PDF AI コパイロット

Aspose.PDF AI Copilot APIは、ユーザーが異なるプロバイダーのLLMを使用してPDFドキュメントを処理できるように設計されています。このAPIは、ユーザーがチャットボットアプリケーションを構築し、PDFソリューションをLLMと統合するのに役立ちます。

主な機能

• ドキュメント要約。
• ドキュメントとのチャット。
• ドキュメントから画像を取得し、説明を提供。

例

現在、以下のコパイロットが利用可能です：

OpenAI Summaryは、ユーザーがドキュメントから要約を生成できるようにします。モデル、温度、トークン数、モデルの指示、ドキュメントの添付ファイルなどのオプションを設定することで、要約を作成する便利な方法を提供します。このコパイロットは、テキスト、ドキュメントとして要約を非同期に生成し、さまざまな形式で要約を保存できます。提供されたデモコードは、OpenAIクライアントの作成、コパイロットオプションの設定、およびSummaryCopilotを使用して要約を生成し保存する方法を示しています。

.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は、ドキュメントとのチャットインタラクションのために設計されたAIコパイロットです。ユーザーのクエリに対する応答を生成し、コンテキストを管理することを容易にします。ユーザーは、モデル、温度、トークン数、モデルの指示、ドキュメントの添付ファイルなどのコパイロットオプションを設定できます。このコパイロットは、単一または複数のクエリに対する応答を提供し、さまざまな形式で応答を保存し、コンテキストを保存および削除できます。

提供されたコードは、OpenAIクライアントの作成、ChatCopilotオプションの設定、およびユーザーのクエリと対話しコンテキストを管理するためのChatCopilotの使用を示しています。

.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は、PDFドキュメント内の画像や別の画像ファイルの画像説明を生成するために設計されたAIコパイロットです。ユーザーは、モデル、温度、トークン数、モデルの指示、ドキュメントの添付ファイルなどのコパイロットオプションを設定できます。このコパイロットは、すべての添付ドキュメントの画像説明を一度に取得する機能を提供します。

提供されたコードスニペットは、OpenAIクライアントの作成、ImageDescriptionCopilotオプションの設定、および添付ドキュメントの画像説明を取得するためのコパイロットの使用を示しています。さらに、添付ドキュメント内の画像に画像説明を追加し、指定されたディレクトリに新しいドキュメントを保存するための拡張メソッドもあります。

.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);
}

Llama Chatは、Llamaチャット完了APIにリクエストを送信するクライアントを作成することを可能にします。

.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は、クライアントがSummary Copilotを作成するために使用できます。

.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);
}