PDF AI 副驾驶

Aspose.PDF AI Copilot API 旨在允许用户使用来自不同提供商的 LLM 处理 PDF 文档。这个 API 将帮助用户构建聊天机器人应用程序并将 PDF 解决方案与 LLM 集成。

主要特性

• 文档摘要。
• 与文档聊天。
• 从文档中获取图像并提供描述。

示例

目前，以下 copilots 可用：

OpenAI 摘要 允许用户从文档中生成摘要。它通过配置选项（如模型、温度、令牌数量、模型指令、文档附件等）提供了一种方便的创建摘要的方式。该 copilot 可以异步生成文本、文档的摘要，并以各种格式保存摘要。提供的演示代码展示了如何创建 OpenAI 客户端、配置 copilot 选项以及使用 SummaryCopilot 生成和保存摘要。

.NET Core 3.1

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task GetSummary()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    // Create AI client
    using (var openAiClient = Aspose.Pdf.AI.OpenAIClient
               .CreateWithApiKey(ApiKey) // Create OpenAI client with the API key
               //.WithProject("proj_123") // Configure optional parameters
               .Build())
    {
        // Create copilot options
        var options = Aspose.Pdf.AI.OpenAISummaryCopilotOptions
            .Create() // Create options like this, or...
            //.Create(options => { options.Model = OpenAIModels.Gpt35Turbo; }) // ...create using delegate
            .WithTemperature(0.5) // Configure other optional parameters
            .WithDocument("SampleDocument.pdf"); // .WithDocument methods allows to add text, pdf and paths to documents
            //.WithDocuments(new List<Aspose.Pdf.AI.TextDocument> { new Aspose.Pdf.AI.TextDocument() }); // .WithDocuments methods allows to add text, pdf and path collections

        // Create summary copilot
        Aspose.Pdf.AI.ISummaryCopilot summaryCopilot = Aspose.Pdf.AI.AICopilotFactory.CreateSummaryCopilot(openAiClient, options);

        // Get summary text
        string summaryText = await summaryCopilot.GetSummaryAsync();

        // Get summary document
        Aspose.Pdf.Document summaryDocument = await summaryCopilot.GetSummaryDocumentAsync();

        // Get summary document with page info
        Aspose.Pdf.Document summaryDocumentWithPageInfo = await summaryCopilot.GetSummaryDocumentAsync(new Aspose.Pdf.PageInfo());

        // Save PDF document
        await summaryCopilot.SaveSummaryAsync(dataDir + "Summary_out.pdf");
    }
}

.NET 8

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task GetSummary()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    // Create AI client
    using var openAiClient = Aspose.Pdf.AI.OpenAIClient
        .CreateWithApiKey(ApiKey) // Create OpenAI client with the API key
        //.WithProject("proj_123") // Configure optional parameters
        .Build();

    // Create copilot options
    var options = Aspose.Pdf.AI.OpenAISummaryCopilotOptions
        .Create() // Create options like this, or...
        //.Create(options => { options.Model = OpenAIModels.Gpt35Turbo; }) // ...create using delegate
        .WithTemperature(0.5) // Configure other optional parameters
        .WithDocument("SampleDocument.pdf"); // .WithDocument methods allows to add text, pdf and paths to documents
        //.WithDocuments(new List<Aspose.Pdf.AI.TextDocument> { new Aspose.Pdf.AI.TextDocument() }); // .WithDocuments methods allows to add text, pdf and path collections

    // Create summary copilot
    Aspose.Pdf.AI.ISummaryCopilot summaryCopilot = Aspose.Pdf.AI.AICopilotFactory.CreateSummaryCopilot(openAiClient, options);

    // Get summary text
    string summaryText = await summaryCopilot.GetSummaryAsync();

    // Get summary document
    Aspose.Pdf.Document summaryDocument = await summaryCopilot.GetSummaryDocumentAsync();

    // Get summary document with page info
    Aspose.Pdf.Document summaryDocumentWithPageInfo = await summaryCopilot.GetSummaryDocumentAsync(new Aspose.Pdf.PageInfo());

    // Save PDF document
    await summaryCopilot.SaveSummaryAsync(dataDir + "Summary_out.pdf");
}

OpenAI Chat 是一个设计用于与文档进行聊天交互的 AI copilot。它促进了对用户查询的响应生成和上下文管理。用户可以配置 copilot 选项，如模型、温度、令牌数量、模型指令、文档附件等。该 copilot 可以对单个或多个查询提供响应，保存响应为各种格式，保存和删除上下文。

提供的代码演示了如何创建 OpenAI 客户端、配置 ChatCopilot 选项以及使用 ChatCopilot 与用户查询进行交互和管理上下文。

.NET Core 3.1

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task ChatWithDocument()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    // Create AI client
    using (var openAiClient = Aspose.Pdf.AI.OpenAIClient
                .CreateWithApiKey(ApiKey) // Create OpenAI client with the API key
                //.WithProject("proj_123") // Configure optional parameters
                //.WithOrganization("org_123")
                .Build()) // Build
    {
        // Create copilot options
        var options = Aspose.Pdf.AI.OpenAIChatCopilotOptions
            .Create() // Create options like this, or...
            //.Create(options => { options.Model = OpenAIModels.Gpt35Turbo; }) // ...create using delegate
            .WithModel(Aspose.Pdf.AI.OpenAIModels.Gpt35Turbo) // Configure other optional parameters
            .WithTemperature(0.5)
            .WithTopP(1)
            //.WithContextBackupJsonPath("ContextBackup.json") // Supply context backup to resume the conversation session
            //.WithRestoreContextFromBackup(true) // If set to true, the context will be restored
            .WithDocument(dataDir + "SampleDocument.pdf"); // Attach documents using .WithDocument(s) methods allows to add text, pdf and paths to documents

        // Create summary copilot
        Aspose.Pdf.AI.IChatCopilot chatCopilot = Aspose.Pdf.AI.AICopilotFactory.CreateChatCopilot(openAiClient, options);

        // Get response on a user query
        string copilotResponse1 = await chatCopilot.GetResponseAsync("Summarize this document.");

        // Get response on a list of queries
        string copilotResponse2 = await chatCopilot.GetResponseAsync(new List<string>
        {
            "What is the subject of this document?",
            "How many words in it?"
        });

        // Save PDF document
        await chatCopilot.SaveResponseAsync("Summarize this document.", dataDir + "ResponseDocument1_out.pdf");

        // Save PDF document
        await chatCopilot.SaveResponseAsync(new List<string>
            {
                "What is the subject of this document?",
                "How many words in it?"
            },
            dataDir + "ResponseDocument2_out.pdf");

        // Save context (ids of assistant, thread, documents)
        await chatCopilot.SaveContextAsync(dataDir + "ContextBackup.json");

        // Deletes the context
        await chatCopilot.DeleteContextAsync();
    }
}

.NET 8

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task ChatWithDocument()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    // Create AI client
    using var openAiClient = Aspose.Pdf.AI.OpenAIClient
        .CreateWithApiKey(ApiKey) // Create OpenAI client with the API key
        //.WithProject("proj_123") // Configure optional parameters
        //.WithOrganization("org_123")
        .Build(); // Build

    // Create copilot options
    var options = Aspose.Pdf.AI.OpenAIChatCopilotOptions
        .Create() // Create options like this, or...
        //.Create(options => { options.Model = OpenAIModels.Gpt35Turbo; }) // ...create using delegate
        .WithModel(Aspose.Pdf.AI.OpenAIModels.Gpt35Turbo) // Configure other optional parameters
        .WithTemperature(0.5)
        .WithTopP(1)
        //.WithContextBackupJsonPath("ContextBackup.json") // Supply context backup to resume the conversation session
        //.WithRestoreContextFromBackup(true) // If set to true, the context will be restored
        .WithDocument(dataDir + "SampleDocument.pdf"); // Attach documents using .WithDocument(s) methods allows to add text, pdf and paths to documents

    // Create summary copilot
    Aspose.Pdf.AI.IChatCopilot chatCopilot = Aspose.Pdf.AI.AICopilotFactory.CreateChatCopilot(openAiClient, options);

    // Get response on a user query
    string copilotResponse1 = await chatCopilot.GetResponseAsync("Summarize this document.");

    // Get response on a list of queries
    string copilotResponse2 = await chatCopilot.GetResponseAsync(new List<string>
    {
        "What is the subject of this document?",
        "How many words in it?"
    });

    // Save summary as PDF document
    await chatCopilot.SaveResponseAsync("Summarize this document.", dataDir + "ResponseDocument1_out.pdf");

    // Save summary as PDF document
    await chatCopilot.SaveResponseAsync(new List<string>
        {
            "What is the subject of this document?",
            "How many words in it?"
        },
        dataDir + "ResponseDocument2_out.pdf");

    // Save context (ids of assistant, thread, documents)
    await chatCopilot.SaveContextAsync(dataDir +"ContextBackup.json");

    // Deletes the context
    await chatCopilot.DeleteContextAsync();
}

OpenAI Image Description 是一个设计用于生成 PDF 文档内图像以及单独图像文件的图像描述的 AI copilot。用户可以配置 copilot 选项，如模型、温度、令牌数量、模型指令、文档附件等。该 copilot 提供了一次获取所有附加文档的图像描述的能力。

提供的代码片段演示了如何创建 OpenAI 客户端、配置 ImageDescriptionCopilot 选项以及使用 copilot 获取附加文档的图像描述。此外，还有一个扩展方法，允许将图像描述添加到附加文档中的图像，并在提供的目录中保存新文档。

.NET Core 3.1

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task CreateImageDescriptions()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    // Create AI client
    using (var openAiClient = Aspose.Pdf.AI.OpenAIClient
        .CreateWithApiKey(ApiKey) // Create OpenAI client with the API key
        //.WithProject("proj_123") // Configure optional parameters
        //.WithOrganization("org_123")
        .Build()) // Build
    {
        // Create copilot options
        var options = Aspose.Pdf.AI.OpenAIImageDescriptionCopilotOptions
            .Create() // Create options like this, or...
            //.Create(options => { options.Model = OpenAIModels.Gpt4Turbo; }) // ...create using delegate
            .WithModel(Aspose.Pdf.AI.OpenAIModels.Gpt4Turbo) // The model should have vision capabilities
            .WithTemperature(0.5)
            .WithTopP(1)
            .WithDocument(new Aspose.Pdf.AI.PdfDocument // Attach documents
            {
                Name = "Another_Pdf_with_images",
                Document = new Aspose.Pdf.Document(dataDir + "Pdf_with_images.pdf")
            })
            .WithDocument(dataDir + "Mona_liza.jpg"); // Attach images
            //.WithDocument("Pdf_with_images.pdf"); // Attach document paths

        // Create copilot
        var copilot = Aspose.Pdf.AI.AICopilotFactory.CreateImageDescriptionCopilot(openAiClient, options);

        // Get Image descriptions
        List<Aspose.Pdf.AI.ImageDescriptionResult> imageDescriptions = await copilot.GetImageDescriptionsAsync();

        // Use extension method to add image descriptions to attached documents
        await copilot.AddPdfImageDescriptionsAsync(dataDir);
    }
}

.NET 8

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task CreateImageDescriptions()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    // Create AI client
    using var openAiClient = Aspose.Pdf.AI.OpenAIClient
        .CreateWithApiKey(ApiKey) // Create OpenAI client with the API key
        //.WithProject("proj_123") // Configure optional parameters
        //.WithOrganization("org_123")
        .Build(); // Build

    // Create copilot options
    var options = Aspose.Pdf.AI.OpenAIImageDescriptionCopilotOptions
        .Create() // Create options like this, or...
        //.Create(options => { options.Model = OpenAIModels.Gpt4Turbo; }) // ...create using delegate
        .WithModel(Aspose.Pdf.AI.OpenAIModels.Gpt4Turbo) // The model should have vision capabilities
        .WithTemperature(0.5)
        .WithTopP(1)
        .WithDocument(new Aspose.Pdf.AI.PdfDocument // Attach documents
        {
            Name = "Another_Pdf_with_images",
            Document = new Aspose.Pdf.Document(dataDir + "Pdf_with_images.pdf")
        })
        .WithDocument(dataDir + "Mona_liza.jpg"); // Attach images
        //.WithDocument(dataDir + "Pdf_with_images.pdf"); // Attach document paths

    // Create copilot
    var copilot = Aspose.Pdf.AI.AICopilotFactory.CreateImageDescriptionCopilot(openAiClient, options);

    // Get Image descriptions
    List<Aspose.Pdf.AI.ImageDescriptionResult> imageDescriptions = await copilot.GetImageDescriptionsAsync();

    // Use extension method to add image descriptions to attached documents
    await copilot.AddPdfImageDescriptionsAsync(dataDir);
}

OpenAI OCR 是一个旨在从扫描文档和图像中提取文本的 AI copilot。用户可以配置 copilot 选项，如模型、温度、令牌数量、模型指令、文档附件等。

提供的代码片段演示了如何创建 OpenAI 客户端、配置 OpenAIOcrCopilotOptions 选项以及使用 copilot 从扫描文档和图像中获取文本。

.NET Core 3.1

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task ExtractText()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    // Create AI client
    using (var openAiClient = Aspose.Pdf.AI.OpenAIClient
        .CreateWithApiKey(ApiKey) // Create OpenAI client with the API key
        //.WithProject("proj_123") // Configure optional parameters
        //.WithOrganization("org_123")
        .Build()) // Build
    {
        // Create copilot options
        var options = Aspose.Pdf.AI.OpenAIOcrCopilotOptions
            .Create() // Create options like this, or...
            //.Create(options => { options.Model = OpenAIModels.Gpt4OMini; }) // ...create using delegate
            .WithModel(Aspose.Pdf.AI.OpenAIModels.Gpt4OMini) // The model should have vision capabilities
            .WithDocument(dataDir + "ScannedDocument.pdf") // Attach document paths
            .WithDocument(dataDir + "ImageWithText.jpg"); // Attach images

        // Create copilot
        Aspose.Pdf.AI.IOcrCopilot copilot = Aspose.Pdf.AI.AICopilotFactory.CreateOcrCopilot(openAiClient, options);

        // Get text recognitions
        List<Aspose.Pdf.AI.TextRecognitionResult> textRecognitions = await copilot.GetTextRecognitionResultAsync();

        // Access to the extracted text
        string text = textRecognitions[0].OcrDetails[0].ExtractedText;
    }
}

.NET 8

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task ExtractText()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    // Create AI client
    using var openAiClient = Aspose.Pdf.AI.OpenAIClient
        .CreateWithApiKey(ApiKey) // Create OpenAI client with the API key
        //.WithProject("proj_123") // Configure optional parameters
        //.WithOrganization("org_123")
        .Build(); // Build

    // Create copilot options
    var options = Aspose.Pdf.AI.OpenAIOcrCopilotOptions
        .Create() // Create options like this, or...
        //.Create(options => { options.Model = OpenAIModels.Gpt4OMini; }) // ...create using delegate
        .WithModel(Aspose.Pdf.AI.OpenAIModels.Gpt4OMini) // The model should have vision capabilities
        .WithDocument(dataDir + "ScannedDocument.pdf") // Attach document paths
        .WithDocument(dataDir + "ImageWithText.jpg"); // Attach images

    // Create copilot
    Aspose.Pdf.AI.IOcrCopilot copilot = Aspose.Pdf.AI.AICopilotFactory.CreateOcrCopilot(openAiClient, options);

    // Get text recognitions
    List<Aspose.Pdf.AI.TextRecognitionResult> textRecognitions = await copilot.GetTextRecognitionResultAsync();

    // Access to the extracted text
    string text = textRecognitions[0].OcrDetails[0].ExtractedText;
}

Llama Chat 允许创建一个客户端以向 Llama 聊天完成 API 发送请求。

.NET Core 3.1

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task ChatWithLlama()
{
    using (var llamaClient = Aspose.Pdf.AI.LlamaClient
        .CreateWithApiKey(ApiKey) // Create Llama client with the API key
        .Build())
    {
        var result = await llamaClient.CreateCompletionAsync(new Aspose.Pdf.AI.LlamaChatCompletionRequest
        {
            Messages = new List<Aspose.Pdf.AI.ChatMessage>
            {
                Aspose.Pdf.AI.ChatMessage.FromUser("Hello!")
            }
        });

        var response = result.Choices[0].Message.Content; // Hello! How can I assist you today?
    }
}

.NET 8

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task ChatWithLlama()
{
    using var llamaClient = Aspose.Pdf.AI.LlamaClient
        .CreateWithApiKey(ApiKey) // Create Llama client with the API key
        .Build();

    var result = await llamaClient.CreateCompletionAsync(new Aspose.Pdf.AI.LlamaChatCompletionRequest
    {
        Messages = new List<Aspose.Pdf.AI.ChatMessage>
        {
            Aspose.Pdf.AI.ChatMessage.FromUser("Hello!")
        }
    });

    var response = result.Choices[0].Message.Content; // Hello! How can I assist you today?
}

Llama Summary 允许客户端用于创建 Summary Copilot。

.NET Core 3.1

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task GenerateSummary()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    using (var llamaClient = Aspose.Pdf.AI.LlamaClient
        .CreateWithApiKey(ApiKey) // Create Llama client with the API key
        .Build())
    {
        // Create copilot options
        var options = Aspose.Pdf.AI.LlamaSummaryCopilotOptions
            .Create() // Create options like this, or...
            //.Create(options => { options.Model = LlamaModels.Llama13BChat; }) // ...create using delegate
            .WithTemperature(0.5) // Configure other optional parameters
            .WithDocument(dataDir + "SampleDocument.pdf"); // .WithDocument methods allow to add text, pdf, and paths to documents
            //.WithDocuments(new List<Aspose.Pdf.AI.TextDocument> { new Aspose.Pdf.AI.TextDocument() }); // .WithDocuments methods allow to add text, pdf and path collections

        // Create summary copilot
        var summaryCopilot = Aspose.Pdf.AI.AICopilotFactory.CreateSummaryCopilot(llamaClient, options);

        // Get summary text
        string summaryText = await summaryCopilot.GetSummaryAsync();

        // Get summary document
        Aspose.Pdf.Document summaryDocument = await summaryCopilot.GetSummaryDocumentAsync();

        // Get the summary document with page info
        Aspose.Pdf.Document summaryDocumentWithPageInfo = await summaryCopilot.GetSummaryDocumentAsync(new Aspose.Pdf.PageInfo());

        // Save the summary as a PDF document
        await summaryCopilot.SaveSummaryAsync(dataDir + "Llama_out.pdf");

        // Save summary with specified format
        await summaryCopilot.SaveSummaryAsync(dataDir + "Llama_out.docx", Aspose.Pdf.SaveFormat.DocX);
    }
}

.NET 8

// For complete examples and data files, visit https://github.com/aspose-pdf/Aspose.PDF-for-.NET
private static async Task GenerateSummary()
{
    // The path to the documents directory
    var dataDir = RunExamples.GetDataDir_AsposePdf_AI();

    using var llamaClient = Aspose.Pdf.AI.LlamaClient
        .CreateWithApiKey(ApiKey) // Create Llama client with the API key
        .Build();

    // Create copilot options
    var options = Aspose.Pdf.AI.LlamaSummaryCopilotOptions
        .Create() // Create options like this, or...
        //.Create(options => { options.Model = LlamaModels.Llama13BChat; }) // ...create using delegate
        .WithTemperature(0.5) // Configure other optional parameters
        .WithDocument(dataDir + "SampleDocument.pdf"); // .WithDocument methods allow to add text, pdf, and paths to documents
        //.WithDocuments(new List<Aspose.Pdf.AI.TextDocument> { new Aspose.Pdf.AI.TextDocument() }); // .WithDocuments methods allow to add text, pdf and path collections

    // Create summary copilot
    var summaryCopilot = Aspose.Pdf.AI.AICopilotFactory.CreateSummaryCopilot(llamaClient, options);

    // Get summary text
    string summaryText = await summaryCopilot.GetSummaryAsync();

    // Get summary document
    Aspose.Pdf.Document summaryDocument = await summaryCopilot.GetSummaryDocumentAsync();

    // Get the summary document with page info
    Aspose.Pdf.Document summaryDocumentWithPageInfo = await summaryCopilot.GetSummaryDocumentAsync(new Aspose.Pdf.PageInfo());

    // Save the summary as a PDF document
    await summaryCopilot.SaveSummaryAsync(dataDir + "Llama_out.pdf");

    // Save summary with specified format
    await summaryCopilot.SaveSummaryAsync(dataDir + "Llama_out.docx", Aspose.Pdf.SaveFormat.DocX);
}