AutoGen实现多代理-Tool_Use_and_Conversational_Chess（四）

于 2024-09-30 09:30:00 发布
阅读量264 收藏 3
点赞数 7
分类专栏： AIGC 文章标签： 人工智能 多代理 语言模型 多代理交互
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/l8947943/article/details/142645260
版权

1. 学习背景 在这里插入图片描述

如图，常见的Agent之间交流对话，可能会涉及到外部工具的调用和嵌套式聊天，这种设计模式就要求代理会使用工具和按序执行代码。本节尝试聊天机器人模拟人类进行下棋，如图所示，Chess Board代理则进行检测合法性并记录移动。
在这里插入图片描述

2. 代码实践

本节实验链接：传送门

2.1 准备环境

llm_config = {"model": "gpt-4-turbo"}

# 导入与棋盘相关的包
import chess
import chess.svg
from typing_extensions import Annotated

# 初始化棋盘
board = chess.Board()
made_move = False

2.2 定义所需的工具

2.2.1 获取合法移动的工具

def get_legal_moves(
    
) -> Annotated[str, "A list of legal moves in UCI format"]:
    return "Possible moves are: " + ",".join(
        [str(move) for move in board.legal_moves]
    )

2.2.2 在棋盘上进行移动的工具

def make_move(
    move: Annotated[str, "A move in UCI format."]
) -> Annotated[str, "Result of the move."]:
    move = chess.Move.from_uci(move)
    board.push_uci(str(move))
    global made_move
    made_move = True
    
    # Display the board.
    display(
        chess.svg.board(
            board,
            arrows=[(move.from_square, move.to_square)],
            fill={move.from_square: "gray"},
            size=200
        )
    )
    
    # Get the piece name.
    piece = board.piece_at(move.to_square)
    piece_symbol = piece.unicode_symbol()
    piece_name = (
        chess.piece_name(piece.piece_type).capitalize()
        if piece_symbol.isupper()
        else chess.piece_name(piece.piece_type)
    )
    return f"Moved {piece_name} ({piece_symbol}) from "\
    f"{chess.SQUARE_NAMES[move.from_square]} to "\
    f"{chess.SQUARE_NAMES[move.to_square]}."

2.3 创建代理

你将为棋盘创建玩家代理和棋盘代理。

from autogen import ConversableAgent

# Player white agent
player_white = ConversableAgent(
    name="Player White",
    system_message="You are a chess player and you play as white. "
    "First call get_legal_moves(), to get a list of legal moves. "
    "Then call make_move(move) to make a move.",
    llm_config=llm_config,
)

# Player black agent
player_black = ConversableAgent(
    name="Player Black",
    system_message="You are a chess player and you play as black. "
    "First call get_legal_moves(), to get a list of legal moves. "
    "Then call make_move(move) to make a move.",
    llm_config=llm_config,
)

def check_made_move(msg):
    global made_move
    if made_move:
        made_move = False
        return True
    else:
        return False

board_proxy = ConversableAgent(
    name="Board Proxy",
    llm_config=False,
    is_termination_msg=check_made_move,
    default_auto_reply="Please make a move.",
    human_input_mode="NEVER",
)

2.4 为代理注册可用的工具

调用工具的代理和执行工具的代理都必须为其注册一个工具。

from autogen import register_function

# 为每个代理都注册合法性检测工具和移动工具，方便代理调用
for caller in [player_white, player_black]:
    register_function(
        get_legal_moves,
        caller=caller,
        executor=board_proxy,
        name="get_legal_moves",
        description="Get legal moves.",
    )
    
    register_function(
        make_move,
        caller=caller,
        executor=board_proxy,
        name="make_move",
        description="Call this tool to make a move.",
    )

可以查看代理绑定工具的情况

player_black.llm_config["tools"]

输出如下：

[{'type': 'function',
  'function': {'description': 'Get legal moves.',
   'name': 'get_legal_moves',
   'parameters': {'type': 'object', 'properties': {}, 'required': []}}},
 {'type': 'function',
  'function': {'description': 'Call this tool to make a move.',
   'name': 'make_move',
   'parameters': {'type': 'object',
    'properties': {'move': {'type': 'string',
      'description': 'A move in UCI format.'}},
    'required': ['move']}}}]

2.5 注册嵌套式聊天模式

每个玩家代理将与棋盘代理进行嵌套式聊天，以便在棋盘上进行移动。

player_white.register_nested_chats(
    trigger=player_black,
    chat_queue=[
        {
            "sender": board_proxy,
            "recipient": player_white,
            "summary_method": "last_msg",
        }
    ],
)

player_black.register_nested_chats(
    trigger=player_white,
    chat_queue=[
        {
            "sender": board_proxy,
            "recipient": player_black,
            "summary_method": "last_msg",
        }
    ],
)

2.6 开始游戏

board = chess.Board()

chat_result = player_black.initiate_chat(
    player_white,
    message="Let's play chess! Your move.",
    max_turns=2,
)

输出如下：

Player Black (to Player White):

Let's play chess! Your move.

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...

********************************************************************************
Starting a new chat....

********************************************************************************
Board Proxy (to Player White):

Let's play chess! Your move.

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player White (to Board Proxy):

***** Suggested tool call (call_Xp271IKAeUo1TxQvoHZ4dWae): get_legal_moves *****
Arguments: 
{}
********************************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION get_legal_moves...
Board Proxy (to Player White):

Board Proxy (to Player White):

***** Response from calling tool (call_Xp271IKAeUo1TxQvoHZ4dWae) *****
Possible moves are: g1h3,g1f3,b1c3,b1a3,h2h3,g2g3,f2f3,e2e3,d2d3,c2c3,b2b3,a2a3,h2h4,g2g4,f2f4,e2e4,d2d4,c2c4,b2b4,a2a4
**********************************************************************

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player White (to Board Proxy):

***** Suggested tool call (call_XanM9YQcrNYeqGH23xy5Mphx): make_move *****
Arguments: 
{"move":"e2e4"}
**************************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION make_move...
Board Proxy (to Player White):

Board Proxy (to Player White):

***** Response from calling tool (call_XanM9YQcrNYeqGH23xy5Mphx) *****
Moved pawn (♙) from e2 to e4.
**********************************************************************

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player White (to Board Proxy):

I've moved my pawn from e2 to e4. Your turn!

--------------------------------------------------------------------------------
Player White (to Player Black):

I've moved my pawn from e2 to e4. Your turn!

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...

********************************************************************************
Starting a new chat....

********************************************************************************
Board Proxy (to Player Black):

I've moved my pawn from e2 to e4. Your turn!

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player Black (to Board Proxy):

***** Suggested tool call (call_iNPpTbzXbXjXvkrBir5eMoXg): get_legal_moves *****
Arguments: 
{}
********************************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION get_legal_moves...
Board Proxy (to Player Black):

Board Proxy (to Player Black):

***** Response from calling tool (call_iNPpTbzXbXjXvkrBir5eMoXg) *****
Possible moves are: g8h6,g8f6,b8c6,b8a6,h7h6,g7g6,f7f6,e7e6,d7d6,c7c6,b7b6,a7a6,h7h5,g7g5,f7f5,e7e5,d7d5,c7c5,b7b5,a7a5
**********************************************************************

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player Black (to Board Proxy):

I'll move my pawn from e7 to e5.

Let's execute that move.
***** Suggested tool call (call_OeCdklYTQCRCf3zbDTN4Iyhh): make_move *****
Arguments: 
{"move":"e7e5"}
**************************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION make_move...
Board Proxy (to Player Black):

Board Proxy (to Player Black):

***** Response from calling tool (call_OeCdklYTQCRCf3zbDTN4Iyhh) *****
Moved pawn (♟) from e7 to e5.
**********************************************************************

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player Black (to Board Proxy):

I moved my pawn from e7 to e5. It's your turn!

--------------------------------------------------------------------------------
Player Black (to Player White):

I moved my pawn from e7 to e5. It's your turn!

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...

********************************************************************************
Starting a new chat....

********************************************************************************
Board Proxy (to Player White):

I moved my pawn from e7 to e5. It's your turn!

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player White (to Board Proxy):

***** Suggested tool call (call_OIZtgdPww7usovnyFNVQuaf2): get_legal_moves *****
Arguments: 
{}
********************************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION get_legal_moves...
Board Proxy (to Player White):

Board Proxy (to Player White):

***** Response from calling tool (call_OIZtgdPww7usovnyFNVQuaf2) *****
Possible moves are: g1h3,g1f3,g1e2,f1a6,f1b5,f1c4,f1d3,f1e2,e1e2,d1h5,d1g4,d1f3,d1e2,b1c3,b1a3,h2h3,g2g3,f2f3,d2d3,c2c3,b2b3,a2a3,h2h4,g2g4,f2f4,d2d4,c2c4,b2b4,a2a4
**********************************************************************

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player White (to Board Proxy):

***** Suggested tool call (call_o8palhCreQqCGalX2GDOezvj): make_move *****
Arguments: 
{"move":"e2e4"}
**************************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION make_move...
Board Proxy (to Player White):

Board Proxy (to Player White):

***** Response from calling tool (call_o8palhCreQqCGalX2GDOezvj) *****
Error: illegal uci: 'e2e4' in rnbqkbnr/pppp1ppp/8/4p3/4P3/8/PPPP1PPP/RNBQKBNR w KQkq - 0 2
**********************************************************************

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player White (to Board Proxy):

It seems there was a misunderstanding; your move, pawn from e7 to e5, has already been reflected on the board. I'll proceed with my move now. Let me make another valid move.
***** Suggested tool call (call_IOkzRwgZ2V61IJeprXUGSbyk): make_move *****
Arguments: 
{"move":"g1f3"}
**************************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION make_move...
Board Proxy (to Player White):

Board Proxy (to Player White):

***** Response from calling tool (call_IOkzRwgZ2V61IJeprXUGSbyk) *****
Moved knight (♘) from g1 to f3.
**********************************************************************

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
Player White (to Board Proxy):

I've moved my knight from g1 to f3. It's your turn!

--------------------------------------------------------------------------------
Player White (to Player Black):

I've moved my knight from g1 to f3. It's your turn!

--------------------------------------------------------------------------------

jupyternotebook执行过程有图，因博客原因，图片复制不出来，见谅。

2.7 为游戏添加有趣的闲聊！

player_white = ConversableAgent(
    name="Player White",
    system_message="You are a chess player and you play as white. "
    "First call get_legal_moves(), to get a list of legal moves. "
    "Then call make_move(move) to make a move. "
    "After a move is made, chitchat to make the game fun.",
    llm_config=llm_config,
)

player_black = ConversableAgent(
    name="Player Black",
    system_message="You are a chess player and you play as black. "
    "First call get_legal_moves(), to get a list of legal moves. "
    "Then call make_move(move) to make a move. "
    "After a move is made, chitchat to make the game fun.",
    llm_config=llm_config,
)

for caller in [player_white, player_black]:
    register_function(
        get_legal_moves,
        caller=caller,
        executor=board_proxy,
        name="get_legal_moves",
        description="Get legal moves.",
    )

    register_function(
        make_move,
        caller=caller,
        executor=board_proxy,
        name="make_move",
        description="Call this tool to make a move.",
    )

player_white.register_nested_chats(
    trigger=player_black,
    chat_queue=[
        {
            "sender": board_proxy,
            "recipient": player_white,
            "summary_method": "last_msg",
            "silent": True,
        }
    ],
)

player_black.register_nested_chats(
    trigger=player_white,
    chat_queue=[
        {
            "sender": board_proxy,
            "recipient": player_black,
            "summary_method": "last_msg",
            "silent": True,
        }
    ],
)

board = chess.Board()

chat_result = player_black.initiate_chat(
    player_white,
    message="Let's play chess! Your move.",
    max_turns=2,
)

输出如下：

Player Black (to Player White):

Let's play chess! Your move.

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...

********************************************************************************
Starting a new chat....

********************************************************************************

>>>>>>>> USING AUTO REPLY...

>>>>>>>> EXECUTING FUNCTION get_legal_moves...

>>>>>>>> USING AUTO REPLY...

>>>>>>>> EXECUTING FUNCTION make_move...

>>>>>>>> USING AUTO REPLY...
Player White (to Player Black):

Great! I've moved my pawn to e4. It's your turn now. How do you like to open your games when you play chess? Do you prefer more aggressive openings or a solid, defensive setup?

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...

********************************************************************************
Starting a new chat....

********************************************************************************

>>>>>>>> USING AUTO REPLY...

>>>>>>>> EXECUTING FUNCTION get_legal_moves...

>>>>>>>> USING AUTO REPLY...

>>>>>>>> EXECUTING FUNCTION make_move...

>>>>>>>> USING AUTO REPLY...
Player Black (to Player White):

The pawn has moved to e5, and now I'm looking forward to seeing your next strategy. It seems like we are gearing up for a classical open game!

Tell me about your chess experiences. Have you had any memorable games or achievements that stand out in your journey as a chess player?

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...

********************************************************************************
Starting a new chat....

********************************************************************************

>>>>>>>> USING AUTO REPLY...

>>>>>>>> EXECUTING FUNCTION get_legal_moves...

>>>>>>>> USING AUTO REPLY...

>>>>>>>> USING AUTO REPLY...

>>>>>>>> EXECUTING FUNCTION make_move...

>>>>>>>> USING AUTO REPLY...
Player White (to Player Black):

I've moved the pawn from d2 to d4, which opens up lines for our bishop and queen and contests the center directly against the black e5 pawn. This is a classic central strategy aiming to seize spatial advantage.

Now, as the game becomes more complex and tense, what’s your favorite opening to play in chess and why? Do you prefer tactical skirmishes or strategic, positional battles?

--------------------------------------------------------------------------------

3. 总结

本篇内容以下棋为案例，模拟了三个不同代理的嵌套式聊天，最大的特点是允许代理可以使用定义的工具，扩展了代理的功能。此外，嵌套式工作流是可以借鉴的地方，如何让多代理进行嵌套式聊天，从而解决问题是一个不错的思路。

l8947943
关注
专栏目录
FFmpeg-master.zip_FFmpeg-master_ffmpeg_ffmpeg 播放
09-14
它包含了多个库，如 libavcodec（编码器库）、libavformat（容器格式库）、libavfilter（过滤器库）和 libavutil（通用工具库）等，提供了解码、编码、转换、流媒体等功能。"FFmpeg-master.zip" 是 FFmpeg 源代码的...
FFmpeg.AutoGen-master_c#ffmpeg_FFmpeg.AutoGen_ffmpeg_ladysef_pay
09-30
FFmpeg.AutoGen是一个C#项目，它为FFmpeg库提供了自动化的绑定，使得在C#环境中可以方便地调用FFmpeg的功能。这个项目的核心在于利用FFmpeg的原生API，并通过自动化工具（如SWIG或者SharpGen）生成C#接口，从而在...
autogen-libopts-5.18-5.el7.x86_64.rpm下载
12-16
本文将详细解析"autogen-libopts-5.18-5.el7.x86_64.rpm"这个软件包，以及它在解决CentOS7系统中ntpd服务启动失败和yum源不可用问题上的作用。 "autogen-libopts-5.18-5.el7.x86_64.rpm"是一个专门为CentOS7设计的...
autogen-5.11.5.tar.gz_autogen
09-21
标题中的"autogen-5.11.5.tar.gz"是一个开源软件的源代码压缩包，其中"autogen"是该软件的名称，而"5.11.5"则是其版本号。这个版本号表明这是autogen软件的第5次大版本更新，第1次次要版本更新和第5次小版本更新。"....
autogen-libopts-5.18-5.el7.x86_64.rpm
12-13
autogen-libopts-5.18-5.el7.x86_64.rpm，启动ntp服务报错，又因为是内网环境，如法上外网，把这个包安装后解决！
扩散模型（2）--1
最新发布
m0_63860007的博客
09-29 479
生成模型通过学习并建模输入数据的分布，从而采集生成新的样木，该模型广泛运用于图片视频生成、文本生成和药物分子生成。扩散模型是一类概率生成模型，扩散模型通过向数据中逐步。
[论文精读]AI-Guardian: Defeating Adversarial Attacks using Backdoors
m0_52911108的博客
09-27 1036
本文提出了AI-Guardian框架，本质是让模型预测结果受到后门控制，而很难被攻击者操纵，从而抵御对抗攻击。在模型训练阶段，AI-Guardian会针对原始训练样本生成一批后门样本（经过后门处理的样本），按标签双射关系为后门样本分配对应标签，实现后门注入；在模型预测阶段，AI-Guardian先对输入数据进行后门处理，然后将其输入到模型，根据模型输出结果、标签双射关系返回对输入数据真实标签的预测结果。
零售业的数字化转型与消费者体验升级
weixin_58820787的博客
09-24 864
在数字化浪潮的推动下，零售业正经历着前所未有的变革。数字化转型不仅为零售商带来了新的商业模式和运营效率的提升，更重要的是，它极大地提升了消费者的购物体验。金智维将探讨零售业如何通过数字化转型，实现线上线下融合、智能推荐、个性化服务等方面的升级，以满足消费者日益增长的需求。
Python精选200Tips：171-175
AnFany
09-26 1127
Python深度学习的项目
Her来了，OpenAI正式发布高级语音模式
shadowcz007的博客
09-25 370
OpenAI宣布将Advanced Voice Mode（AVM）扩展到更多付费用户，包括ChatGPT的Plus和Teams用户。该功能旨在使ChatGPT的语音交互更加自然，并新增五种新声音和改进的口音识别能力。虽然Sky声音因法律争议被移除，但AVM的整体设计和性能得到了提升。新版本不仅能在用户打断时立即停止回答，还能根据用户语音中的情感调整响应。此外，视频和屏幕共享功能暂未推出。AVM目前...
OpenCV视频I/O(13）视频采集类VideoCapture之等待多个视频流中的任意一个变为可用状态函数waitAny()的使用
jndingxin的专栏
09-29 315
cv::VideoCapture::waitAny 是一个静态成员函数，用于等待多个视频流中的任意一个变为可用状态。这个函数可以用于同步多个视频流，直到其中一个流准备好为止。
持续学习与创新能力的双重提升
thinkers
09-25 422
  ---
高维空间的维数灾难问题
Python领域优质萌新创作者
09-23 304
高维空间的维数灾难问题是指在处理高维数据时，随着维度的增加，数据的性质发生了显著变化，从而导致许多传统的机器学习和统计方法失效的现象。
OTP语音播报芯片-实现智能门铃叮咚方案
jiuxin666的博客
09-26 487
九芯电子推出功能强大、外围简单的门铃叮咚芯片NV040D，音质优美，性能稳定，可定制性强，满足智能门铃行业快速简捷设计需求，引领智能门铃市场向智能化、个性化发展。
从0到1训练私有大模型技能与应用实现
DAMOXINGHAITAI的博客
09-25 1246
LoRA，英文全称 Low-Rank Adaptation of Large Language Models（中文为大语言模型的低阶适应）。这是微软的研究人员为了解决大语言模型微调而开发的一项技术，其 github 地址为 https://github.com/microsoft/LoRA ，当前已经得到HuggingFace 的 PEFT库 https://github.com/huggingface/peft 的支持。对于大语音模型来说，其参数量非常多。
sheng的学习笔记-AI-K-摇臂赌博机（K-armed bandit）
coldstarry的专栏
09-24 690
强化学习。
超分之SPIN
weixin_44342777的博客
09-23 1035
SPIN利用超像素将局部相似像素分组为可解释的局部区域。SPIN采用ISPA模块来促进不规则局部超像素区域内的局部信息交互，采用SPCA模块通过超像素的替代来促进远程信息交互。
变电站设备检测系统源码分享
xuehaisj的博客
09-25 1108
数据集信息展示在现代电力系统中，变电站作为电力传输和分配的重要环节，其设备的实时监测与故障检测显得尤为重要。为此，我们构建了一个名为“Substation elements detection”的数据集，旨在为改进YOLOv8的变电站设备检测系统提供强有力的支持。该数据集包含14个类别，涵盖了变电站中常见的设备类型，这些设备在电力系统的运行和维护中扮演着关键角色。
Hi3403V100 4K/60 芯片及开发板
WeChat_zhanggong345的博客
09-26 302
Hi3403V100是一颗面向监控市场推出的专业 Ultra-HD Smart IP Camera SOC，该芯片最高支持四路sensor输入，支持最高4K60的ISP图像处理能力，支持3F WDR、多级降噪、六轴防抖、硬件拼接等多种图像增强和处理算法，为用户提供了卓越的图像处理能力。Hi3403V100内置四核A55，提供高效且丰富和灵活的CPU资源，以满足客户计算和控制需求。Hi3403V100集成了高效的神经网络推理单元，最高10TOPS INT8，并支持业界主流的神经网络框架。
linux android glib库交叉编译 autogen.sh命令 --sysroot配置
05-27
在 Linux 平台上交叉编译 Android 平台的 GLib 库时，需要通过 autogen.sh 脚本进行初始化和配置。在这个过程中，需要通过 --host 和 --target 参数指定 GLib 库的主机架构和目标架构。 同时，为了能够正确地在 Android 平台上运行 GLib 库，还需要使用 --sysroot 参数指定 Android 平台的根目录。这个参数告诉编译器和链接器去哪里寻找 Android 平台的系统库和头文件。 例如，假设您的 Android 平台根目录为 /path/to/android/root，GLib 库的主机架构为 x86_64，目标架构为 arm64-v8a，则可以使用如下命令进行配置： ``` ./autogen.sh --host=x86_64-linux-gnu --target=aarch64-linux-android --disable-shared --enable-static --with-sysroot=/path/to/android/root ``` 其中，--disable-shared 和 --enable-static 参数用于指定编译静态库，--with-sysroot 参数用于指定 Android 平台的根目录。 希望这些信息能够对您有所帮助。
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
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

打赏作者

l8947943

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付：¥1
获取中
扫码支付

您的余额不足，请更换扫码支付或充值

打赏作者

实付元
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