llamaindex 集成查询引擎指南

集成查询引擎指南

概述

在构建检索增强生成（RAG）应用程序时，通常需要尝试不同的查询管道（例如，top-k 检索、关键词搜索、知识图谱）。如果我们能够同时尝试多种策略，并让语言模型（LLM）评估每个查询的相关性并综合结果，那将会非常有用。

本指南展示了如何在《了不起的盖茨比》上进行集成检索。我们通过不同的分块大小和不同的索引进行集成检索。

安装依赖

首先，我们需要安装必要的库：

%pip install llama-index-llms-openai
!pip install llama-index

设置

在 Jupyter Notebook 中，我们需要应用 nest_asyncio 以允许嵌套的异步查询：

import nest_asyncio
nest_asyncio.apply()

下载数据

下载《了不起的盖茨比》的全文：

!wget 'https://raw.githubusercontent.com/jerryjliu/llama_index/main/examples/gatsby/gatsby_full.txt' -O 'gatsby_full.txt'

加载数据

我们将文档转换为节点，并插入到文档存储中：

from llama_index.core import SimpleDirectoryReader

documents = SimpleDirectoryReader(
    input_files=["./gatsby_full.txt"]
).load_data()

定义查询引擎

初始化设置（设置分块大小）：

from llama_index.llms.openai import OpenAI
from llama_index.core import Settings

Settings.llm = OpenAI(model="gpt-3.5-turbo")
Settings.chunk_size = 1024

nodes = Settings.node_parser.get_nodes_from_documents(documents)

初始化存储上下文（默认是内存中的）：

from llama_index.core import StorageContext

storage_context = StorageContext.from_defaults()
storage_context.docstore.add_documents(nodes)

初始化关键词表索引和向量存储索引：

from llama_index.core import SimpleKeywordTableIndex, VectorStoreIndex

keyword_index = SimpleKeywordTableIndex(
    nodes,
    storage_context=storage_context,
    show_progress=True,
)
vector_index = VectorStoreIndex(
    nodes,
    storage_context=storage_context,
    show_progress=True,
)

定义查询模板：

from llama_index.core import PromptTemplate

QA_PROMPT_TMPL = (
    "Context information is below.\n"
    "---------------------\n"
    "{context_str}\n"
    "---------------------\n"
    "Given the context information and not prior knowledge, "
    "answer the question. If the answer is not in the context, inform "
    "the user that you can't answer the question - DO NOT MAKE UP AN ANSWER.\n"
    "In addition to returning the answer, also return a relevance score as to "
    "how relevant the answer is to the question. "
    "Question: {query_str}\n"
    "Answer (including relevance score): "
)
QA_PROMPT = PromptTemplate(QA_PROMPT_TMPL)

初始化关键词查询引擎和向量查询引擎：

keyword_query_engine = keyword_index.as_query_engine(
    text_qa_template=QA_PROMPT
)
vector_query_engine = vector_index.as_query_engine(text_qa_template=QA_PROMPT)

测试查询引擎：

response = vector_query_engine.query(
    "Describe and summarize the interactions between Gatsby and Daisy"
)
print(response)

response = keyword_query_engine.query(
    "Describe and summarize the interactions between Gatsby and Daisy"
)
print(response)

定义路由器查询引擎

初始化查询引擎工具：

from llama_index.core.tools import QueryEngineTool

keyword_tool = QueryEngineTool.from_defaults(
    query_engine=keyword_query_engine,
    description="Useful for answering questions about this essay",
)
vector_tool = QueryEngineTool.from_defaults(
    query_engine=vector_query_engine,
    description="Useful for answering questions about this essay",
)

初始化路由器查询引擎：

from llama_index.core.query_engine import RouterQueryEngine
from llama_index.core.selectors import LLMSingleSelector, LLMMultiSelector
from llama_index.core.selectors import (
    PydanticMultiSelector,
    PydanticSingleSelector,
)
from llama_index.core.response_synthesizers import TreeSummarize

TREE_SUMMARIZE_PROMPT_TMPL = (
    "Context information from multiple sources is below. Each source may or"
    " may not have \na relevance score attached to"
    " it.\n---------------------\n{context_str}\n---------------------\nGiven"
    " the information from multiple sources and their associated relevance"
    " scores (if provided) and not prior knowledge, answer the question. If"
    " the answer is not in the context, inform the user that you can't answer"
    " the question.\nQuestion: {query_str}\nAnswer: "
)

tree_summarize = TreeSummarize(
    summary_template=PromptTemplate(TREE_SUMMARIZE_PROMPT_TMPL)
)

query_engine = RouterQueryEngine(
    selector=LLMMultiSelector.from_defaults(),
    query_engine_tools=[
        keyword_tool,
        vector_tool,
    ],
    summarizer=tree_summarize,
)

实验查询

使用异步查询进行实验：

response = await query_engine.aquery(
    "Describe and summarize the interactions between Gatsby and Daisy"
)
print(response)

response = await query_engine.aquery(
    "What part of his past is Gatsby trying to recapture?"
)
print(response)

拓展内容

集成查询引擎的优势：

1. 多样性：通过集成多种查询策略，可以提高检索结果的多样性和全面性。
2. 鲁棒性：不同的查询策略可能在不同的场景下表现更好，集成查询引擎可以提高系统的鲁棒性。
3. 自动化评估：通过 LLM 自动评估查询结果的相关性，可以减少人工干预，提高效率。

实际应用场景：

1. 问答系统：在问答系统中，集成查询引擎可以提供更准确和全面的答案。
2. 信息检索：在信息检索系统中，集成查询引擎可以提高检索结果的质量。
3. 知识图谱：在知识图谱中，集成查询引擎可以提供更丰富的知识关联和推理。

通过这些详细的讲解和示例，学生们可以更好地理解和掌握集成查询引擎的定义与实现方法，从而在实际项目中高效地应用。