检索器路由查询引擎¶

在本教程中，我们将基于检索器定义一个路由查询引擎。该检索器会选取一组节点，我们将据此选择最合适的查询引擎。

为此我们使用全新的 ToolRetrieverRouterQueryEngine 类！

安装配置¶

如果您在 Colab 上打开此 Notebook，可能需要安装 LlamaIndex 🦙。

In [ ]:

!pip install llama-index

!pip install llama-index

In [ ]:

# NOTE: This is ONLY necessary in jupyter notebook.
# Details: Jupyter runs an event-loop behind the scenes.
#          This results in nested event-loops when we start an event-loop to make async queries.
#          This is normally not allowed, we use nest_asyncio to allow it for convenience.
import nest_asyncio

nest_asyncio.apply()

# NOTE: This is ONLY necessary in jupyter notebook. # Details: Jupyter runs an event-loop behind the scenes. # This results in nested event-loops when we start an event-loop to make async queries. # This is normally not allowed, we use nest_asyncio to allow it for convenience. import nest_asyncio nest_asyncio.apply()

In [ ]:

import logging
import sys

logging.basicConfig(stream=sys.stdout, level=logging.INFO)
logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))

from llama_index.core import (
    VectorStoreIndex,
    SimpleDirectoryReader,
    StorageContext,
)
from llama_index.core import SummaryIndex

import logging import sys logging.basicConfig(stream=sys.stdout, level=logging.INFO) logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout)) from llama_index.core import ( VectorStoreIndex, SimpleDirectoryReader, StorageContext, ) from llama_index.core import SummaryIndex

INFO:numexpr.utils:Note: NumExpr detected 12 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
Note: NumExpr detected 12 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
INFO:numexpr.utils:NumExpr defaulting to 8 threads.
NumExpr defaulting to 8 threads.

/Users/jerryliu/Programming/gpt_index/.venv/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from .autonotebook import tqdm as notebook_tqdm

下载数据

In [ ]:

!mkdir -p 'data/paul_graham/'
!wget 'https://raw.githubusercontent.com/run-llama/llama_index/main/docs/docs/examples/data/paul_graham/paul_graham_essay.txt' -O 'data/paul_graham/paul_graham_essay.txt'

!mkdir -p 'data/paul_graham/' !wget 'https://raw.githubusercontent.com/run-llama/llama_index/main/docs/docs/examples/data/paul_graham/paul_graham_essay.txt' -O 'data/paul_graham/paul_graham_essay.txt'

加载数据¶

我们首先演示如何将文档转换为节点集合，并将其插入文档存储库。

In [ ]:

# load documents
documents = SimpleDirectoryReader("./data/paul_graham").load_data()

# load documents documents = SimpleDirectoryReader("./data/paul_graham").load_data()

In [ ]:

from llama_index.core import Settings

# initialize settings (set chunk size)
Settings.chunk_size = 1024
nodes = Settings.node_parser.get_nodes_from_documents(documents)

from llama_index.core import Settings # initialize settings (set chunk size) Settings.chunk_size = 1024 nodes = Settings.node_parser.get_nodes_from_documents(documents)

In [ ]:

# initialize storage context (by default it's in-memory)
storage_context = StorageContext.from_defaults()
storage_context.docstore.add_documents(nodes)

# initialize storage context (by default it's in-memory) storage_context = StorageContext.from_defaults() storage_context.docstore.add_documents(nodes)

在同一数据集上定义摘要索引与向量索引¶

In [ ]:

summary_index = SummaryIndex(nodes, storage_context=storage_context)
vector_index = VectorStoreIndex(nodes, storage_context=storage_context)

summary_index = SummaryIndex(nodes, storage_context=storage_context) vector_index = VectorStoreIndex(nodes, storage_context=storage_context)

INFO:llama_index.token_counter.token_counter:> [build_index_from_nodes] Total LLM token usage: 0 tokens
> [build_index_from_nodes] Total LLM token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [build_index_from_nodes] Total embedding token usage: 0 tokens
> [build_index_from_nodes] Total embedding token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [build_index_from_nodes] Total LLM token usage: 0 tokens
> [build_index_from_nodes] Total LLM token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [build_index_from_nodes] Total embedding token usage: 17038 tokens
> [build_index_from_nodes] Total embedding token usage: 17038 tokens

为这些索引定义查询引擎与工具¶

我们为每个索引定义一个查询引擎，随后使用 QueryEngineTool 对这些引擎进行封装。

In [ ]:

from llama_index.core.tools import QueryEngineTool

list_query_engine = summary_index.as_query_engine(
    response_mode="tree_summarize", use_async=True
)
vector_query_engine = vector_index.as_query_engine(
    response_mode="tree_summarize", use_async=True
)

list_tool = QueryEngineTool.from_defaults(
    query_engine=list_query_engine,
    description="Useful for questions asking for a biography of the author.",
)
vector_tool = QueryEngineTool.from_defaults(
    query_engine=vector_query_engine,
    description=(
        "Useful for retrieving specific snippets from the author's life, like"
        " his time in college, his time in YC, or more."
    ),
)

from llama_index.core.tools import QueryEngineTool list_query_engine = summary_index.as_query_engine( response_mode="tree_summarize", use_async=True ) vector_query_engine = vector_index.as_query_engine( response_mode="tree_summarize", use_async=True ) list_tool = QueryEngineTool.from_defaults( query_engine=list_query_engine, description="Useful for questions asking for a biography of the author.", ) vector_tool = QueryEngineTool.from_defaults( query_engine=vector_query_engine, description=( "Useful for retrieving specific snippets from the author's life, like" " his time in college, his time in YC, or more." ), )

定义检索增强型路由查询引擎¶

我们定义了一个通过检索机制增强的路由查询引擎，用于处理选项集合过大的情况。

具体实现时，首先在查询引擎工具集上定义一个ObjectIndex。该ObjectIndex基于底层索引数据结构（如向量索引、关键词索引），并能够将QueryEngineTool对象与我们的索引进行序列化/反序列化转换。

随后使用ToolRetrieverRouterQueryEngine类，并传入基于QueryEngineTool对象的ObjectRetriever。这个ObjectRetriever与我们的ObjectIndex相对应。

该检索器能够在查询时动态获取相关的查询引擎。通过这种方式，我们可以传入任意数量的查询引擎工具，而无需担心提示词的长度限制问题。

In [ ]:

from llama_index.core import VectorStoreIndex
from llama_index.core.objects import ObjectIndex

obj_index = ObjectIndex.from_objects(
    [list_tool, vector_tool],
    index_cls=VectorStoreIndex,
)

from llama_index.core import VectorStoreIndex from llama_index.core.objects import ObjectIndex obj_index = ObjectIndex.from_objects( [list_tool, vector_tool], index_cls=VectorStoreIndex, )

INFO:llama_index.token_counter.token_counter:> [build_index_from_nodes] Total LLM token usage: 0 tokens
> [build_index_from_nodes] Total LLM token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [build_index_from_nodes] Total embedding token usage: 59 tokens
> [build_index_from_nodes] Total embedding token usage: 59 tokens

In [ ]:

from llama_index.core.query_engine import ToolRetrieverRouterQueryEngine

query_engine = ToolRetrieverRouterQueryEngine(obj_index.as_retriever())

from llama_index.core.query_engine import ToolRetrieverRouterQueryEngine query_engine = ToolRetrieverRouterQueryEngine(obj_index.as_retriever())

In [ ]:

response = query_engine.query("What is a biography of the author's life?")

response = query_engine.query("What is a biography of the author's life?")

INFO:llama_index.token_counter.token_counter:> [retrieve] Total LLM token usage: 0 tokens
> [retrieve] Total LLM token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total embedding token usage: 10 tokens
> [retrieve] Total embedding token usage: 10 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total LLM token usage: 0 tokens
> [retrieve] Total LLM token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total embedding token usage: 0 tokens
> [retrieve] Total embedding token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total LLM token usage: 2111 tokens
> [get_response] Total LLM token usage: 2111 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total embedding token usage: 0 tokens
> [get_response] Total embedding token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total LLM token usage: 0 tokens
> [retrieve] Total LLM token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total embedding token usage: 0 tokens
> [retrieve] Total embedding token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total LLM token usage: 2148 tokens
> [get_response] Total LLM token usage: 2148 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total embedding token usage: 0 tokens
> [get_response] Total embedding token usage: 0 tokens
INFO:llama_index.query_engine.router_query_engine:Combining responses from multiple query engines.
Combining responses from multiple query engines.
INFO:llama_index.token_counter.token_counter:> [get_response] Total LLM token usage: 1063 tokens
> [get_response] Total LLM token usage: 1063 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total embedding token usage: 0 tokens
> [get_response] Total embedding token usage: 0 tokens

In [ ]:

print(str(response))

print(str(response))

The author is a creative person who has had a varied and interesting life. They grew up in the US and went to college, but then decided to take a break and pursue their passion for art. They applied to two art schools, RISD in the US and the Accademia di Belli Arti in Florence, and were accepted to both. They chose to go to Florence, where they took the entrance exam and passed. They then spent a year living in Florence, studying art at the Accademia and painting still lives in their bedroom. After their year in Florence, the author returned to the US and completed their BFA program at RISD. They then went on to pursue a PhD in computer science at MIT, where they wrote a dissertation on the evolution of computers. During their time at MIT, they also did consulting work and wrote essays on topics they had been thinking about. After completing their PhD, the author started a software company, Viaweb, which was eventually acquired by Yahoo. They then went on to write essays and articles about their experiences in the tech industry. They also wrote an essay about how to choose what to work on, which was based on their own experience. The author then moved back to Florence, where they found a rent-stabilized apartment and continued to pursue their interest in art. They wrote about their experiences in the art world, and experienced the reactions of readers to their essays. The author is now a successful writer and continues to write essays and articles about topics they are passionate about. 

In summary, the author's life has been a journey of exploration and creativity. They have experienced a wide range of different things in their life, from art school to computer science to the tech industry, and have used their experiences to inform their writing. They have pursued their passion for art, and have used their knowledge and experience to create meaningful work.

In [ ]:

response

response

Out[ ]:

"\nThe author is a creative person who has had a varied and interesting life. They grew up in the US and went to college, but then decided to take a break and pursue their passion for art. They applied to two art schools, RISD in the US and the Accademia di Belli Arti in Florence, and were accepted to both. They chose to go to Florence, where they took the entrance exam and passed. They then spent a year living in Florence, studying art at the Accademia and painting still lives in their bedroom. After their year in Florence, the author returned to the US and completed their BFA program at RISD. They then went on to pursue a PhD in computer science at MIT, where they wrote a dissertation on the evolution of computers. During their time at MIT, they also did consulting work and wrote essays on topics they had been thinking about. After completing their PhD, the author started a software company, Viaweb, which was eventually acquired by Yahoo. They then went on to write essays and articles about their experiences in the tech industry. They also wrote an essay about how to choose what to work on, which was based on their own experience. The author then moved back to Florence, where they found a rent-stabilized apartment and continued to pursue their interest in art. They wrote about their experiences in the art world, and experienced the reactions of readers to their essays. The author is now a successful writer and continues to write essays and articles about topics they are passionate about. \n\nIn summary, the author's life has been a journey of exploration and creativity. They have experienced a wide range of different things in their life, from art school to computer science to the tech industry, and have used their experiences to inform their writing. They have pursued their passion for art, and have used their knowledge and experience to create meaningful work."

In [ ]:

response = query_engine.query(
    "What did Paul Graham do during his time in college?"
)

response = query_engine.query( "What did Paul Graham do during his time in college?" )

INFO:llama_index.token_counter.token_counter:> [retrieve] Total LLM token usage: 0 tokens
> [retrieve] Total LLM token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total embedding token usage: 11 tokens
> [retrieve] Total embedding token usage: 11 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total LLM token usage: 0 tokens
> [retrieve] Total LLM token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total embedding token usage: 0 tokens
> [retrieve] Total embedding token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total LLM token usage: 1947 tokens
> [get_response] Total LLM token usage: 1947 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total embedding token usage: 0 tokens
> [get_response] Total embedding token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total LLM token usage: 0 tokens
> [retrieve] Total LLM token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [retrieve] Total embedding token usage: 0 tokens
> [retrieve] Total embedding token usage: 0 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total LLM token usage: 1947 tokens
> [get_response] Total LLM token usage: 1947 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total embedding token usage: 0 tokens
> [get_response] Total embedding token usage: 0 tokens
INFO:llama_index.query_engine.router_query_engine:Combining responses from multiple query engines.
Combining responses from multiple query engines.
INFO:llama_index.token_counter.token_counter:> [get_response] Total LLM token usage: 316 tokens
> [get_response] Total LLM token usage: 316 tokens
INFO:llama_index.token_counter.token_counter:> [get_response] Total embedding token usage: 0 tokens
> [get_response] Total embedding token usage: 0 tokens

In [ ]:

print(str(response))

print(str(response))

Paul Graham studied philosophy in college, but he did not pursue AI. He continued to work on programming outside of school, writing simple games, a program to predict how high his model rockets would fly, and a word processor. He eventually convinced his father to buy him a TRS-80 computer, which he used to further his programming skills.