large numbers. He was the older brother of Johann Bernoulli, another influential mathematician, and the two had a complex relationship that involved both collaboration and rivalry.",
        "triplets": [
            ["Jakob Bernoulli", "made significant contributions to", "calculus"],
            [
                "Jakob Bernoulli",
                "made significant contributions to",
                "the theory of probability",
            ],
            ["Jakob Bernoulli", "is known for", "the Bernoulli numbers"],
            ["Jakob Bernoulli", "is known for", "the Bernoulli theorem"],
            ["The Bernoulli theorem", "is a precursor to", "the law of large numbers"],
            ["Jakob Bernoulli", "was the older brother of", "Johann Bernoulli"],
        ],
    },
    {
        "passage": "Johann Bernoulli (1667–1748): Johann, Jakob’s younger brother, was also a major figure in the development of calculus. He worked on infinitesimal calculus and was instrumental in spreading the ideas of Leibniz across Europe. Johann also contributed to the calculus of variations and was known for his work on the brachistochrone problem, which is the curve of fastest descent between two points.",
        "triplets": [
            [
                "Johann Bernoulli",
                "was a major figure of",
                "the development of calculus",
            ],
            ["Johann Bernoulli", "was", "Jakob's younger brother"],
            ["Johann Bernoulli", "worked on", "infinitesimal calculus"],
            ["Johann Bernoulli", "was instrumental in spreading", "Leibniz's ideas"],
            ["Johann Bernoulli", "contributed to", "the calculus of variations"],
            ["Johann Bernoulli", "was known for", "the brachistochrone problem"],
        ],
    },
    {
        "passage": "Daniel Bernoulli (1700–1782): The son of Johann Bernoulli, Daniel made major contributions to fluid dynamics, probability, and statistics. He is most famous for Bernoulli’s principle, which describes the behavior of fluid flow and is fundamental to the understanding of aerodynamics.",
        "triplets": [
            ["Daniel Bernoulli", "was the son of", "Johann Bernoulli"],
            ["Daniel Bernoulli", "made major contributions to", "fluid dynamics"],
            ["Daniel Bernoulli", "made major contributions to", "probability"],
            ["Daniel Bernoulli", "made major contributions to", "statistics"],
            ["Daniel Bernoulli", "is most famous for", "Bernoulli’s principle"],
            [
                "Bernoulli’s principle",
                "is fundamental to",
                "the understanding of aerodynamics",
            ],
        ],
    },
    {
        "passage": "Leonhard Euler (1707–1783) was one of the greatest mathematicians of all time, and his relationship with the Bernoulli family was significant. Euler was born in Basel and was a student of Johann Bernoulli, who recognized his exceptional talent and mentored him in mathematics. Johann Bernoulli’s influence on Euler was profound, and Euler later expanded upon many of the ideas and methods he learned from the Bernoullis.",
        "triplets": [
            [
                "Leonhard Euler",
                "had a significant relationship with",
                "the Bernoulli family",
            ],
            ["leonhard Euler", "was born in", "Basel"],
            ["Leonhard Euler", "was a student of", "Johann Bernoulli"],
            ["Johann Bernoulli's influence", "was profound on", "Euler"],
        ],
    },
]

Construimos las entidades y relaciones de la siguiente manera:

  • La entidad es el sujeto o el objeto de la tripleta, por lo que los extraemos directamente de las tripletas.
  • Aquí construimos el concepto de relación concatenando directamente el sujeto, el predicado y el objeto con un espacio intermedio.

También preparamos un dict para asignar el id de entidad al id de relación, y otro dict para asignar el id de relación al id de pasaje para su uso posterior.

entityid_2_relationids = defaultdict(list)
relationid_2_passageids = defaultdict(list)

entities = []
relations = []
passages = []
for passage_id, dataset_info in enumerate(nano_dataset):
    passage, triplets = dataset_info["passage"], dataset_info["triplets"]
    passages.append(passage)
    for triplet in triplets:
        if triplet[0] not in entities:
            entities.append(triplet[0])
        if triplet[2] not in entities:
            entities.append(triplet[2])
        relation = " ".join(triplet)
        if relation not in relations:
            relations.append(relation)
            entityid_2_relationids[entities.index(triplet[0])].append(
                len(relations) - 1
            )
            entityid_2_relationids[entities.index(triplet[2])].append(
                len(relations) - 1
            )
        relationid_2_passageids[relations.index(relation)].append(passage_id)

Inserción de datos

Cree colecciones Milvus para entidad, relación y pasaje. La colección de entidades y la colección de relaciones se utilizan como colecciones principales para la construcción de grafos en nuestro método, mientras que la colección de pasajes se utiliza como comparación de recuperación RAG ingenua o propósito auxiliar.

embedding_dim = len(embedding_model.embed_query("foo"))


def create_milvus_collection(collection_name: str):
    if milvus_client.has_collection(collection_name=collection_name):
        milvus_client.drop_collection(collection_name=collection_name)
    milvus_client.create_collection(
        collection_name=collection_name,
        dimension=embedding_dim,
        consistency_level="Strong",
    )


entity_col_name = "entity_collection"
relation_col_name = "relation_collection"
passage_col_name = "passage_collection"
create_milvus_collection(entity_col_name)
create_milvus_collection(relation_col_name)
create_milvus_collection(passage_col_name)

Inserte los datos con su información de metadatos en las colecciones Milvus, incluidas las colecciones de entidades, relaciones y pasajes. La información de metadatos incluye el id de pasaje y el id de entidad o relación de adyacencia.

def milvus_insert(
    collection_name: str,
    text_list: list[str],
):
    batch_size = 512
    for row_id in tqdm(range(0, len(text_list), batch_size), desc="Inserting"):
        batch_texts = text_list[row_id : row_id + batch_size]
        batch_embeddings = embedding_model.embed_documents(batch_texts)

        batch_ids = [row_id + j for j in range(len(batch_texts))]
        batch_data = [
            {
                "id": id_,
                "text": text,
                "vector": vector,
            }
            for id_, text, vector in zip(batch_ids, batch_texts, batch_embeddings)
        ]
        milvus_client.insert(
            collection_name=collection_name,
            data=batch_data,
        )


milvus_insert(
    collection_name=relation_col_name,
    text_list=relations,
)

milvus_insert(
    collection_name=entity_col_name,
    text_list=entities,
)

milvus_insert(
    collection_name=passage_col_name,
    text_list=passages,
)
Inserting: 100%|███████████████████████████████████| 1/1 [00:00<00:00,  1.02it/s]
Inserting: 100%|███████████████████████████████████| 1/1 [00:00<00:00,  1.39it/s]
Inserting: 100%|███████████████████████████████████| 1/1 [00:00<00:00,  2.28it/s]

Consulta en línea

Recuperación de similitudes

Recuperamos las K entidades y relaciones más similares a partir de la consulta de Milvus.

Al realizar la recuperación de entidades, primero debemos extraer las entidades de la consulta del texto de la consulta utilizando algún método específico como el NER (reconocimiento de entidades con nombre). Para simplificar, preparamos aquí los resultados del NER. En la práctica, puede utilizar cualquier otro modelo o método para extraer las entidades de la consulta.

query = "What contribution did the son of Euler's teacher make?"

query_ner_list = ["Euler"]
# query_ner_list = ner(query) # In practice, replace it with your custom NER approach

query_ner_embeddings = [
    embedding_model.embed_query(query_ner) for query_ner in query_ner_list
]

top_k = 3

entity_search_res = milvus_client.search(
    collection_name=entity_col_name,
    data=query_ner_embeddings,
    limit=top_k,
    output_fields=["id"],
)

query_embedding = embedding_model.embed_query(query)

relation_search_res = milvus_client.search(
    collection_name=relation_col_name,
    data=[query_embedding],
    limit=top_k,
    output_fields=["id"],
)[0]

Expandir subgrafos

Utilizamos las entidades y relaciones recuperadas para expandir el subgrafo y obtener las relaciones candidatas y, a continuación, fusionarlas de las dos formas. A continuación se muestra un diagrama de flujo del proceso de expansión del subgrafo:

Aquí construimos una matriz de adyacencia y utilizamos la multiplicación de matrices para calcular la información de mapeo de adyacencia en unos pocos grados. De este modo, podemos obtener rápidamente información de cualquier grado de expansión.

# Construct the adjacency matrix of entities and relations where the value of the adjacency matrix is 1 if an entity is related to a relation, otherwise 0.
entity_relation_adj = np.zeros((len(entities), len(relations)))
for entity_id, entity in enumerate(entities):
    entity_relation_adj[entity_id, entityid_2_relationids[entity_id]] = 1

# Convert the adjacency matrix to a sparse matrix for efficient computation.
entity_relation_adj = csr_matrix(entity_relation_adj)

# Use the entity-relation adjacency matrix to construct 1 degree entity-entity and relation-relation adjacency matrices.
entity_adj_1_degree = entity_relation_adj @ entity_relation_adj.T
relation_adj_1_degree = entity_relation_adj.T @ entity_relation_adj

# Specify the target degree of the subgraph to be expanded.
# 1 or 2 is enough for most cases.
target_degree = 1

# Compute the target degree adjacency matrices using matrix multiplication.
entity_adj_target_degree = entity_adj_1_degree
for _ in range(target_degree - 1):
    entity_adj_target_degree = entity_adj_target_degree * entity_adj_1_degree
relation_adj_target_degree = relation_adj_1_degree
for _ in range(target_degree - 1):
    relation_adj_target_degree = relation_adj_target_degree * relation_adj_1_degree

entity_relation_adj_target_degree = entity_adj_target_degree @ entity_relation_adj

Tomando el valor de la matriz de expansión de grado objetivo, podemos expandir fácilmente el grado correspondiente de la entidad y las relaciones recuperadas para obtener todas las relaciones del subgrafo.

expanded_relations_from_relation = set()
expanded_relations_from_entity = set()
# You can set the similarity threshold here to guarantee the quality of the retrieved ones.
# entity_sim_filter_thresh = ...
# relation_sim_filter_thresh = ...

filtered_hit_relation_ids = [
    relation_res["entity"]["id"]
    for relation_res in relation_search_res
    # if relation_res['distance'] > relation_sim_filter_thresh
]
for hit_relation_id in filtered_hit_relation_ids:
    expanded_relations_from_relation.update(
        relation_adj_target_degree[hit_relation_id].nonzero()[1].tolist()
    )

filtered_hit_entity_ids = [
    one_entity_res["entity"]["id"]
    for one_entity_search_res in entity_search_res
    for one_entity_res in one_entity_search_res
    # if one_entity_res['distance'] > entity_sim_filter_thresh
]

for filtered_hit_entity_id in filtered_hit_entity_ids:
    expanded_relations_from_entity.update(
        entity_relation_adj_target_degree[filtered_hit_entity_id].nonzero()[1].tolist()
    )

# Merge the expanded relations from the relation and entity retrieval ways.
relation_candidate_ids = list(
    expanded_relations_from_relation | expanded_relations_from_entity
)

relation_candidate_texts = [
    relations[relation_id] for relation_id in relation_candidate_ids
]

Hemos obtenido las relaciones candidatas expandiendo el subgrafo, que serán reordenadas por LLM en el siguiente paso.

Nueva clasificación LLM

En esta etapa, desplegamos el potente mecanismo de autoatención de LLM para filtrar y refinar aún más el conjunto de relaciones candidatas. Empleamos una pregunta única, incorporando la consulta y el conjunto de relaciones candidatas en la pregunta, e instruimos a LLM para que seleccione las relaciones potenciales que podrían ayudar a responder la consulta. Dado que algunas consultas pueden ser complejas, adoptamos el enfoque de la cadena de pensamiento, permitiendo a LLM articular su proceso de pensamiento en su respuesta. Estipulamos que la respuesta de LLM esté en formato json para facilitar el análisis sintáctico.

query_prompt_one_shot_input = """I will provide you with a list of relationship descriptions. Your task is to select 3 relationships that may be useful to answer the given question. Please return a JSON object containing your thought process and a list of the selected relationships in order of their relevance.

Question:
When was the mother of the leader of the Third Crusade born?

Relationship descriptions:
[1] Eleanor was born in 1122.
[2] Eleanor married King Louis VII of France.
[3] Eleanor was the Duchess of Aquitaine.
[4] Eleanor participated in the Second Crusade.
[5] Eleanor had eight children.
[6] Eleanor was married to Henry II of England.
[7] Eleanor was the mother of Richard the Lionheart.
[8] Richard the Lionheart was the King of England.
[9] Henry II was the father of Richard the Lionheart.
[10] Henry II was the King of England.
[11] Richard the Lionheart led the Third Crusade.

"""
query_prompt_one_shot_output = """{"thought_process": "To answer the question about the birth of the mother of the leader of the Third Crusade, I first need to identify who led the Third Crusade and then determine who his mother was. After identifying his mother, I can look for the relationship that mentions her birth.", "useful_relationships": ["[11] Richard the Lionheart led the Third Crusade", "[7] Eleanor was the mother of Richard the Lionheart", "[1] Eleanor was born in 1122"]}"""

query_prompt_template = """Question:
{question}

Relationship descriptions:
{relation_des_str}

"""


def rerank_relations(
    query: str, relation_candidate_texts: list[str], relation_candidate_ids: list[str]
) -> list[int]:
    relation_des_str = "\n".join(
        map(
            lambda item: f"[{item[0]}] {item[1]}",
            zip(relation_candidate_ids, relation_candidate_texts),
        )
    ).strip()
    rerank_prompts = ChatPromptTemplate.from_messages(
        [
            HumanMessage(query_prompt_one_shot_input),
            AIMessage(query_prompt_one_shot_output),
            HumanMessagePromptTemplate.from_template(query_prompt_template),
        ]
    )
    rerank_chain = (
        rerank_prompts
        | llm.bind(response_format={"type": "json_object"})
        | JsonOutputParser()
    )
    rerank_res = rerank_chain.invoke(
        {"question": query, "relation_des_str": relation_des_str}
    )
    rerank_relation_ids = []
    rerank_relation_lines = rerank_res["useful_relationships"]
    id_2_lines = {}
    for line in rerank_relation_lines:
        id_ = int(line[line.find("[") + 1 : line.find("]")])
        id_2_lines[id_] = line.strip()
        rerank_relation_ids.append(id_)
    return rerank_relation_ids


rerank_relation_ids = rerank_relations(
    query,
    relation_candidate_texts=relation_candidate_texts,
    relation_candidate_ids=relation_candidate_ids,
)

Obtener resultados finales

Podemos obtener los pasajes finales recuperados de las relaciones reordenadas.

final_top_k = 2

final_passages = []
final_passage_ids = []
for relation_id in rerank_relation_ids:
    for passage_id in relationid_2_passageids[relation_id]:
        if passage_id not in final_passage_ids:
            final_passage_ids.append(passage_id)
            final_passages.append(passages[passage_id])
passages_from_our_method = final_passages[:final_top_k]

Podemos comparar los resultados con el método ingenuo RAG, que recupera los pasajes topK basados en la incrustación de la consulta directamente de la colección de pasajes.

naive_passage_res = milvus_client.search(
    collection_name=passage_col_name,
    data=[query_embedding],
    limit=final_top_k,
    output_fields=["text"],
)[0]
passages_from_naive_rag = [res["entity"]["text"] for res in naive_passage_res]

print(
    f"Passages retrieved from naive RAG: \n{passages_from_naive_rag}\n\n"
    f"Passages retrieved from our method: \n{passages_from_our_method}\n\n"
)


prompt = ChatPromptTemplate.from_messages(
    [
        (
            "human",
            """Use the following pieces of retrieved context to answer the question. If there is not enough information in the retrieved context to answer the question, just say that you don't know.
Question: {question}
Context: {context}
Answer:""",
        )
    ]
)

rag_chain = prompt | llm | StrOutputParser()

answer_from_naive_rag = rag_chain.invoke(
    {"question": query, "context": "\n".join(passages_from_naive_rag)}
)
answer_from_our_method = rag_chain.invoke(
    {"question": query, "context": "\n".join(passages_from_our_method)}
)

print(
    f"Answer from naive RAG: {answer_from_naive_rag}\n\nAnswer from our method: {answer_from_our_method}"
)
Passages retrieved from naive RAG: 
['Leonhard Euler (1707–1783) was one of the greatest mathematicians of all time, and his relationship with the Bernoulli family was significant. Euler was born in Basel and was a student of Johann Bernoulli, who recognized his exceptional talent and mentored him in mathematics. Johann Bernoulli’s influence on Euler was profound, and Euler later expanded upon many of the ideas and methods he learned from the Bernoullis.', 'Johann Bernoulli (1667–1748): Johann, Jakob’s younger brother, was also a major figure in the development of calculus. He worked on infinitesimal calculus and was instrumental in spreading the ideas of Leibniz across Europe. Johann also contributed to the calculus of variations and was known for his work on the brachistochrone problem, which is the curve of fastest descent between two points.']

Passages retrieved from our method: 
['Leonhard Euler (1707–1783) was one of the greatest mathematicians of all time, and his relationship with the Bernoulli family was significant. Euler was born in Basel and was a student of Johann Bernoulli, who recognized his exceptional talent and mentored him in mathematics. Johann Bernoulli’s influence on Euler was profound, and Euler later expanded upon many of the ideas and methods he learned from the Bernoullis.', 'Daniel Bernoulli (1700–1782): The son of Johann Bernoulli, Daniel made major contributions to fluid dynamics, probability, and statistics. He is most famous for Bernoulli’s principle, which describes the behavior of fluid flow and is fundamental to the understanding of aerodynamics.']


Answer from naive RAG: I don't know. The retrieved context does not provide information about the contributions made by the son of Euler's teacher.

Answer from our method: The son of Euler's teacher, Daniel Bernoulli, made major contributions to fluid dynamics, probability, and statistics. He is most famous for Bernoulli’s principle, which describes the behavior of fluid flow and is fundamental to the understanding of aerodynamics.

Como podemos ver, los pasajes recuperados con el método ingenuo RAG omitieron un pasaje verdadero, lo que condujo a una respuesta errónea. Los pasajes recuperados con nuestro método son correctos, y ayudan a obtener una respuesta precisa a la pregunta.

Traducido porDeepLogo

Feedback

¿Fue útil esta página?