Autogen

AutoGen 是一个框架，支持使用多个代理开发 LLM 应用程序，这些代理可以相互对话以解决任务。

在这里找到有关 Autogen 的更多信息： https://microsoft.github.io/autogen/docs/Getting-Started

快速开始集成

Autogen 支持一种 config_list 的概念，允许定义将要使用的 LLM 提供者和模型。Portkey 通过我们创建的自定义配置无缝集成到 Autogen 框架中。

使用最小配置的示例

from autogen import AssistantAgent, UserProxyAgent, config_list_from_json

# Import the portkey library to fetch helper functions
from portkey_ai import PORTKEY_GATEWAY_URL, createHeaders

config_list = [
    {
        "api_key": 'Your OpenAI Key', 
        "model": "gpt-3.5-turbo",
        "base_url": PORTKEY_GATEWAY_URL,
        "api_type": "openai",
        "default_headers": createHeaders(
            api_key = "Your Portkey API Key",
            provider = "openai",
        )
    }
]

assistant = AssistantAgent("assistant", llm_config={"config_list": config_list})
user_proxy = UserProxyAgent("user_proxy", code_execution_config={"work_dir": "coding", "use_docker": False}) # IMPORTANT: set to True to run code in docker, recommended
user_proxy.initiate_chat(assistant, message="Say this is also a test - part 2.")
# This initiates an automated chat between the two agents to solve the task

请注意，我们将 base_url 更新为 Portkey 的 AI 网关，然后添加了 default_headers 以启用 Portkey 特定功能。

当我们执行这个脚本时，它将产生与不使用 Portkey 相同的结果，但现在每个请求都可以在 Portkey 分析与日志用户界面中进行检查，包括令牌、成本、准确性计算。

Portkey 支持的所有配置参数都可以作为头的一部分使用。让我们看一些示例：

通过 Portkey 在 Autogen 中使用 100+ 个模型

由于 Portkey 无缝连接到 150+ 个模型，您可以轻松连接其中任何一个以在 Autogen 中运行。

让我们看一个使用 Mistral-7B 在 Anyscale 上 无缝运行 Autogen 的示例：

from autogen import AssistantAgent, UserProxyAgent, config_list_from_json

# Import the portkey library to fetch helper functions
from portkey_ai import PORTKEY_GATEWAY_URL, createHeaders

config_list = [
    {
        "api_key": 'Your Anyscale API Key', 
        "model": "mistralai/Mistral-7B-Instruct-v0.1",
        "base_url": PORTKEY_GATEWAY_URL,
        "api_type": "openai", # Portkey conforms to the openai api_type
        "default_headers": createHeaders(
            api_key = "Your Portkey API Key",
            provider = "anyscale",
        )
    }
]

assistant = AssistantAgent("assistant", llm_config={"config_list": config_list})
user_proxy = UserProxyAgent("user_proxy", code_execution_config={"work_dir": "coding", "use_docker": False}) # IMPORTANT: set to True to run code in docker, recommended
user_proxy.initiate_chat(assistant, message="Say this is also a test - part 2.")
# This initiates an automated chat between the two agents to solve the task

使用虚拟密钥

虚拟密钥 在 Portkey 中允许您轻松切换提供商，而无需手动存储和更改他们的 API 密钥。我们使用上面的 Mistral 示例，但这次使用虚拟密钥。

from autogen import AssistantAgent, UserProxyAgent, config_list_from_json

# Import the portkey library to fetch helper functions
from portkey_ai import PORTKEY_GATEWAY_URL, createHeaders

config_list = [
    {
        # Set a dummy value, since we'll pick the API key from the virtual key
        "api_key": 'X',
        
        # Pick the model from the provider of your choice
        "model": "mistralai/Mistral-7B-Instruct-v0.1",
        "base_url": PORTKEY_GATEWAY_URL,
        "api_type": "openai", # Portkey conforms to the openai api_type

         "default_headers": createHeaders(
            api_key = "Your Portkey API Key",
            # Add your virtual key here
            virtual_key = "Your Anyscale Virtual Key",
        )
    }
]

assistant = AssistantAgent("assistant", llm_config={"config_list": config_list})
user_proxy = UserProxyAgent("user_proxy", code_execution_config={"work_dir": "coding", "use_docker": False}) # IMPORTANT: set to True to run code in docker, recommended
user_proxy.initiate_chat(assistant, message="Say this is also a test - part 2.")
# This initiates an automated chat between the two agents to solve the task

使用配置

配置在Portkey中解锁了高级管理和路由功能，包括负载均衡、回退、金丝雀测试、切换模型等。

您可以像这样在Autogen中使用Portkey配置：

from autogen import AssistantAgent, UserProxyAgent, config_list_from_json

# Import the portkey library to fetch helper functions
from portkey_ai import PORTKEY_GATEWAY_URL, createHeaders

config_list = [
    {
        # Set a dummy value, since we'll pick the API key from the virtual key
        "api_key": 'X',
        
        # Pick the model from the provider of your choice
        "model": "mistralai/Mistral-7B-Instruct-v0.1",
        "base_url": PORTKEY_GATEWAY_URL,
        "api_type": "openai", # Portkey conforms to the openai api_type

         "default_headers": createHeaders(
            api_key = "Your Portkey API Key",
            # Add your Portkey config id
            config = "Your Config ID",
        )
    }
]

assistant = AssistantAgent("assistant", llm_config={"config_list": config_list})
user_proxy = UserProxyAgent("user_proxy", code_execution_config={"work_dir": "coding", "use_docker": False}) # IMPORTANT: set to True to run code in docker, recommended
user_proxy.initiate_chat(assistant, message="Say this is also a test - part 2.")
# This initiates an automated chat between the two agents to solve the task