Semantic Kernel:图识文

多模态是每个LLM具有的能力，图片又是最常见的信息载体，GPT对图片的识别也很早就有了，随着GPT版本的迭代，效果越来越好。SK也是在很多就适配了图识文，只不过最近版本才支持本地图片的上传。（有点晚）
图片场景识别：

using Microsoft.SemanticKernel.ChatCompletion;using Microsoft.SemanticKernel;using Microsoft.SemanticKernel.Connectors.OpenAI;
var chatModelId = "gpt-4o";var key = File.ReadAllText(@"C:GPTkey.txt");#pragma warning disable SKEXP0070#pragma warning disable SKEXP0010#pragma warning disable SKEXP0001#pragma warning disable SKEXP0110var kernel = Kernel.CreateBuilder()   .AddOpenAIChatCompletion(chatModelId, key)   .Build();
var chat = kernel.GetRequiredService<IChatCompletionService>();var chatHistory = new ChatHistory();chatHistory.AddUserMessage(new ChatMessageContentItemCollection{     new TextContent("请说明这是那里，什么样的天气，大家在干什么?一共有多少人"),     new ImageContent(File.ReadAllBytes("tam.jpg"),"image/jpeg")});var settings = new Dictionary<string, object>{    ["max_tokens"] = 1000,    ["temperature"] = 0.2,    ["top_p"] = 0.8,    ["presence_penalty"] = 0.0,    ["frequency_penalty"] = 0.0};
var content = chat.GetStreamingChatMessageContentsAsync(chatHistory, new PromptExecutionSettings{    ExtensionData = settings});await foreach (var item in content){    Console.Write(item.Content);}Console.ReadLine();

图片：

结果：

文字识别：

using Microsoft.SemanticKernel.ChatCompletion;using Microsoft.SemanticKernel;using Microsoft.SemanticKernel.Connectors.OpenAI;
var chatModelId = "gpt-4o";var key = File.ReadAllText(@"C:GPTkey.txt");#pragma warning disable SKEXP0070#pragma warning disable SKEXP0010#pragma warning disable SKEXP0001#pragma warning disable SKEXP0110var kernel = Kernel.CreateBuilder()   .AddOpenAIChatCompletion(chatModelId, key)   .Build();
var chat = kernel.GetRequiredService<IChatCompletionService>();var chatHistory = new ChatHistory();chatHistory.AddUserMessage(new ChatMessageContentItemCollection{     new TextContent("请识别图片上的文字，并输出"),     new ImageContent(File.ReadAllBytes("japancard.png"),"image/jpeg")});var settings = new Dictionary<string, object>{    ["max_tokens"] = 1000,    ["temperature"] = 0.2,    ["top_p"] = 0.8,    ["presence_penalty"] = 0.0,    ["frequency_penalty"] = 0.0};
var content = chat.GetStreamingChatMessageContentsAsync(chatHistory, new PromptExecutionSettings{    ExtensionData = settings});await foreach (var item in content){    Console.Write(item.Content);}Console.ReadLine();

图片：

结果：