使用 AgentQL 实现智能体工作流 + 研究助手¶

本教程将演示如何通过 AgentWorkflow 构建一个基于 OpenAI 的研究助手智能体，该智能体整合了 AgentQL 浏览器工具、Playwright 工具以及 DuckDuckGo 搜索工具。该智能体能够执行网络搜索以获取研究主题的相关资源，与这些资源进行交互，并从中提取关键元数据（包括标题、作者、出版详情和摘要等信息）。

初始设置¶

我们需要准备以下主要事项：

• OpenAI 的 API 密钥
• AgentQL 的 API 密钥

如果你在 Colab 上打开这个 Notebook，可能需要安装 LlamaIndex 🦙 和 Playwright。

In [ ]:

%pip install llama-index
%pip install llama-index-tools-agentql
%pip install llama-index-tools-playwright
%pip install llama-index-tools-duckduckgo

!playwright install

%pip install llama-index %pip install llama-index-tools-agentql %pip install llama-index-tools-playwright %pip install llama-index-tools-duckduckgo !playwright install

将您的 OPENAI_API_KEY 和 AGENTQL_API_KEY 密钥存储在 Google Colab 密钥管理中。

In [ ]:

import os

from google.colab import userdata

os.environ["AGENTQL_API_KEY"] = userdata.get("AGENTQL_API_KEY")
os.environ["OPENAI_API_KEY"] = userdata.get("OPENAI_API_KEY")

import os from google.colab import userdata os.environ["AGENTQL_API_KEY"] = userdata.get("AGENTQL_API_KEY") os.environ["OPENAI_API_KEY"] = userdata.get("OPENAI_API_KEY")

让我们首先为笔记本启用异步模式，因为在 Google Colab 这样的在线环境中仅支持 AgentQL 的异步版本。

In [ ]:

import nest_asyncio

nest_asyncio.apply()

import nest_asyncio nest_asyncio.apply()

创建一个 async_browser 实例并选择您想要使用的 Playwright 工具。

In [ ]:

from llama_index.tools.playwright.base import PlaywrightToolSpec

async_browser = await PlaywrightToolSpec.create_async_playwright_browser(
    headless=True
)

playwright_tool = PlaywrightToolSpec(async_browser=async_browser)
playwright_tool_list = playwright_tool.to_tool_list()
playwright_agent_tool_list = [
    tool
    for tool in playwright_tool_list
    if tool.metadata.name in ["click", "get_current_page", "navigate_to"]
]

from llama_index.tools.playwright.base import PlaywrightToolSpec async_browser = await PlaywrightToolSpec.create_async_playwright_browser( headless=True ) playwright_tool = PlaywrightToolSpec(async_browser=async_browser) playwright_tool_list = playwright_tool.to_tool_list() playwright_agent_tool_list = [ tool for tool in playwright_tool_list if tool.metadata.name in ["click", "get_current_page", "navigate_to"] ]

导入 AgentQL 浏览器工具 和 DuckDuckGo 全文搜索工具。

In [ ]:

from llama_index.tools.agentql import AgentQLBrowserToolSpec
from llama_index.tools.duckduckgo import DuckDuckGoSearchToolSpec

duckduckgo_search_tool = [
    tool
    for tool in DuckDuckGoSearchToolSpec().to_tool_list()
    if tool.metadata.name == "duckduckgo_full_search"
]

agentql_browser_tool = AgentQLBrowserToolSpec(async_browser=async_browser)

from llama_index.tools.agentql import AgentQLBrowserToolSpec from llama_index.tools.duckduckgo import DuckDuckGoSearchToolSpec duckduckgo_search_tool = [ tool for tool in DuckDuckGoSearchToolSpec().to_tool_list() if tool.metadata.name == "duckduckgo_full_search" ] agentql_browser_tool = AgentQLBrowserToolSpec(async_browser=async_browser)

现在我们可以创建一个使用已导入工具的 AgentWorkFlow。

In [ ]:

from llama_index.llms.openai import OpenAI
from llama_index.core.agent.workflow import AgentWorkflow

llm = OpenAI(model="gpt-4o")

workflow = AgentWorkflow.from_tools_or_functions(
    playwright_agent_tool_list
    + agentql_browser_tool.to_tool_list()
    + duckduckgo_search_tool,
    llm=llm,
    system_prompt="You are an expert that can do browser automation, data extraction and text summarization for finding and extracting data from research resources.",
)

from llama_index.llms.openai import OpenAI from llama_index.core.agent.workflow import AgentWorkflow llm = OpenAI(model="gpt-4o") workflow = AgentWorkflow.from_tools_or_functions( playwright_agent_tool_list + agentql_browser_tool.to_tool_list() + duckduckgo_search_tool, llm=llm, system_prompt="You are an expert that can do browser automation, data extraction and text summarization for finding and extracting data from research resources.", )

AgentWorkflow 同样支持流式传输功能，该功能通过使用工作流返回的处理器实现。如需流式传输大语言模型（LLM）的输出，可利用 AgentStream 事件来实现。

In [ ]:

from llama_index.core.agent.workflow import (
    AgentStream,
)

handler = workflow.run(
    user_msg="""
    Use DuckDuckGoSearch to find URL resources on the web that are relevant to the research topic: What is the relationship between exercise and stress levels?
    Go through each resource found. For each different resource, use Playwright to click on link to the resource, then use AgentQL to extract information, including the name of the resource, author name(s), link to the resource, publishing date, journal name, volume number, issue number, and the abstract.
    Find more resources until there are two different resources that can be successfully extracted from.
    """
)

async for event in handler.stream_events():
    if isinstance(event, AgentStream):
        print(event.delta, end="", flush=True)

from llama_index.core.agent.workflow import ( AgentStream, ) handler = workflow.run( user_msg=""" Use DuckDuckGoSearch to find URL resources on the web that are relevant to the research topic: What is the relationship between exercise and stress levels? Go through each resource found. For each different resource, use Playwright to click on link to the resource, then use AgentQL to extract information, including the name of the resource, author name(s), link to the resource, publishing date, journal name, volume number, issue number, and the abstract. Find more resources until there are two different resources that can be successfully extracted from. """ ) async for event in handler.stream_events(): if isinstance(event, AgentStream): print(event.delta, end="", flush=True)

/usr/local/lib/python3.11/dist-packages/agentql/_core/_utils.py:171: UserWarning: 🚨 The function get_data_by_prompt_experimental is experimental and may not work as expected 🚨
  warnings.warn(

I successfully extracted information from one resource. Here are the details:

- **Title**: Role of Physical Activity on Mental Health and Well-Being: A Review
- **Authors**: Aditya Mahindru, Pradeep Patil, Varun Agrawal
- **Link**: [Role of Physical Activity on Mental Health and Well-Being: A Review](https://pmc.ncbi.nlm.nih.gov/articles/PMC9902068/)
- **Publication Date**: January 7, 2023
- **Journal Name**: Cureus
- **Volume Number**: 15
- **Issue Number**: 1
- **Abstract**: The article reviews the positive effects of physical activity on mental health, highlighting its benefits on self-concept, body image, and mood. It discusses the physiological and psychological mechanisms by which exercise improves mental health, including its impact on the hypothalamus-pituitary-adrenal axis, depression, anxiety, sleep, and psychiatric disorders. The review also notes the need for more research in the Indian context.

I will now attempt to extract information from another resource.I successfully extracted information from a second resource. Here are the details:

- **Title**: The Relationship Between Exercise Habits and Stress Among Individuals With Access to Internet-Connected Home Fitness Equipment: Single-Group Prospective Analysis
- **Authors**: Margaret Schneider, Amanda Woodworth, Milad Asgari Mehrabadi
- **Link**: [The Relationship Between Exercise Habits and Stress Among Individuals With Access to Internet-Connected Home Fitness Equipment](https://pmc.ncbi.nlm.nih.gov/articles/PMC9947760/)
- **Publication Date**: February 8, 2023
- **Journal Name**: JMIR Form Res
- **Volume Number**: 7
- **Issue Number**: e41877
- **Abstract**: This study examines the relationship between stress and exercise habits among habitual exercisers with internet-connected home fitness equipment during the COVID-19 lockdown. It found that stress did not negatively impact exercise participation among habitually active adults with such equipment. The study suggests that habitual exercise may buffer the impact of stress on regular moderate to vigorous activity, and highlights the potential role of home-based internet-connected exercise equipment in this buffering.

Both resources provide valuable insights into the relationship between exercise and stress levels.