多AI Agent代理:使用LangGraph和LangChain创建多代理工作流

前言

在AI时代,每个人都是一个超级个体,AI Agent智能体的出现,为我们打造超级个体提供了可能。如果说2024年将是AI应用的元年,那么AI Agent将是这个AI应用元年里最为闪亮的那颗星。之前有读者留言,希望多分享一些AI Agent智能体的搭建方法,在上一篇推文中也从实战案例角度分享了怎么用天工AI快速搭建一套属于我们自己的AI Agent智能体,天工AI的多模态和AI搜索能力相信已经能满足大多数人的使用需求。今天就从代码实战上来分享如何使用LangGraph和LangChain创建多代理工作流。

自我反省

自我反省的AI代理使用一个生成器来产生输出,并用一个反射器来审查生成器的工作。生成器和反射器都使用相同的大型语言模型(LLM),但各自有不同的提示,从而形成了一个自我反省的AI代理。

这种在循环方式中以两种不同角色使用同一LLM的方法,是通过LangChain的LangGraph框架 facilitated 的。LangGraph 框架也可以用来创建多代理工作流。就像在自我反思的 AI 代理中一样,LLM 可以扮演多个角色,每个角色都充当一个不同的 AI 代理。这就是多代理的概念。

多代理

一个多代理系统涉及到将独立的行动者连接起来,每个行动者都由一个大型语言模型提供支持,按照特定的排列组合。

每个代理可以有自己的提示、LLM、工具和其他自定义代码来与其他代理协作。然而,同一个 LLM 也可以基于提供的提示扮演不同的角色。

使用LangGraph进行多代理工作流

LangGraph非常适合创建多代理工作流,因为它允许将两个或更多代理连接成一个图。每个代理都是一个独立的行动者,代理之间的连接由边缘表示。每个连接边缘都可以有一个控制条件,指导信息从一个代理流向另一个代理。每个代理都有一个状态,可以在每次流动过程中用信息更新。

协作多代理

就像人类合作一样,协作多代理工作流中的不同AI代理通过共享的留言板进行沟通。这使得每个代理都可以查看其他代理的工作并观察所有独立步骤。

这些不同的AI代理可能基于同一个大型语言模型,但扮演不同的角色。

我们现在将创建一个多代理工作流,用于生成过去五年马来西亚GDP的图表。此工作流要求研究人员在互联网上搜索并找出过去五年马来西亚GDP的数据,以及一个图表生成器,该生成器使用Python代码根据研究人员获得的数据生成图表。

因此,参与者包括:

1.研究人员 — 使用互联网搜索工具来研究问题。2.图表生成器 — 根据研究人员提供的数据执行代码创建图表。

图表蓝图

我们将构造一个图表蓝图,展示代理节点及其之间的连接。

图中的信息流动

A. 用户将输入发送给研究人员。

B. 研究人员将消息发送给路由器。

C. 根据消息的内容,路由器要么决定调用一个工具,要么发送消息给图表生成器。

D. 工具完成其工作并将输出发送给研究人员。研究人员可能再次调用该工具或发送消息给图表生成器。相应地,路由器调用工具或将消息发送给图表生成器。

E. 图表生成器将回复发送给路由器,路由器可能调用工具或将回复发送回研究人员。F. 最终,当研究者或图表生成器发送消息作为最终答案时,路由器将最终答案路由给用户。

现在,根据上述图表,让我们创建一个多代理工作流程。

安装包

首先,我们需要安装所有langchain包。

pip install langchain langchain_openai langgraph langchain_core langsmith langchain_experimental

我们还需要安装一些包用于存储数据(pandas)和创建图表(matplotlib)。

pip install pandas matplotlib

导入库

接下来,我们将导入创建多代理工作流所需的库。

import json``   ``from langchain_core.messages import (` `AIMessage,` `BaseMessage,` `ChatMessage,` `FunctionMessage,` `HumanMessage,``)``from langchain.tools.render import format_tool_to_openai_function``from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder``from langgraph.graph import END, StateGraph``from langgraph.prebuilt.tool_executor import ToolExecutor, ToolInvocation

创建工具

Tavily搜索工具

接下来,我们将设置工具。我们需要一个互联网搜索工具。我们将使用tavily来实现这一点。

import os``os.environ['TAVILY_API_KEY'] = "<Your Tavily API Key here>"``   ``from langchain_core.tools import tool``from langchain_community.tools.tavily_search import TavilySearchResults``tavily_tool = TavilySearchResults(max_results=5)

Python 工具

我们还需要一个工具来执行 Python 代码。我们将使用来自 langchain_experimental.utilities 的一个名为 PythonREPL 的工具。

from langchain_experimental.utilities import PythonREPL``from typing import Annotated``   ``repl = PythonREPL()``@tool``def python_repl(code: Annotated[str, "The python code to execute to generate your chart."]):` ``"""Use this to execute python code. If you want to see the output of a value, you should print it out with `print(...)`. This is visible to the user."""`` `try:` `result = repl.run(code)` `except BaseException as e:` `return f"Failed to execute. Error: {repr(e)}"` ``return f"Succesfully executed:\\\\n`python\\\\\\\\n{code}\\\\\\\\n`\\\\nStdout: {result}"`

我们将工具列入一个列表中。

tools = [tavily_tool, python_repl]

创建状态

接下来,我们将为状态创建对象。状态将是从一个节点传递到另一个节点的消息列表。

import operator``from typing import Annotated, List, Sequence, Tuple, TypedDict, Union``   ``from langchain.agents import create_openai_functions_agent``from langchain.tools.render import format_tool_to_openai_function``from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholderfrom langchain_openai import ChatOpenAI``from typing_extensions import TypedDict# This defines the object that is passed between each node

在图中。我们将为每个代理和工具创建不同的节点

class AgentState(TypedDict):` `messages: Annotated[Sequence[BaseMessage], operator.add]` `sender: str

创建工具节点

我们将根据消息的内容使用每个工具。我们将创建一个 tool_node 函数,该函数将根据消息执行一个工具。

tool_executor = ToolExecutor(tools)``   ``def tool_node(state):` `"""This runs tools in the graph It takes in an agent action and calls that tool and returns the result."""` `messages = state["messages"]` `# Based on the continue condition` `# we know the last message involves a function call` `last_message = messages[-1]` `# We construct an ToolInvocation from the function_call` `tool_input = json.loads(` `last_message.additional_kwargs["function_call"]["arguments"]` `)` `# We can pass single-arg inputs by value` `if len(tool_input) == 1 and "__arg1" in tool_input:` `tool_input = next(iter(tool_input.values()))` `tool_name = last_message.additional_kwargs["function_call"]["name"]` `action = ToolInvocation(` `tool=tool_name,` `tool_input=tool_input,` `)` `# We call the tool_executor and get back a response` `response = tool_executor.invoke(action)` `# We use the response to create a FunctionMessage` `function_message = FunctionMessage(` `content=f"{tool_name} response: {str(response)}", name=action.tool` `)` `# We return a list, because this will get added to the existing list` `return {"messages": [function_message]}

创建路由器

现在我们已经设置好了工具节点,接下来让我们创建一个路由器,它将根据来自状态的消息内容决定将信息发送到哪里。

任一代理都可以决定结束

def router(state):` `# This is the router` `messages = state["messages"]` `last_message = messages[-1]` `if "function_call" in last_message.additional_kwargs:` `# The previus agent is invoking a tool` `return "call_tool"` `if "FINAL ANSWER" in last_message.content:` `# Any agent decided the work is done` `return "end"` `return "continue"

接下来,我们要设置AI代理研究员和图表生成器。

创建LLM

首先,让我们创建llm。

from langchain_openai import ChatOpenAI``llm = ChatOpenAI(api_key="<Your API Key here>")

创建代理

首先,我们创建一个 create_agent 函数,该函数将创建一个代理并追加函数提供的系统消息。我们稍后将使用此函数来创建研究员和图表生成器。

import json``   ``from langchain_core.messages import (` `AIMessage,` `BaseMessage,` `ChatMessage,` `FunctionMessage,` `HumanMessage,``)``from langchain_core.utils.function_calling import convert_to_openai_function``from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder``from langgraph.graph import END, StateGraph``from langgraph.prebuilt.tool_executor import ToolExecutor, ToolInvocationdef create_agent(llm, tools, system_message: str):` `"""Create an agent."""` `functions = [convert_to_openai_function(t) for t in tools] prompt = ChatPromptTemplate.from_messages(` `[` `(` `"system",` `"You are a helpful AI assistant, collaborating with other assistants."` `" Use the provided tools to progress towards answering the question."` `" If you are unable to fully answer, that's OK, another assistant with different tools "` `" will help where you left off. Execute what you can to make progress."` `" If you or any of the other assistants have the final answer or deliverable,"` `" prefix your response with FINAL ANSWER so the team knows to stop."` `" You have access to the following tools: {tool_names}.\\\\n{system_message}",` `),` `MessagesPlaceholder(variable_name="messages"),` `]` `)` `prompt = prompt.partial(system_message=system_message)` `prompt = prompt.partial(tool_names=", ".join([tool.name for tool in tools]))` `return prompt | llm.bind_functions(functions)

接下来,我们创建一个名为agent_node的函数,该函数负责添加消息和发送者的名称。这将用于路由目的。

创建给定代理的节点的辅助函数

def agent_node(state, agent, name):` `result = agent.invoke(state)` `# We convert the agent output into a format that is suitable to append to the global state` `if isinstance(result, FunctionMessage):` `pass` `else:` `result = HumanMessage(**result.dict(exclude={"type", "name"}), name=name)` `return {` `"messages": [result],` `# Since we have a strict workflow, we can` `# track the sender so we know who to pass to next.` `"sender": name,` `}``   ``Now, we create the researcher agent and the chart generator agent.``   ``research_agent= create_agent(` `llm,` `[tavily_tool],` `system_message="You should provide accurate data for the chart generator to use.",``)``   ``chart_agent= create_agent(` `llm,` `[python_repl],` `system_message="Any charts you display will be visible by the user.",``)

从代理节点创建研究者节点和图表节点

然后,我们从 research_agent 和 chart_agent 函数创建 research_node 函数和 chart_node 函数。我们使用 functools 库从另一个函数 agent_node 创建函数。

import functools``research_node= functools.partial(agent_node, agent=research_agent, name="Researcher")``   ``chart_node= functools.partial(agent_node, agent=chart_agent, name="Chart Generator")

创建图表

我们现在可以将所有节点放入名为workflow的图中,这包括research_node(研究节点)、chart_node(图表节点)和tool_node(工具节点)。我们还添加了条件边缘,这些边缘将根据满足的条件路由消息。最后,我们将研究者节点设置为入口点。因此,用户输入将首先进入研究者节点。最后,我们编译图表。

workflow= StateGraph(AgentState)``   ``workflow.add_node("Researcher", research_node)``workflow.add_node("Chart Generator", chart_node)``workflow.add_node("call_tool", tool_node)workflow.add_conditional_edges(` `"Researcher",` `router,` `{"continue": "Chart Generator", "call_tool": "call_tool", "end": END},``)``workflow.add_conditional_edges(` `"Chart Generator",` `router,` `{"continue": "Researcher", "call_tool": "call_tool", "end": END},``)workflow.add_conditional_edges(` `"call_tool",#每个代理节点更新'sender'字段``#调用工具的节点不会更新,意味着``#这条边将回路到最初调用工具的代理节点`  `lambda x: x["sender"],``{`    `"Researcher": "Researcher",`    `"Chart Generator": "Chart Generator",``},``)``workflow.set_entry_point("Researcher")``graph= workflow.compile()

调用多代理工作流程

我们现在准备好调用多代理工作流程了。

for s in graph.stream(` `{` `"messages": [` `HumanMessage(` `content="Fetch the Malaysia's GDP over the past 5 years,"` `" then draw a line graph of it."` `" Once you code it up, finish."` `)` `],` `},` `# Maximum number of steps to take in the graph` `{"recursion_limit": 150},``):` `print(s)` `print("----")

输出为:

{'Researcher': {'messages': [HumanMessage(content='', additional_kwargs={'function_call': {'arguments': '{"query":"Malaysia GDP by year"}', 'name': 'tavily_search_results_json'}}, response_metadata={'token_usage': {'completion_tokens': 22, 'prompt_tokens': 221, 'total_tokens': 243}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'function_call', 'logprobs': None}, name='Researcher', id='run-313ac64c-fed0-4505-b00d-715c7ff0aa59-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'call_tool': {'messages': [FunctionMessage(content="tavily_search_results_json response: [{'url': '<https://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG?locations=MY>', 'content': 'GDP growth (annual %) - Malaysia from The World Bank: Data.'}, {'url': '<https://tradingeconomics.com/malaysia/gdp>', 'content': 'GDP in Malaysia averaged 109.79 USD Billion from 1960 until 2022, reaching an all time high of 407.03 USD Billion in 2022 and a record low of 1.90 USD Billion\\\\xa0...'}, {'url': '<https://www.macrotrends.net/global-metrics/countries/MYS/malaysia/gdp-gross-domestic-product>', 'content': 'Malaysia gdp for 2019 was $365.18B, a 1.78% increase from 2018. Download Historical Data Save as Image\\\\xa0...'}, {'url': '<https://data.worldbank.org/indicator/NY.GDP.MKTP.CD?locations=MY>', 'content': '1960 1980 2000 2020 0.00 225b 450b Malaysia. 1960 - 2022. 2022, 2021, 2020, 2019, 2018, 2017, 2016, 2015, 2014, 2013, 2012, 2011, 2010, 2009, 2008, 2007\\\\xa0...'}, {'url': '<https://www.worldometers.info/gdp/malaysia-gdp/>', 'content': 'Nominal (current) Gross Domestic Product (GDP) of Malaysia is $406,306,000,000 (USD) as of 2022. · Real GDP (constant, inflation adjusted) of Malaysia reached\\\\xa0...'}]", name='tavily_search_results_json')]}}``   ``{'Researcher': {'messages': [HumanMessage(content='', additional_kwargs={'function_call': {'arguments': '{"query":"Malaysia GDP historical data"}', 'name': 'tavily_search_results_json'}}, response_metadata={'token_usage': {'completion_tokens': 22, 'prompt_tokens': 632, 'total_tokens': 654}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'function_call', 'logprobs': None}, name='Researcher', id='run-442c7d99-8b3c-4a32-99ec-e0739ac059dd-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'call_tool': {'messages': [FunctionMessage(content='tavily_search_results_json response: [{\\'url\\': \\'<https://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG?locations=MY&view=chart\\>', \\'content\\': \\'GDP growth (annual %) - Malaysia. World Bank national accounts data, and OECD National Accounts data files. License : CC BY-4.0. LineBarMap. Also Show Share Details. Label. 1961 - 2022.\\'}, {\\'url\\': \\'<https://tradingeconomics.com/malaysia/gdp\\>', \\'content\\': \\'The Gross Domestic Product (GDP) in Malaysia was worth 407.03 billion US dollars in 2022, according to official data from the World Bank. The GDP value of Malaysia represents 0.17 percent of the world economy. This page provides - Malaysia GDP - actual values, historical data, forecast, chart, statistics, economic calendar and news.\\'}, {\\'url\\': \\'<https://www.macrotrends.net/global-metrics/countries/MYS/malaysia/gdp-growth-rate\\>', \\'content\\': \\'Malaysia gdp growth rate for 2022 was 8.65%, a 5.35% increase from 2021. Malaysia gdp growth rate for 2021 was 3.30%, a 8.75% increase from 2020. Malaysia gdp growth rate for 2020 was -5.46%, a 9.87% decline from 2019. Malaysia gdp growth rate for 2019 was 4.41%, a 0.43% decline from 2018. Download Historical Data. Save as Image.\\'}, {\\'url\\': \\'<https://data.worldbank.org/indicator/NY.GDP.MKTP.CD?locations=MY&view=chart\\>', \\'content\\': \\'All Countries and Economies. Country. Most Recent Year. Most Recent Value(Millions) GDP (current US$) - Malaysia from The World Bank: Data.\\'}, {\\'url\\': \\'<https://www.macrotrends.net/global-metrics/countries/MYS/malaysia/gdp-gross-domestic-product\\>', \\'content\\': "Malaysia gdp for 2022 was $407.03B, a 8.88% increase from 2021. Malaysia gdp for 2021 was $373.83B, a 10.78% increase from 2020. Malaysia gdp for 2020 was $337.46B, a 7.59% decline from 2019. Malaysia gdp for 2019 was $365.18B, a 1.78% increase from 2018. GDP at purchaser\\'s prices is the sum of gross value added by all resident producers in the ..."}]', name='tavily_search_results_json')]}}``   ``{'Researcher': {'messages': [HumanMessage(content="Based on the search results, I found the GDP data for Malaysia for the past 5 years as follows:\\n\\n- 2018: $365.18 billion\\n- 2019: $337.46 billion\\n- 2020: $373.83 billion\\n- 2021: $407.03 billion\\n- 2022: $407.03 billion\\n\\nNow, I will proceed to create a line graph to visualize this data.\\nI will now generate a line graph to visualize Malaysia's GDP over the past 5 years.", response_metadata={'token_usage': {'completion_tokens': 113, 'prompt_tokens': 1210, 'total_tokens': 1323}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Researcher', id='run-e498d73b-c688-4e38-aa9b-b4cfa39c34c7-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'Chart Generator': {'messages': [HumanMessage(content='', additional_kwargs={'function_call': {'arguments': '{"code":"import matplotlib.pyplot as plt\\\\n\\\\ndata = {\\\\n \\'Year\\': [2018, 2019, 2020, 2021, 2022],\\\\n \\'GDP (in billion USD)\\': [365.18, 337.46, 373.83, 407.03, 407.03]\\\\n}\\\\n\\\\nplt.figure(figsize=(10, 6))\\\\nplt.plot(data[\\'Year\\'], data[\\'GDP (in billion USD)\\'], marker=\\'o\\', color=\\'b\\', linestyle=\\'-\\', linewidth=2)\\\\nplt.title(\\'Malaysia GDP Over the Past 5 Years\\')\\\\nplt.xlabel(\\'Year\\')\\\\nplt.ylabel(\\'GDP (in billion USD)\\')\\\\nplt.grid(True)\\\\nplt.show()"}', 'name': 'python_repl'}}, response_metadata={'token_usage': {'completion_tokens': 175, 'prompt_tokens': 1338, 'total_tokens': 1513}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'function_call', 'logprobs': None}, name='Chart Generator', id='run-11c4756f-097b-4733-8ecf-6091f5be5bd4-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Chart Generator'}}``   ```{'call_tool': {'messages': [FunctionMessage(content="python_repl response: Succesfully executed:\\\\n`python\\\\\\\\nimport matplotlib.pyplot as plt\\n\\ndata = {\\n 'Year': [2018, 2019, 2020, 2021, 2022],\\n 'GDP (in billion USD)': [365.18, 337.46, 373.83, 407.03, 407.03]\\n}\\n\\nplt.figure(figsize=(10, 6))\\nplt.plot(data['Year'], data['GDP (in billion USD)'], marker='o', color='b', linestyle='-', linewidth=2)\\nplt.title('Malaysia GDP Over the Past 5 Years')\\nplt.xlabel('Year')\\nplt.ylabel('GDP (in billion USD)')\\nplt.grid(True)\\nplt.show()\\\\\\\\n`\\\\nStdout: ", name='python_repl')]}}`````{'Chart Generator': {'messages': [HumanMessage(content="The line graph showing Malaysia's GDP over the past 5 years has been successfully generated. The graph displays the GDP (in billion USD) on the y-axis and the years from 2018 to 2022 on the x-axis. The GDP values are plotted as points connected by lines.\\n\\nIs there anything else you would like to analyze or any other assistance needed?", response_metadata={'token_usage': {'completion_tokens': 75, 'prompt_tokens': 1514, 'total_tokens': 1589}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Chart Generator', id='run-a6d15c89-6cb8-4f98-9cbb-67bb92214dc5-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Chart Generator'}}``   ``{'Researcher': {'messages': [HumanMessage(content="I have successfully generated the line graph showing Malaysia's GDP over the past 5 years. If you have any more data to analyze or need further assistance, feel free to let me know!", response_metadata={'token_usage': {'completion_tokens': 39, 'prompt_tokens': 1579, 'total_tokens': 1618}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Researcher', id='run-9c1709e6-f6ee-4a8c-b0ae-f363d65a5a2f-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'Chart Generator': {'messages': [HumanMessage(content="Great! If you have any more data to analyze or any other tasks you'd like assistance with, feel free to ask. I'm here to help!", response_metadata={'token_usage': {'completion_tokens': 32, 'prompt_tokens': 1634, 'total_tokens': 1666}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Chart Generator', id='run-1f869abb-930c-4b19-ab7e-c068a407f134-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Chart Generator'}}``   ``{'Researcher': {'messages': [HumanMessage(content="I'm glad the line graph for Malaysia's GDP over the past 5 years was successfully generated. If you have any more data to analyze or any other tasks you'd like assistance with, feel free to ask. I'm here to help!", response_metadata={'token_usage': {'completion_tokens': 50, 'prompt_tokens': 1656, 'total_tokens': 1706}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Researcher', id='run-4c614614-5fc6-4d6b-abd8-5b1b85607f5a-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'Chart Generator': {'messages': [HumanMessage(content='Thank you! If you have any more data to analyze or if you need assistance with any other tasks in the future, feel free to reach out. Have a great day!', response_metadata={'token_usage': {'completion_tokens': 36, 'prompt_tokens': 1722, 'total_tokens': 1758}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Chart Generator', id='run-659e7259-1906-4c07-8bb5-53c8d6ac7dd2-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Chart Generator'}}``   ``{'Researcher': {'messages': [HumanMessage(content='Thank you! If you have any more data to analyze or if you need assistance with any other tasks in the future, feel free to reach out. Have a great day!', response_metadata={'token_usage': {'completion_tokens': 36, 'prompt_tokens': 1748, 'total_tokens': 1784}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Researcher', id='run-f87114d6-08aa-4a54-8490-1a1145a30263-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'Chart Generator': {'messages': [HumanMessage(content='Great collaboration! If you have any more questions or need assistance in the future, feel free to ask. Have a wonderful day!', response_metadata={'token_usage': {'completion_tokens': 27, 'prompt_tokens': 1800, 'total_tokens': 1827}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Chart Generator', id='run-b12ec2b6-61bb-485c-81cf-279ade302e02-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Chart Generator'}}``   ``{'Researcher': {'messages': [HumanMessage(content="Thank you! It was a pleasure collaborating with you. If you have any more questions or need assistance in the future, don't hesitate to ask. Have a wonderful day!", response_metadata={'token_usage': {'completion_tokens': 36, 'prompt_tokens': 1817, 'total_tokens': 1853}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Researcher', id='run-3c7cf207-ce49-416f-bb02-81a2a3410d5a-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'Chart Generator': {'messages': [HumanMessage(content='It was a pleasure collaborating with you! If you have any more questions or need assistance in the future, feel free to reach out. Have a great day!', response_metadata={'token_usage': {'completion_tokens': 33, 'prompt_tokens': 1869, 'total_tokens': 1902}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Chart Generator', id='run-1d5d0ed0-c20a-4745-a16f-6adcaa8c2c81-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Chart Generator'}}``   ``{'Researcher': {'messages': [HumanMessage(content='Likewise! If you ever have more questions or need help in the future, feel free to ask. Have a fantastic day!', response_metadata={'token_usage': {'completion_tokens': 27, 'prompt_tokens': 1892, 'total_tokens': 1919}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Researcher', id='run-bb88218f-ab2f-4ee5-b381-d75e1bd1d1cf-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'Chart Generator': {'messages': [HumanMessage(content='Likewise! If you ever have more questions or need help in the future, feel free to ask. Have a fantastic day!', response_metadata={'token_usage': {'completion_tokens': 27, 'prompt_tokens': 1935, 'total_tokens': 1962}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Chart Generator', id='run-50293d69-34ff-455d-8b34-3a20a0a1bb91-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Chart Generator'}}``   ``{'Researcher': {'messages': [HumanMessage(content="I'm glad we were able to collaborate effectively! If you have any more questions or need assistance in the future, feel free to reach out. Have a wonderful day!", response_metadata={'token_usage': {'completion_tokens': 35, 'prompt_tokens': 1952, 'total_tokens': 1987}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Researcher', id='run-67208ed1-0d31-4b43-b2e8-92619b806deb-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'Researcher': {'messages': [HumanMessage(content="Thank you! It was a pleasure collaborating with you. If you have any more questions or need assistance in the future, don't hesitate to ask. Have a wonderful day!", response_metadata={'token_usage': {'completion_tokens': 36, 'prompt_tokens': 2087, 'total_tokens': 2123}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Researcher', id='run-bc8186f2-1f7e-477e-b1ff-0020c7171aa3-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Researcher'}}``   ``{'Chart Generator': {'messages': [HumanMessage(content='FINAL ANSWER', response_metadata={'token_usage': {'completion_tokens': 4, 'prompt_tokens': 2139, 'total_tokens': 2143}, 'model_name': 'gpt-3.5-turbo', 'system_fingerprint': 'fp_3b956da36b', 'finish_reason': 'stop', 'logprobs': None}, name='Chart Generator', id='run-45177a17-7ae8-43b0-b73e-f9d3d5c5d456-0', tool_calls=[], invalid_tool_calls=[])], 'sender': 'Chart Generator'}

恭喜您!您已经创建了一个多代理工作流。现在,您正朝着创建多个AI代理共同完成复杂任务的方向前进。

如何学习大模型 AI ?

由于新岗位的生产效率,要优于被取代岗位的生产效率,所以实际上整个社会的生产效率是提升的。

但是具体到个人,只能说是:

“最先掌握AI的人,将会比较晚掌握AI的人有竞争优势”。

这句话,放在计算机、互联网、移动互联网的开局时期,都是一样的道理。

我在一线互联网企业工作十余年里,指导过不少同行后辈。帮助很多人得到了学习和成长。

我意识到有很多经验和知识值得分享给大家,也可以通过我们的能力和经验解答大家在人工智能学习中的很多困惑,所以在工作繁忙的情况下还是坚持各种整理和分享。但苦于知识传播途径有限,很多互联网行业朋友无法获得正确的资料得到学习提升,故此将并将重要的AI大模型资料包括AI大模型入门学习思维导图、精品AI大模型学习书籍手册、视频教程、实战学习等录播视频免费分享出来。

在这里插入图片描述

第一阶段(10天):初阶应用

该阶段让大家对大模型 AI有一个最前沿的认识,对大模型 AI 的理解超过 95% 的人,可以在相关讨论时发表高级、不跟风、又接地气的见解,别人只会和 AI 聊天,而你能调教 AI,并能用代码将大模型和业务衔接。

  • 大模型 AI 能干什么?
  • 大模型是怎样获得「智能」的?
  • 用好 AI 的核心心法
  • 大模型应用业务架构
  • 大模型应用技术架构
  • 代码示例:向 GPT-3.5 灌入新知识
  • 提示工程的意义和核心思想
  • Prompt 典型构成
  • 指令调优方法论
  • 思维链和思维树
  • Prompt 攻击和防范

第二阶段(30天):高阶应用

该阶段我们正式进入大模型 AI 进阶实战学习,学会构造私有知识库,扩展 AI 的能力。快速开发一个完整的基于 agent 对话机器人。掌握功能最强的大模型开发框架,抓住最新的技术进展,适合 Python 和 JavaScript 程序员。

  • 为什么要做 RAG
  • 搭建一个简单的 ChatPDF
  • 检索的基础概念
  • 什么是向量表示(Embeddings)
  • 向量数据库与向量检索
  • 基于向量检索的 RAG
  • 搭建 RAG 系统的扩展知识
  • 混合检索与 RAG-Fusion 简介
  • 向量模型本地部署

第三阶段(30天):模型训练

恭喜你,如果学到这里,你基本可以找到一份大模型 AI相关的工作,自己也能训练 GPT 了!通过微调,训练自己的垂直大模型,能独立训练开源多模态大模型,掌握更多技术方案。

到此为止,大概2个月的时间。你已经成为了一名“AI小子”。那么你还想往下探索吗?

  • 为什么要做 RAG
  • 什么是模型
  • 什么是模型训练
  • 求解器 & 损失函数简介
  • 小实验2:手写一个简单的神经网络并训练它
  • 什么是训练/预训练/微调/轻量化微调
  • Transformer结构简介
  • 轻量化微调
  • 实验数据集的构建

第四阶段(20天):商业闭环

对全球大模型从性能、吞吐量、成本等方面有一定的认知,可以在云端和本地等多种环境下部署大模型,找到适合自己的项目/创业方向,做一名被 AI 武装的产品经理。

  • 硬件选型
  • 带你了解全球大模型
  • 使用国产大模型服务
  • 搭建 OpenAI 代理
  • 热身:基于阿里云 PAI 部署 Stable Diffusion
  • 在本地计算机运行大模型
  • 大模型的私有化部署
  • 基于 vLLM 部署大模型
  • 案例:如何优雅地在阿里云私有部署开源大模型
  • 部署一套开源 LLM 项目
  • 内容安全
  • 互联网信息服务算法备案

学习是一个过程,只要学习就会有挑战。天道酬勤,你越努力,就会成为越优秀的自己。

如果你能在15天内完成所有的任务,那你堪称天才。然而,如果你能完成 60-70% 的内容,你就已经开始具备成为一名大模型 AI 的正确特征了。

这份完整版的大模型 AI 学习资料已经上传CSDN,朋友们如果需要可以微信扫描下方CSDN官方认证二维码免费领取【保证100%免费

在这里插入图片描述

  • 9
    点赞
  • 21
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值