掘金 人工智能 05月08日 16:03
使用C#构建一个同时问多个LLM并总结的小工具
index_new5.html
../../../zaker_core/zaker_tpl_static/wap/tpl_guoji1.html

 

本文介绍了一种利用多个AI模型并行处理问题,并总结最常见解决方案的方法。通过将问题同时提交给多个AI,然后分析它们返回的结果,可以快速找到最可行的方案。文章详细描述了如何使用C#和Avalonia UI实现这一过程,包括界面设计、ViewModel实现以及并行和流式处理的优化。这种方法旨在提高问题解决的效率,尤其是在面对不熟悉的领域时,能够快速获取多种解决方案并进行筛选。

💡**多AI并行提问:** 将问题同时提交给多个AI模型(例如Qwen、GLM、DeepSeek),收集它们的解决方案,避免了逐个尝试不同AI的时间浪费。

⚡️**流式处理优化体验:** 最初的实现方式是非流式且非并行的,改为并行处理后,进一步优化为流式输出,使得用户可以更快地看到各个AI的回答,提升用户体验。

📊**方案总结与筛选:** 通过分析多个AI返回的结果,选择其中出现频率最高的3种方案,提高找到可行解决方案的概率。使用额外的AI模型(如Qwen2.5-72B-Instruct)进行总结。

💻**界面与代码实现:** 文章提供了Avalonia UI的界面布局代码,展示了如何并排显示多个AI的回答结果,并使用C#代码实现了与AI模型的交互,包括API调用和数据处理。

前言

在AI编程时代,如果自己能够知道一些可行的解决方案,那么描述清楚交给AI,可以有很大的帮助。

但是我们往往不知道真正可行的解决方案是什么?

我自己有过这样的经历,遇到一个需求,我不知道有哪些解决方案,就去问AI,然后AI输出一大堆东西,我一个个去试,然后再换个AI问,又提出了不同的解决方案。

在换AI问与一个个试的过程中好像浪费了很多时间。

突然出现了一个想法,不是可以一下子把问题丢给多个AI,然后再总结一下出现最多的三个方案。那么这三个方案可行的概率会大一点。然后再丢给Cursor或者Cline等AI编程工具帮我们实现一下。

这样做的缺点是比起直接在网页上问,调用API需要耗费Token,但是硅基流动给我赠送了很多额度还没用完,随便玩一下。

实现效果:

实现方案

实现方案也很简单,如下图所示:

先设计一下布局:

<UserControl xmlns="https://github.com/avaloniaui"             xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"             xmlns:d="http://schemas.microsoft.com/expression/blend/2008"             xmlns:mc="http://schemas.openxmlformats.org/markup-compatibility/2006"             mc:Ignorable="d" d:DesignWidth="800" d:DesignHeight="450"             xmlns:vm="using:AIE_Studio.ViewModels"             x:DataType="vm:DuoWenViewModel"             x:Class="AIE_Studio.Views.DuoWenView">    <StackPanel>        <TextBox Text="{Binding Question}"></TextBox>        <Button Content="提问" Command="{Binding DuoWenStreamingParallelCommand}" Margin="5"/>        <ScrollViewer VerticalScrollBarVisibility="Auto">        <Grid>            <Grid.RowDefinitions>                <RowDefinition Height="*"/>                <RowDefinition Height="*"/>            </Grid.RowDefinitions>            <Grid.ColumnDefinitions>                <ColumnDefinition Width="*"/>                <ColumnDefinition Width="*"/>                <ColumnDefinition Width="*"/>            </Grid.ColumnDefinitions>            <!-- Row 1, Column 1 -->            <StackPanel Grid.Row="0" Grid.Column="0">                <TextBlock Text="{Binding Title1}" Margin="5"/>                <ScrollViewer VerticalScrollBarVisibility="Auto">                    <TextBox Text="{Binding Result1}" AcceptsReturn="True" Margin="5" Height="300"/>                </ScrollViewer>            </StackPanel>            <!-- Row 1, Column 2 -->            <StackPanel Grid.Row="0" Grid.Column="1">                <TextBlock Text="{Binding Title2}" Margin="5"/>                <ScrollViewer VerticalScrollBarVisibility="Auto">                    <TextBox Text="{Binding Result2}" AcceptsReturn="True" Margin="5" Height="300"/>                </ScrollViewer>            </StackPanel>            <!-- Row 1, Column 3 -->            <StackPanel Grid.Row="0" Grid.Column="2">                <TextBlock Text="{Binding Title3}" Margin="5"/>                <ScrollViewer VerticalScrollBarVisibility="Auto">                    <TextBox Text="{Binding Result3}" AcceptsReturn="True" Margin="5" Height="300"/>                </ScrollViewer>            </StackPanel>            <!-- Row 2, Column 1 -->            <StackPanel Grid.Row="1" Grid.Column="0">                <TextBlock Text="{Binding Title4}" Margin="5"/>                <ScrollViewer VerticalScrollBarVisibility="Auto">                    <TextBox Text="{Binding Result4}" AcceptsReturn="True" Margin="5" Height="300"/>                </ScrollViewer>            </StackPanel>            <!-- Row 2, Column 2 -->            <StackPanel Grid.Row="1" Grid.Column="1">                <TextBlock Text="{Binding Title5}" Margin="5"/>                <ScrollViewer VerticalScrollBarVisibility="Auto">                    <TextBox Text="{Binding Result5}" AcceptsReturn="True" Margin="5" Height="300"/>                </ScrollViewer>            </StackPanel>            <!-- Row 2, Column 3 -->            <StackPanel Grid.Row="1" Grid.Column="2">                <TextBlock Text="{Binding Title6}" Margin="5"/>                <ScrollViewer VerticalScrollBarVisibility="Auto">                    <TextBox Text="{Binding Result6}" AcceptsReturn="True" Margin="5" Height="300"/>                </ScrollViewer>            </StackPanel>        </Grid>        </ScrollViewer>    </StackPanel></UserControl>

在ViewModel中先来看一下最原始的显示结果的方式:

 [RelayCommand] private async Task DuoWen() {     ApiKeyCredential apiKeyCredential = new ApiKeyCredential("your api key");     OpenAIClientOptions openAIClientOptions = new OpenAIClientOptions();     openAIClientOptions.Endpoint = new Uri("https://api.siliconflow.cn/v1");        IChatClient client1 =     new OpenAI.Chat.ChatClient("Qwen/Qwen2.5-72B-Instruct", apiKeyCredential, openAIClientOptions).AsChatClient();     var result1 = await client1.GetResponseAsync(Question);     Result1 = result1.ToString();     IChatClient client2 =     new OpenAI.Chat.ChatClient("Qwen/Qwen3-235B-A22B", apiKeyCredential, openAIClientOptions).AsChatClient();     var result2 = await client2.GetResponseAsync(Question);     Result2 = result2.ToString();     IChatClient client3 =     new OpenAI.Chat.ChatClient("THUDM/GLM-Z1-32B-0414", apiKeyCredential, openAIClientOptions).AsChatClient();     var result3 = await client3.GetResponseAsync(Question);     Result3 = result3.ToString();     IChatClient client4 =     new OpenAI.Chat.ChatClient("THUDM/GLM-4-32B-0414", apiKeyCredential, openAIClientOptions).AsChatClient();     var result4 = await client4.GetResponseAsync(Question);     Result4 = result4.ToString();     IChatClient client5 =    new OpenAI.Chat.ChatClient("deepseek-ai/DeepSeek-R1", apiKeyCredential, openAIClientOptions).AsChatClient();     var result5 = await client5.GetResponseAsync(Question);     Result5 = result5.ToString();     IChatClient client6 =     new OpenAI.Chat.ChatClient("deepseek-ai/DeepSeek-V3", apiKeyCredential, openAIClientOptions).AsChatClient();     var result6 = await client6.GetResponseAsync(Question);     Result6 = result6.ToString();

这种最简单的方式是非流式的并且也不是并行的,你会发现一个结束了才会继续向下一个提问。

但至少已经成功显示结果了,现在想要实现的是有一个窗体进行总结。

窗体设计:

<Window xmlns="https://github.com/avaloniaui"        xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"        xmlns:d="http://schemas.microsoft.com/expression/blend/2008"        xmlns:mc="http://schemas.openxmlformats.org/markup-compatibility/2006"        mc:Ignorable="d" d:DesignWidth="600" d:DesignHeight="450"        xmlns:vm="using:AIE_Studio.ViewModels"        x:Class="AIE_Studio.Views.ShowResultWindow"        x:DataType="vm:ShowResultWindowViewModel"        Title="ShowResultWindow">    <StackPanel>        <TextBlock Text="最终结果:" Margin="5" />        <ScrollViewer VerticalScrollBarVisibility="Auto">            <TextBox Text="{Binding ReceivedValue}" AcceptsReturn="True" Margin="5" Height="400"/>        </ScrollViewer>    </StackPanel></Window>

窗体的ViewModel:

public partial class ShowResultWindowViewModel : ViewModelBase{    [ObservableProperty]    private string? receivedValue;      }

然后只要在全部都有结果之后,再进行一下总结即可。

IChatClient client7 =new OpenAI.Chat.ChatClient("Qwen/Qwen2.5-72B-Instruct", apiKeyCredential, openAIClientOptions).AsChatClient();List<Microsoft.Extensions.AI.ChatMessage> messages = new List<Microsoft.Extensions.AI.ChatMessage>();string prompt = $"""      请分析以下各个助手给出的方案,选择其中提到最多的3种方案。      助手1:{result1}      助手2:{result2}      助手3:{result3}      助手4:{result4}      助手5:{result5}      助手6:{result6}      """;messages.Add(new Microsoft.Extensions.AI.ChatMessage(ChatRole.User, prompt));var result7 = await client7.GetResponseAsync(messages);var showWindow = _serviceProvider.GetRequiredService<ShowResultWindow>();var showWindowViewModel = _serviceProvider.GetRequiredService<ShowResultWindowViewModel>();showWindowViewModel.ReceivedValue = result7.ToString();showWindow.DataContext = showWindowViewModel;showWindow.Show();

以上就成功实现了。

但是还是有可以改进的地方,首先是并行,一个一个问不如同时问。

 [RelayCommand] private async Task DuoWenParallel() {     ApiKeyCredential apiKeyCredential = new ApiKeyCredential("your api key");     OpenAIClientOptions openAIClientOptions = new OpenAIClientOptions();     openAIClientOptions.Endpoint = new Uri("https://api.siliconflow.cn/v1");     // 创建一个列表来存储所有的任务     var tasks = new List<Task<string>>();     // 向每个助手发送请求并将任务添加到列表中     tasks.Add(GetResponseFromClient("Qwen/Qwen2.5-72B-Instruct", apiKeyCredential, openAIClientOptions));     tasks.Add(GetResponseFromClient("Qwen/Qwen3-235B-A22B", apiKeyCredential, openAIClientOptions));     tasks.Add(GetResponseFromClient("THUDM/GLM-Z1-32B-0414", apiKeyCredential, openAIClientOptions));     tasks.Add(GetResponseFromClient("THUDM/GLM-4-32B-0414", apiKeyCredential, openAIClientOptions));     tasks.Add(GetResponseFromClient("deepseek-ai/DeepSeek-R1", apiKeyCredential, openAIClientOptions));     tasks.Add(GetResponseFromClient("deepseek-ai/DeepSeek-V3", apiKeyCredential, openAIClientOptions));     // 等待所有任务完成     var results = await Task.WhenAll(tasks);     // 将结果分配给相应的属性     Result1 = results[0];     Result2 = results[1];     Result3 = results[2];     Result4 = results[3];     Result5 = results[4];     Result6 = results[5]; }  private async Task<string> GetResponseFromClient(string model, ApiKeyCredential apiKeyCredential, OpenAIClientOptions options)  {      IChatClient client = new OpenAI.Chat.ChatClient(model, apiKeyCredential, options).AsChatClient();      var result = await client.GetResponseAsync(Question);      return result.ToString();  }

现在虽然是并行了,但是只有等到所有助手都回答了之后,才会统一显示,用户体验也不好。

改成流式:

[RelayCommand]private async Task DuoWenStreaming(){    ApiKeyCredential apiKeyCredential = new ApiKeyCredential("your api key");    OpenAIClientOptions openAIClientOptions = new OpenAIClientOptions();    openAIClientOptions.Endpoint = new Uri("https://api.siliconflow.cn/v1");    //string question = "C#如何获取鼠标滑动选中的值?请告诉我一些可能的方案,每个方案只需用一句话描述即可,不用展开说明。";    IChatClient client1 =    new OpenAI.Chat.ChatClient("Qwen/Qwen2.5-72B-Instruct", apiKeyCredential, openAIClientOptions).AsChatClient();    await foreach (var item in client1.GetStreamingResponseAsync(Question))    {       Result1 += item.ToString();    }          }

现在查看效果:

最后再改造成流式+并行就好了。

 [RelayCommand] private async Task DuoWenStreamingParallel() {     ApiKeyCredential apiKeyCredential = new ApiKeyCredential("your api key");     OpenAIClientOptions openAIClientOptions = new OpenAIClientOptions();     openAIClientOptions.Endpoint = new Uri("https://api.siliconflow.cn/v1");     // Clear previous results     Result1 = Result2 = Result3 = Result4 = Result5 = Result6 = string.Empty;     // Create a list of tasks for parallel processing     var tasks = new List<Task>     {         ProcessStreamingResponse("Qwen/Qwen2.5-72B-Instruct", apiKeyCredential, openAIClientOptions, (text) => Result1 += text),         ProcessStreamingResponse("Qwen/Qwen3-235B-A22B", apiKeyCredential, openAIClientOptions, (text) => Result2 += text),         ProcessStreamingResponse("THUDM/GLM-Z1-32B-0414", apiKeyCredential, openAIClientOptions, (text) => Result3 += text),         ProcessStreamingResponse("THUDM/GLM-4-32B-0414", apiKeyCredential, openAIClientOptions, (text) => Result4 += text),         ProcessStreamingResponse("deepseek-ai/DeepSeek-R1", apiKeyCredential, openAIClientOptions, (text) => Result5 += text),         ProcessStreamingResponse("deepseek-ai/DeepSeek-V3", apiKeyCredential, openAIClientOptions, (text) => Result6 += text)     };     // Wait for all streaming responses to complete     await Task.WhenAll(tasks);     IChatClient client7 =     new OpenAI.Chat.ChatClient("Qwen/Qwen2.5-72B-Instruct", apiKeyCredential, openAIClientOptions).AsChatClient();     List<Microsoft.Extensions.AI.ChatMessage> messages = new List<Microsoft.Extensions.AI.ChatMessage>();     string prompt = $"""           请分析以下各个助手给出的方案,选择其中提到最多的3种方案。           助手1:{Result1}           助手2:{Result2}           助手3:{Result3}           助手4:{Result4}           助手5:{Result5}           助手6:{Result6}           """;     messages.Add(new Microsoft.Extensions.AI.ChatMessage(ChatRole.User, prompt));     var result7 = await client7.GetResponseAsync(messages);     var showWindow = _serviceProvider.GetRequiredService<ShowResultWindow>();     var showWindowViewModel = _serviceProvider.GetRequiredService<ShowResultWindowViewModel>();     showWindowViewModel.ReceivedValue = result7.ToString();     showWindow.DataContext = showWindowViewModel;     showWindow.Show(); }private async Task ProcessStreamingResponse(string model, ApiKeyCredential apiKeyCredential, OpenAIClientOptions options, Action<string> updateResult){    IChatClient client = new OpenAI.Chat.ChatClient(model, apiKeyCredential, options).AsChatClient();        await foreach (var item in client.GetStreamingResponseAsync(Question))    {        updateResult(item.ToString());    }}

这里使用了一个带有一个参数的委托来更新每个助手回复的结果。

现在再查看效果:

Qwen/Qwen3-235B-A22B、THUDM/GLM-Z1-32B-0414、deepseek-ai/DeepSeek-R1有思考过程,返回结果比较慢。

目前Microsoft.Extensions.AI.OpenAI好像还无法获取思考内容。

等待久一会之后,可以看到结果都出来了:

然后总结窗口会显示最终的总结内容:

确定方案之后可以让Cursor或者Cline帮我们写一下试试。

Fish AI Reader

Fish AI Reader

AI辅助创作,多种专业模板,深度分析,高质量内容生成。从观点提取到深度思考,FishAI为您提供全方位的创作支持。新版本引入自定义参数,让您的创作更加个性化和精准。

FishAI

FishAI

鱼阅,AI 时代的下一个智能信息助手,助你摆脱信息焦虑

联系邮箱 441953276@qq.com

相关标签

AI编程 并行处理 流式输出 Avalonia UI 多模型总结
相关文章