Traitement des PDF par l'IA en C# : Résumer, extraire et analyser des documents avec IronPDF

PM > Install-Package IronPdf
PM > Install-Package IronPdf.Extensions.AI
PM > Install-Package Microsoft.SemanticKernel
PM > Install-Package Microsoft.SemanticKernel.Plugins.Memory

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;

// Initialize IronPDF AI with Azure OpenAI credentials

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel with Azure OpenAI
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

// Create memory store for document embeddings
var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

// Initialize IronPDF AI
IronDocumentAI.Initialize(kernel, memory);

Console.WriteLine("IronPDF AI initialized successfully with Azure OpenAI");

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI

' Initialize IronPDF AI with Azure OpenAI credentials

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel with Azure OpenAI
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

' Create memory store for document embeddings
Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

' Initialize IronPDF AI
IronDocumentAI.Initialize(kernel, memory)

Console.WriteLine("IronPDF AI initialized successfully with Azure OpenAI")

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;

// Synthesize insights across multiple related documents (e.g., quarterly reports into annual summary)

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

// Define documents to synthesize
string[] documentPaths = {
    "Q1-report.pdf",
    "Q2-report.pdf",
    "Q3-report.pdf",
    "Q4-report.pdf"
};

var documentSummaries = new List<string>();

// Summarize each document
foreach (string path in documentPaths)
{
    var pdf = PdfDocument.FromFile(path);
    string summary = await pdf.Summarize();
    documentSummaries.Add($"=== {Path.GetFileName(path)} ===\n{summary}");
    Console.WriteLine($"Processed: {path}");
}

// Combine and synthesize across all documents
string combinedSummaries = string.Join("\n\n", documentSummaries);

var synthesisDoc = PdfDocument.FromFile(documentPaths[0]);

string synthesisQuery = @"Based on the quarterly summaries below, provide an annual synthesis:
ll trends across quarters
chievements and challenges
over-year patterns

s:
inedSummaries;

string synthesis = await synthesisDoc.Query(synthesisQuery);

Console.WriteLine("\n=== Annual Synthesis ===");
Console.WriteLine(synthesis);

File.WriteAllText("annual-synthesis.txt", synthesis);

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.IO

' Synthesize insights across multiple related documents (e.g., quarterly reports into annual summary)

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

' Define documents to synthesize
Dim documentPaths As String() = {
    "Q1-report.pdf",
    "Q2-report.pdf",
    "Q3-report.pdf",
    "Q4-report.pdf"
}

Dim documentSummaries = New List(Of String)()

' Summarize each document
For Each path As String In documentPaths
    Dim pdf = PdfDocument.FromFile(path)
    Dim summary As String = Await pdf.Summarize()
    documentSummaries.Add($"=== {Path.GetFileName(path)} ==={vbCrLf}{summary}")
    Console.WriteLine($"Processed: {path}")
Next

' Combine and synthesize across all documents
Dim combinedSummaries As String = String.Join(vbCrLf & vbCrLf, documentSummaries)

Dim synthesisDoc = PdfDocument.FromFile(documentPaths(0))

Dim synthesisQuery As String = "Based on the quarterly summaries below, provide an annual synthesis:" & vbCrLf &
    "Overall trends across quarters" & vbCrLf &
    "Key achievements and challenges" & vbCrLf &
    "Year-over-year patterns" & vbCrLf & vbCrLf &
    combinedSummaries

Dim synthesis As String = Await synthesisDoc.Query(synthesisQuery)

Console.WriteLine(vbCrLf & "=== Annual Synthesis ===")
Console.WriteLine(synthesis)

File.WriteAllText("annual-synthesis.txt", synthesis)

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;

// Generate executive summary from strategic documents for C-suite leadership

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

var pdf = PdfDocument.FromFile("strategic-plan.pdf");

string executiveQuery = @"Create an executive summary for C-suite leadership. Include:

cisions Required:**
ny decisions needing executive approval

al Findings:**
5 most important findings (bullet points)

ial Impact:**
e/cost implications if mentioned

ssessment:**
riority risks identified

ended Actions:**
ate next steps

er 500 words. Use business language appropriate for board presentation.";

string executiveSummary = await pdf.Query(executiveQuery);

File.WriteAllText("executive-summary.txt", executiveSummary);
Console.WriteLine("Executive summary saved to executive-summary.txt");

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI

' Generate executive summary from strategic documents for C-suite leadership

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

Dim pdf = PdfDocument.FromFile("strategic-plan.pdf")

Dim executiveQuery As String = "Create an executive summary for C-suite leadership. Include:

cisions Required:**
ny decisions needing executive approval

al Findings:**
5 most important findings (bullet points)

ial Impact:**
e/cost implications if mentioned

ssessment:**
riority risks identified

ended Actions:**
ate next steps

er 500 words. Use business language appropriate for board presentation."

Dim executiveSummary As String = Await pdf.Query(executiveQuery)

File.WriteAllText("executive-summary.txt", executiveSummary)
Console.WriteLine("Executive summary saved to executive-summary.txt")

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using System.Text.Json;

// Analyze contract clauses and identify key terms, risks, and critical dates

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

var pdf = PdfDocument.FromFile("contract.pdf");

// Define JSON schema for contract analysis
string clauseQuery = @"Analyze this contract and identify key clauses. Return JSON:

ntractType"": ""string"",
rties"": [""string""],
fectiveDate"": ""string"",
auses"": [
{
    ""type"": ""Termination|Liability|Indemnification|Confidentiality|IP|Payment|Warranty|Other"",
    ""title"": ""string"",
    ""summary"": ""string"",
    ""riskLevel"": ""Low|Medium|High"",
    ""keyTerms"": [""string""]
}

iticalDates"": [
{
    ""description"": ""string"",
    ""date"": ""string""
}

erallRiskAssessment"": ""Low|Medium|High"",
commendations"": [""string""]


: termination rights, liability caps, indemnification, IP ownership, confidentiality, payment terms.
NLY valid JSON.";

string analysisJson = await pdf.Query(clauseQuery);

try
{
    var analysis = JsonSerializer.Deserialize<JsonElement>(analysisJson);
    string formatted = JsonSerializer.Serialize(analysis, new JsonSerializerOptions { WriteIndented = true });

    Console.WriteLine("Contract Clause Analysis:");
    Console.WriteLine(formatted);

    File.WriteAllText("contract-analysis.json", formatted);

    // Display high-risk clauses
    Console.WriteLine("\n=== High Risk Clauses ===");
    foreach (var clause in analysis.GetProperty("clauses").EnumerateArray())
    {
        if (clause.GetProperty("riskLevel").GetString() == "High")
        {
            Console.WriteLine($"- {clause.GetProperty("type")}: {clause.GetProperty("summary")}");
        }
    }
}
catch (JsonException)
{
    Console.WriteLine("Unable to parse contract analysis");
    File.WriteAllText("contract-analysis-raw.txt", analysisJson);
}

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.Text.Json

' Analyze contract clauses and identify key terms, risks, and critical dates

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

Dim pdf = PdfDocument.FromFile("contract.pdf")

' Define JSON schema for contract analysis
Dim clauseQuery As String = "Analyze this contract and identify key clauses. Return JSON:

ntractType"": ""string"",
rties"": [""string""],
fectiveDate"": ""string"",
auses"": [
{
    ""type"": ""Termination|Liability|Indemnification|Confidentiality|IP|Payment|Warranty|Other"",
    ""title"": ""string"",
    ""summary"": ""string"",
    ""riskLevel"": ""Low|Medium|High"",
    ""keyTerms"": [""string""]
}

iticalDates"": [
{
    ""description"": ""string"",
    ""date"": ""string""
}

erallRiskAssessment"": ""Low|Medium|High"",
commendations"": [""string""]

: termination rights, liability caps, indemnification, IP ownership, confidentiality, payment terms.
NLY valid JSON."

Dim analysisJson As String = Await pdf.Query(clauseQuery)

Try
    Dim analysis = JsonSerializer.Deserialize(Of JsonElement)(analysisJson)
    Dim formatted As String = JsonSerializer.Serialize(analysis, New JsonSerializerOptions With {.WriteIndented = True})

    Console.WriteLine("Contract Clause Analysis:")
    Console.WriteLine(formatted)

    File.WriteAllText("contract-analysis.json", formatted)

    ' Display high-risk clauses
    Console.WriteLine(vbCrLf & "=== High Risk Clauses ===")
    For Each clause In analysis.GetProperty("clauses").EnumerateArray()
        If clause.GetProperty("riskLevel").GetString() = "High" Then
            Console.WriteLine($"- {clause.GetProperty("type")}: {clause.GetProperty("summary")}")
        End If
    Next
Catch ex As JsonException
    Console.WriteLine("Unable to parse contract analysis")
    File.WriteAllText("contract-analysis-raw.txt", analysisJson)
End Try

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using System.Text.Json;

// Extract financial metrics from annual reports and earnings documents

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

var pdf = PdfDocument.FromFile("annual-report.pdf");

// Define JSON schema for financial extraction (numbers in millions)
string financialQuery = @"Extract financial metrics from this document. Return JSON:

portPeriod"": ""string"",
mpany"": ""string"",
rrency"": ""string"",
comeStatement"": {
""revenue"": number,
""costOfRevenue"": number,
""grossProfit"": number,
""operatingExpenses"": number,
""operatingIncome"": number,
""netIncome"": number,
""eps"": number

lanceSheet"": {
""totalAssets"": number,
""totalLiabilities"": number,
""shareholdersEquity"": number,
""cash"": number,
""totalDebt"": number

yMetrics"": {
""revenueGrowthYoY"": ""string"",
""grossMargin"": ""string"",
""operatingMargin"": ""string"",
""netMargin"": ""string"",
""debtToEquity"": number

idance"": {
""nextQuarterRevenue"": ""string"",
""fullYearRevenue"": ""string"",
""notes"": ""string""



 for unavailable data. Numbers in millions unless stated.
NLY valid JSON.";

string financialJson = await pdf.Query(financialQuery);

try
{
    var financials = JsonSerializer.Deserialize<JsonElement>(financialJson);
    string formatted = JsonSerializer.Serialize(financials, new JsonSerializerOptions { WriteIndented = true });

    Console.WriteLine("Extracted Financial Data:");
    Console.WriteLine(formatted);

    File.WriteAllText("financial-data.json", formatted);
}
catch (JsonException)
{
    Console.WriteLine("Unable to parse financial data");
    File.WriteAllText("financial-raw.txt", financialJson);
}

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.Text.Json

' Extract financial metrics from annual reports and earnings documents

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

Dim pdf = PdfDocument.FromFile("annual-report.pdf")

' Define JSON schema for financial extraction (numbers in millions)
Dim financialQuery As String = "Extract financial metrics from this document. Return JSON:

portPeriod"": ""string"",
mpany"": ""string"",
rrency"": ""string"",
comeStatement"": {
""revenue"": number,
""costOfRevenue"": number,
""grossProfit"": number,
""operatingExpenses"": number,
""operatingIncome"": number,
""netIncome"": number,
""eps"": number

lanceSheet"": {
""totalAssets"": number,
""totalLiabilities"": number,
""shareholdersEquity"": number,
""cash"": number,
""totalDebt"": number

yMetrics"": {
""revenueGrowthYoY"": ""string"",
""grossMargin"": ""string"",
""operatingMargin"": ""string"",
""netMargin"": ""string"",
""debtToEquity"": number

idance"": {
""nextQuarterRevenue"": ""string"",
""fullYearRevenue"": ""string"",
""notes"": ""string""



 for unavailable data. Numbers in millions unless stated.
NLY valid JSON."

Dim financialJson As String = Await pdf.Query(financialQuery)

Try
    Dim financials = JsonSerializer.Deserialize(Of JsonElement)(financialJson)
    Dim formatted As String = JsonSerializer.Serialize(financials, New JsonSerializerOptions With {.WriteIndented = True})

    Console.WriteLine("Extracted Financial Data:")
    Console.WriteLine(formatted)

    File.WriteAllText("financial-data.json", formatted)
Catch ex As JsonException
    Console.WriteLine("Unable to parse financial data")
    File.WriteAllText("financial-raw.txt", financialJson)
End Try

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using System.Text.Json;

// Extract structured research metadata from academic papers

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

var pdf = PdfDocument.FromFile("research-paper.pdf");

// Define JSON schema for research paper extraction
string researchQuery = @"Extract structured information from this research paper. Return JSON:

tle"": ""string"",
thors"": [""string""],
stitution"": ""string"",
blicationDate"": ""string"",
stract"": ""string"",
searchQuestion"": ""string"",
thodology"": {
""type"": ""Quantitative|Qualitative|Mixed Methods"",
""approach"": ""string"",
""sampleSize"": ""string"",
""dataCollection"": ""string""

yFindings"": [
{
    ""finding"": ""string"",
    ""significance"": ""string"",
    ""confidence"": ""High|Medium|Low""
}

mitations"": [""string""],
tureWork"": [""string""],
ywords"": [""string""]


 extracting verifiable claims and noting uncertainty.
NLY valid JSON.";

string extractionResult = await pdf.Query(researchQuery);

try
{
    var research = JsonSerializer.Deserialize<JsonElement>(extractionResult);
    string formatted = JsonSerializer.Serialize(research, new JsonSerializerOptions { WriteIndented = true });

    Console.WriteLine("Research Paper Extraction:");
    Console.WriteLine(formatted);

    File.WriteAllText("research-extraction.json", formatted);

    // Display key findings with confidence levels
    Console.WriteLine("\n=== Key Findings ===");
    foreach (var finding in research.GetProperty("keyFindings").EnumerateArray())
    {
        string confidence = finding.GetProperty("confidence").GetString() ?? "Unknown";
        Console.WriteLine($"[{confidence}] {finding.GetProperty("finding")}");
    }
}
catch (JsonException)
{
    Console.WriteLine("Unable to parse research extraction");
    File.WriteAllText("research-raw.txt", extractionResult);
}

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.Text.Json

' Extract structured research metadata from academic papers

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

Dim pdf = PdfDocument.FromFile("research-paper.pdf")

' Define JSON schema for research paper extraction
Dim researchQuery As String = "Extract structured information from this research paper. Return JSON:

tle"": ""string"",
thors"": [""string""],
stitution"": ""string"",
blicationDate"": ""string"",
stract"": ""string"",
searchQuestion"": ""string"",
thodology"": {
""type"": ""Quantitative|Qualitative|Mixed Methods"",
""approach"": ""string"",
""sampleSize"": ""string"",
""dataCollection"": ""string""

yFindings"": [
{
    ""finding"": ""string"",
    ""significance"": ""string"",
    ""confidence"": ""High|Medium|Low""
}

mitations"": [""string""],
tureWork"": [""string""],
ywords"": [""string""]

 extracting verifiable claims and noting uncertainty.
NLY valid JSON."

Dim extractionResult As String = Await pdf.Query(researchQuery)

Try
    Dim research = JsonSerializer.Deserialize(Of JsonElement)(extractionResult)
    Dim formatted As String = JsonSerializer.Serialize(research, New JsonSerializerOptions With {.WriteIndented = True})

    Console.WriteLine("Research Paper Extraction:")
    Console.WriteLine(formatted)

    File.WriteAllText("research-extraction.json", formatted)

    ' Display key findings with confidence levels
    Console.WriteLine(vbCrLf & "=== Key Findings ===")
    For Each finding In research.GetProperty("keyFindings").EnumerateArray()
        Dim confidence As String = finding.GetProperty("confidence").GetString() OrElse "Unknown"
        Console.WriteLine($"[{confidence}] {finding.GetProperty("finding")}")
    Next
Catch ex As JsonException
    Console.WriteLine("Unable to parse research extraction")
    File.WriteAllText("research-raw.txt", extractionResult)
End Try

using IronPdf;

// Split long documents into overlapping chunks for RAG systems

var pdf = PdfDocument.FromFile("long-document.pdf");

// Chunking configuration
int maxChunkTokens = 4000;      // Leave room for prompts and responses
int overlapTokens = 200;        // Overlap for context continuity
int approxCharsPerToken = 4;    // Rough estimate for tokenization

int maxChunkChars = maxChunkTokens * approxCharsPerToken;
int overlapChars = overlapTokens * approxCharsPerToken;

var chunks = new List<DocumentChunk>();
var currentChunk = new System.Text.StringBuilder();
int chunkStartPage = 1;
int currentPage = 1;

for (int i = 0; i < pdf.PageCount; i++)
{
    string pageText = pdf.Pages[i].Text;
    currentPage = i + 1;

    if (currentChunk.Length + pageText.Length > maxChunkChars && currentChunk.Length > 0)
    {
        chunks.Add(new DocumentChunk
        {
            Text = currentChunk.ToString(),
            StartPage = chunkStartPage,
            EndPage = currentPage - 1,
            ChunkIndex = chunks.Count
        });

        // Create overlap with previous chunk for continuity
        string overlap = currentChunk.Length > overlapChars
            ? currentChunk.ToString().Substring(currentChunk.Length - overlapChars)
            : currentChunk.ToString();

        currentChunk.Clear();
        currentChunk.Append(overlap);
        chunkStartPage = currentPage - 1;
    }

    currentChunk.AppendLine($"\n--- Page {currentPage} ---\n");
    currentChunk.Append(pageText);
}

if (currentChunk.Length > 0)
{
    chunks.Add(new DocumentChunk
    {
        Text = currentChunk.ToString(),
        StartPage = chunkStartPage,
        EndPage = currentPage,
        ChunkIndex = chunks.Count
    });
}

Console.WriteLine($"Document chunked into {chunks.Count} segments");
foreach (var chunk in chunks)
{
    Console.WriteLine($"  Chunk {chunk.ChunkIndex + 1}: Pages {chunk.StartPage}-{chunk.EndPage} ({chunk.Text.Length} chars)");
}

// Save chunk metadata for RAG indexing
File.WriteAllText("chunks-metadata.json", System.Text.Json.JsonSerializer.Serialize(
    chunks.Select(c => new { c.ChunkIndex, c.StartPage, c.EndPage, Length = c.Text.Length }),
    new System.Text.Json.JsonSerializerOptions { WriteIndented = true }
));


ic class DocumentChunk

public string Text { get; set; } = "";
public int StartPage { get; set; }
public int EndPage { get; set; }
public int ChunkIndex { get; set; }

Imports IronPdf
Imports System.IO
Imports System.Text
Imports System.Text.Json

' Split long documents into overlapping chunks for RAG systems

Dim pdf = PdfDocument.FromFile("long-document.pdf")

' Chunking configuration
Dim maxChunkTokens As Integer = 4000      ' Leave room for prompts and responses
Dim overlapTokens As Integer = 200        ' Overlap for context continuity
Dim approxCharsPerToken As Integer = 4    ' Rough estimate for tokenization

Dim maxChunkChars As Integer = maxChunkTokens * approxCharsPerToken
Dim overlapChars As Integer = overlapTokens * approxCharsPerToken

Dim chunks As New List(Of DocumentChunk)()
Dim currentChunk As New StringBuilder()
Dim chunkStartPage As Integer = 1
Dim currentPage As Integer = 1

For i As Integer = 0 To pdf.PageCount - 1
    Dim pageText As String = pdf.Pages(i).Text
    currentPage = i + 1

    If currentChunk.Length + pageText.Length > maxChunkChars AndAlso currentChunk.Length > 0 Then
        chunks.Add(New DocumentChunk With {
            .Text = currentChunk.ToString(),
            .StartPage = chunkStartPage,
            .EndPage = currentPage - 1,
            .ChunkIndex = chunks.Count
        })

        ' Create overlap with previous chunk for continuity
        Dim overlap As String = If(currentChunk.Length > overlapChars,
            currentChunk.ToString().Substring(currentChunk.Length - overlapChars),
            currentChunk.ToString())

        currentChunk.Clear()
        currentChunk.Append(overlap)
        chunkStartPage = currentPage - 1
    End If

    currentChunk.AppendLine(vbCrLf & "--- Page " & currentPage & " ---" & vbCrLf)
    currentChunk.Append(pageText)
Next

If currentChunk.Length > 0 Then
    chunks.Add(New DocumentChunk With {
        .Text = currentChunk.ToString(),
        .StartPage = chunkStartPage,
        .EndPage = currentPage,
        .ChunkIndex = chunks.Count
    })
End If

Console.WriteLine($"Document chunked into {chunks.Count} segments")
For Each chunk In chunks
    Console.WriteLine($"  Chunk {chunk.ChunkIndex + 1}: Pages {chunk.StartPage}-{chunk.EndPage} ({chunk.Text.Length} chars)")
Next

' Save chunk metadata for RAG indexing
File.WriteAllText("chunks-metadata.json", JsonSerializer.Serialize(
    chunks.Select(Function(c) New With {.ChunkIndex = c.ChunkIndex, .StartPage = c.StartPage, .EndPage = c.EndPage, .Length = c.Text.Length}),
    New JsonSerializerOptions With {.WriteIndented = True}
))

Public Class DocumentChunk
    Public Property Text As String = ""
    Public Property StartPage As Integer
    Public Property EndPage As Integer
    Public Property ChunkIndex As Integer
End Class

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;

// Retrieval-Augmented Generation (RAG) system for querying across multiple indexed documents

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

// Index all documents in folder
string[] documentPaths = Directory.GetFiles("documents/", "*.pdf");

Console.WriteLine($"Indexing {documentPaths.Length} documents...\n");

// Memorize each document (creates embeddings for retrieval)
foreach (string path in documentPaths)
{
    var pdf = PdfDocument.FromFile(path);
    await pdf.Memorize();
    Console.WriteLine($"Indexed: {Path.GetFileName(path)} ({pdf.PageCount} pages)");
}

Console.WriteLine("\n=== RAG System Ready ===\n");

// Query across all indexed documents
string query = "What are the key compliance requirements for data retention?";

Console.WriteLine($"Query: {query}\n");

var searchPdf = PdfDocument.FromFile(documentPaths[0]);
string answer = await searchPdf.Query(query);

Console.WriteLine($"Answer: {answer}");

// Interactive query loop
Console.WriteLine("\n--- Enter questions (type 'exit' to quit) ---\n");

while (true)
{
    Console.Write("Question: ");
    string? userQuery = Console.ReadLine();

    if (string.IsNullOrWhiteSpace(userQuery) || userQuery.ToLower() == "exit")
        break;

    string response = await searchPdf.Query(userQuery);
    Console.WriteLine($"\nAnswer: {response}\n");
}

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.IO

' Retrieval-Augmented Generation (RAG) system for querying across multiple indexed documents

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

' Index all documents in folder
Dim documentPaths As String() = Directory.GetFiles("documents/", "*.pdf")

Console.WriteLine($"Indexing {documentPaths.Length} documents..." & vbCrLf)

' Memorize each document (creates embeddings for retrieval)
For Each path As String In documentPaths
    Dim pdf = PdfDocument.FromFile(path)
    Await pdf.Memorize()
    Console.WriteLine($"Indexed: {Path.GetFileName(path)} ({pdf.PageCount} pages)")
Next

Console.WriteLine(vbCrLf & "=== RAG System Ready ===" & vbCrLf)

' Query across all indexed documents
Dim query As String = "What are the key compliance requirements for data retention?"

Console.WriteLine($"Query: {query}" & vbCrLf)

Dim searchPdf = PdfDocument.FromFile(documentPaths(0))
Dim answer As String = Await searchPdf.Query(query)

Console.WriteLine($"Answer: {answer}")

' Interactive query loop
Console.WriteLine(vbCrLf & "--- Enter questions (type 'exit' to quit) ---" & vbCrLf)

While True
    Console.Write("Question: ")
    Dim userQuery As String = Console.ReadLine()

    If String.IsNullOrWhiteSpace(userQuery) OrElse userQuery.ToLower() = "exit" Then
        Exit While
    End If

    Dim response As String = Await searchPdf.Query(userQuery)
    Console.WriteLine(vbCrLf & $"Answer: {response}" & vbCrLf)
End While

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using System;
using System.Collections.Concurrent;
using System.Text;

// Process multiple documents in parallel with rate limiting

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

// Configure parallel processing with rate limiting
int maxConcurrency = 3;
string inputFolder = "documents/";
string outputFolder = "summaries/";

Directory.CreateDirectory(outputFolder);

string[] pdfFiles = Directory.GetFiles(inputFolder, "*.pdf");
Console.WriteLine($"Processing {pdfFiles.Length} documents...\n");

var results = new ConcurrentBag<ProcessingResult>();
var semaphore = new SemaphoreSlim(maxConcurrency);

var tasks = pdfFiles.Select(async filePath =>
{
    await semaphore.WaitAsync();
    var result = new ProcessingResult { FilePath = filePath };

    try
    {
        var stopwatch = System.Diagnostics.Stopwatch.StartNew();

        var pdf = PdfDocument.FromFile(filePath);
        string summary = await pdf.Summarize();

        string outputPath = Path.Combine(outputFolder,
            Path.GetFileNameWithoutExtension(filePath) + "-summary.txt");
        await File.WriteAllTextAsync(outputPath, summary);

        stopwatch.Stop();
        result.Success = true;
        result.ProcessingTime = stopwatch.Elapsed;
        result.OutputPath = outputPath;

        Console.WriteLine($"[OK] {Path.GetFileName(filePath)} ({stopwatch.ElapsedMilliseconds}ms)");
    }
    catch (Exception ex)
    {
        result.Success = false;
        result.ErrorMessage = ex.Message;
        Console.WriteLine($"[ERROR] {Path.GetFileName(filePath)}: {ex.Message}");
    }
    finally
    {
        semaphore.Release();
        results.Add(result);
    }
}).ToArray();

await Task.WhenAll(tasks);

// Generate processing report
var successful = results.Where(r => r.Success).ToList();
var failed = results.Where(r => !r.Success).ToList();

var report = new StringBuilder();
report.AppendLine("=== Batch Processing Report ===");
report.AppendLine($"Successful: {successful.Count}");
report.AppendLine($"Failed: {failed.Count}");

if (successful.Any())
{
    var avgTime = TimeSpan.FromMilliseconds(successful.Average(r => r.ProcessingTime.TotalMilliseconds));
    report.AppendLine($"Average processing time: {avgTime.TotalSeconds:F1}s");
}

if (failed.Any())
{
    report.AppendLine("\nFailed documents:");
    foreach (var fail in failed)
        report.AppendLine($"  - {Path.GetFileName(fail.FilePath)}: {fail.ErrorMessage}");
}

string reportText = report.ToString();
Console.WriteLine($"\n{reportText}");
File.WriteAllText(Path.Combine(outputFolder, "processing-report.txt"), reportText);


s ProcessingResult

public string FilePath { get; set; } = "";
public bool Success { get; set; }
public TimeSpan ProcessingTime { get; set; }
public string OutputPath { get; set; } = "";
public string ErrorMessage { get; set; } = "";

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System
Imports System.Collections.Concurrent
Imports System.Text
Imports System.IO
Imports System.Linq
Imports System.Threading
Imports System.Threading.Tasks

' Process multiple documents in parallel with rate limiting

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

' Configure parallel processing with rate limiting
Dim maxConcurrency As Integer = 3
Dim inputFolder As String = "documents/"
Dim outputFolder As String = "summaries/"

Directory.CreateDirectory(outputFolder)

Dim pdfFiles As String() = Directory.GetFiles(inputFolder, "*.pdf")
Console.WriteLine($"Processing {pdfFiles.Length} documents...{vbCrLf}")

Dim results = New ConcurrentBag(Of ProcessingResult)()
Dim semaphore = New SemaphoreSlim(maxConcurrency)

Dim tasks = pdfFiles.Select(Async Function(filePath)
                                Await semaphore.WaitAsync()
                                Dim result = New ProcessingResult With {.FilePath = filePath}

                                Try
                                    Dim stopwatch = System.Diagnostics.Stopwatch.StartNew()

                                    Dim pdf = PdfDocument.FromFile(filePath)
                                    Dim summary As String = Await pdf.Summarize()

                                    Dim outputPath = Path.Combine(outputFolder, Path.GetFileNameWithoutExtension(filePath) & "-summary.txt")
                                    Await File.WriteAllTextAsync(outputPath, summary)

                                    stopwatch.Stop()
                                    result.Success = True
                                    result.ProcessingTime = stopwatch.Elapsed
                                    result.OutputPath = outputPath

                                    Console.WriteLine($"[OK] {Path.GetFileName(filePath)} ({stopwatch.ElapsedMilliseconds}ms)")
                                Catch ex As Exception
                                    result.Success = False
                                    result.ErrorMessage = ex.Message
                                    Console.WriteLine($"[ERROR] {Path.GetFileName(filePath)}: {ex.Message}")
                                Finally
                                    semaphore.Release()
                                    results.Add(result)
                                End Try
                            End Function).ToArray()

Await Task.WhenAll(tasks)

' Generate processing report
Dim successful = results.Where(Function(r) r.Success).ToList()
Dim failed = results.Where(Function(r) Not r.Success).ToList()

Dim report = New StringBuilder()
report.AppendLine("=== Batch Processing Report ===")
report.AppendLine($"Successful: {successful.Count}")
report.AppendLine($"Failed: {failed.Count}")

If successful.Any() Then
    Dim avgTime = TimeSpan.FromMilliseconds(successful.Average(Function(r) r.ProcessingTime.TotalMilliseconds))
    report.AppendLine($"Average processing time: {avgTime.TotalSeconds:F1}s")
End If

If failed.Any() Then
    report.AppendLine($"{vbCrLf}Failed documents:")
    For Each fail In failed
        report.AppendLine($"  - {Path.GetFileName(fail.FilePath)}: {fail.ErrorMessage}")
    Next
End If

Dim reportText As String = report.ToString()
Console.WriteLine($"{vbCrLf}{reportText}")
File.WriteAllText(Path.Combine(outputFolder, "processing-report.txt"), reportText)

Public Class ProcessingResult
    Public Property FilePath As String = ""
    Public Property Success As Boolean
    Public Property ProcessingTime As TimeSpan
    Public Property OutputPath As String = ""
    Public Property ErrorMessage As String = ""
End Class

using IronPdf;
using System.Text.Json;
using System.Net.Http.Headers;

// Use OpenAI Batch API for 50% cost savings on large-scale document processing

string openAiApiKey = "your-openai-api-key";
string inputFolder = "documents/";

// Prepare batch requests in JSONL format
var batchRequests = new List<string>();
string[] pdfFiles = Directory.GetFiles(inputFolder, "*.pdf");

Console.WriteLine($"Preparing batch for {pdfFiles.Length} documents...\n");

foreach (string filePath in pdfFiles)
{
    var pdf = PdfDocument.FromFile(filePath);
    string pdfText = pdf.ExtractAllText();

    // Truncate to stay within batch API limits
    if (pdfText.Length > 100000)
        pdfText = pdfText.Substring(0, 100000) + "\n[Truncated...]";

    var request = new
    {
        custom_id = Path.GetFileNameWithoutExtension(filePath),
        method = "POST",
        url = "/v1/chat/completions",
        body = new
        {
            model = "gpt-4o",
            messages = new[]
            {
                new { role = "system", content = "Summarize the following document concisely." },
                new { role = "user", content = pdfText }
            },
            max_tokens = 1000
        }
    };

    batchRequests.Add(JsonSerializer.Serialize(request));
}

// Create JSONL file
string batchFilePath = "batch-requests.jsonl";
File.WriteAllLines(batchFilePath, batchRequests);
Console.WriteLine($"Created batch file with {batchRequests.Count} requests");

// Upload file to OpenAI
using var httpClient = new HttpClient();
httpClient.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", openAiApiKey);

using var fileContent = new MultipartFormDataContent();
fileContent.Add(new ByteArrayContent(File.ReadAllBytes(batchFilePath)), "file", "batch-requests.jsonl");
fileContent.Add(new StringContent("batch"), "purpose");

var uploadResponse = await httpClient.PostAsync("https://api.openai.com/v1/files", fileContent);
var uploadResult = JsonSerializer.Deserialize<JsonElement>(await uploadResponse.Content.ReadAsStringAsync());
string fileId = uploadResult.GetProperty("id").GetString()!;
Console.WriteLine($"Uploaded file: {fileId}");

// Create batch job (24-hour completion window for 50% discount)
var batchJobRequest = new
{
    input_file_id = fileId,
    endpoint = "/v1/chat/completions",
    completion_window = "24h"
};

var batchResponse = await httpClient.PostAsync(
    "https://api.openai.com/v1/batches",
    new StringContent(JsonSerializer.Serialize(batchJobRequest), System.Text.Encoding.UTF8, "application/json")
);

var batchResult = JsonSerializer.Deserialize<JsonElement>(await batchResponse.Content.ReadAsStringAsync());
string batchId = batchResult.GetProperty("id").GetString()!;

Console.WriteLine($"\nBatch job created: {batchId}");
Console.WriteLine("Job will complete within 24 hours");
Console.WriteLine($"Check status: GET https://api.openai.com/v1/batches/{batchId}");

File.WriteAllText("batch-job-id.txt", batchId);
Console.WriteLine("\nBatch ID saved to batch-job-id.txt");

Imports IronPdf
Imports System.Text.Json
Imports System.Net.Http.Headers

' Use OpenAI Batch API for 50% cost savings on large-scale document processing

Dim openAiApiKey As String = "your-openai-api-key"
Dim inputFolder As String = "documents/"

' Prepare batch requests in JSONL format
Dim batchRequests As New List(Of String)()
Dim pdfFiles As String() = Directory.GetFiles(inputFolder, "*.pdf")

Console.WriteLine($"Preparing batch for {pdfFiles.Length} documents..." & vbCrLf)

For Each filePath As String In pdfFiles
    Dim pdf = PdfDocument.FromFile(filePath)
    Dim pdfText As String = pdf.ExtractAllText()

    ' Truncate to stay within batch API limits
    If pdfText.Length > 100000 Then
        pdfText = pdfText.Substring(0, 100000) & vbCrLf & "[Truncated...]"
    End If

    Dim request = New With {
        .custom_id = Path.GetFileNameWithoutExtension(filePath),
        .method = "POST",
        .url = "/v1/chat/completions",
        .body = New With {
            .model = "gpt-4o",
            .messages = New Object() {
                New With {.role = "system", .content = "Summarize the following document concisely."},
                New With {.role = "user", .content = pdfText}
            },
            .max_tokens = 1000
        }
    }

    batchRequests.Add(JsonSerializer.Serialize(request))
Next

' Create JSONL file
Dim batchFilePath As String = "batch-requests.jsonl"
File.WriteAllLines(batchFilePath, batchRequests)
Console.WriteLine($"Created batch file with {batchRequests.Count} requests")

' Upload file to OpenAI
Using httpClient As New HttpClient()
    httpClient.DefaultRequestHeaders.Authorization = New AuthenticationHeaderValue("Bearer", openAiApiKey)

    Using fileContent As New MultipartFormDataContent()
        fileContent.Add(New ByteArrayContent(File.ReadAllBytes(batchFilePath)), "file", "batch-requests.jsonl")
        fileContent.Add(New StringContent("batch"), "purpose")

        Dim uploadResponse = Await httpClient.PostAsync("https://api.openai.com/v1/files", fileContent)
        Dim uploadResult = JsonSerializer.Deserialize(Of JsonElement)(Await uploadResponse.Content.ReadAsStringAsync())
        Dim fileId As String = uploadResult.GetProperty("id").GetString()
        Console.WriteLine($"Uploaded file: {fileId}")

        ' Create batch job (24-hour completion window for 50% discount)
        Dim batchJobRequest = New With {
            .input_file_id = fileId,
            .endpoint = "/v1/chat/completions",
            .completion_window = "24h"
        }

        Dim batchResponse = Await httpClient.PostAsync(
            "https://api.openai.com/v1/batches",
            New StringContent(JsonSerializer.Serialize(batchJobRequest), System.Text.Encoding.UTF8, "application/json")
        )

        Dim batchResult = JsonSerializer.Deserialize(Of JsonElement)(Await batchResponse.Content.ReadAsStringAsync())
        Dim batchId As String = batchResult.GetProperty("id").GetString()

        Console.WriteLine(vbCrLf & $"Batch job created: {batchId}")
        Console.WriteLine("Job will complete within 24 hours")
        Console.WriteLine($"Check status: GET https://api.openai.com/v1/batches/{batchId}")

        File.WriteAllText("batch-job-id.txt", batchId)
        Console.WriteLine(vbCrLf & "Batch ID saved to batch-job-id.txt")
    End Using
End Using

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using System;
using System.Collections.Generic;
using System.Security.Cryptography;
using System.Text.Json;

// Cache AI processing results using file hashes to avoid reprocessing unchanged documents

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

// Configure caching
string cacheFolder = "ai-cache/";
string documentsFolder = "documents/";

Directory.CreateDirectory(cacheFolder);

var cacheManager = new DocumentCacheManager(cacheFolder);

// Process documents with caching
string[] pdfFiles = Directory.GetFiles(documentsFolder, "*.pdf");
int cached = 0, processed = 0;

foreach (string filePath in pdfFiles)
{
    string fileName = Path.GetFileName(filePath);
    string fileHash = cacheManager.ComputeFileHash(filePath);

    var cachedResult = cacheManager.GetCachedResult(fileName, fileHash);

    if (cachedResult != null)
    {
        Console.WriteLine($"[CACHE HIT] {fileName}");
        cached++;
        continue;
    }

    Console.WriteLine($"[PROCESSING] {fileName}");
    var pdf = PdfDocument.FromFile(filePath);
    string summary = await pdf.Summarize();

    cacheManager.CacheResult(fileName, fileHash, summary);
    processed++;
}

Console.WriteLine($"\nProcessing complete: {cached} cached, {processed} newly processed");
Console.WriteLine($"Cost savings: {(cached * 100.0 / Math.Max(1, cached + processed)):F1}% served from cache");


ash-based cache manager with JSON index
s DocumentCacheManager

private readonly string _cacheFolder;
private readonly string _indexPath;
private Dictionary<string, CacheEntry> _index;

public DocumentCacheManager(string cacheFolder)
{
    _cacheFolder = cacheFolder;
    _indexPath = Path.Combine(cacheFolder, "cache-index.json");
    _index = LoadIndex();
}

private Dictionary<string, CacheEntry> LoadIndex()
{
    if (File.Exists(_indexPath))
    {
        string json = File.ReadAllText(_indexPath);
        return JsonSerializer.Deserialize<Dictionary<string, CacheEntry>>(json) ?? new();
    }
    return new Dictionary<string, CacheEntry>();
}

private void SaveIndex()
{
    string json = JsonSerializer.Serialize(_index, new JsonSerializerOptions { WriteIndented = true });
    File.WriteAllText(_indexPath, json);
}

// SHA256 hash to detect file changes
public string ComputeFileHash(string filePath)
{
    using var sha256 = SHA256.Create();
    using var stream = File.OpenRead(filePath);
    byte[] hash = sha256.ComputeHash(stream);
    return Convert.ToHexString(hash);
}

public string? GetCachedResult(string fileName, string currentHash)
{
    if (_index.TryGetValue(fileName, out var entry))
    {
        if (entry.FileHash == currentHash && File.Exists(entry.CachePath))
        {
            entry.LastAccessed = DateTime.UtcNow;
            SaveIndex();
            return File.ReadAllText(entry.CachePath);
        }
    }
    return null;
}

public void CacheResult(string fileName, string fileHash, string result)
{
    string cachePath = Path.Combine(_cacheFolder, $"{Path.GetFileNameWithoutExtension(fileName)}-{fileHash[..8]}.txt");
    File.WriteAllText(cachePath, result);

    _index[fileName] = new CacheEntry
    {
        FileHash = fileHash,
        CachePath = cachePath,
        CreatedAt = DateTime.UtcNow,
        LastAccessed = DateTime.UtcNow
    };

    SaveIndex();
}


s CacheEntry

public string FileHash { get; set; } = "";
public string CachePath { get; set; } = "";
public DateTime CreatedAt { get; set; }
public DateTime LastAccessed { get; set; }

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System
Imports System.Collections.Generic
Imports System.Security.Cryptography
Imports System.Text.Json

' Cache AI processing results using file hashes to avoid reprocessing unchanged documents

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

' Configure caching
Dim cacheFolder As String = "ai-cache/"
Dim documentsFolder As String = "documents/"

Directory.CreateDirectory(cacheFolder)

Dim cacheManager = New DocumentCacheManager(cacheFolder)

' Process documents with caching
Dim pdfFiles As String() = Directory.GetFiles(documentsFolder, "*.pdf")
Dim cached As Integer = 0, processed As Integer = 0

For Each filePath As String In pdfFiles
    Dim fileName As String = Path.GetFileName(filePath)
    Dim fileHash As String = cacheManager.ComputeFileHash(filePath)

    Dim cachedResult As String = cacheManager.GetCachedResult(fileName, fileHash)

    If cachedResult IsNot Nothing Then
        Console.WriteLine($"[CACHE HIT] {fileName}")
        cached += 1
        Continue For
    End If

    Console.WriteLine($"[PROCESSING] {fileName}")
    Dim pdf = PdfDocument.FromFile(filePath)
    Dim summary As String = Await pdf.Summarize()

    cacheManager.CacheResult(fileName, fileHash, summary)
    processed += 1
Next

Console.WriteLine(vbCrLf & $"Processing complete: {cached} cached, {processed} newly processed")
Console.WriteLine($"Cost savings: {(cached * 100.0 / Math.Max(1, cached + processed)):F1}% served from cache")

' Hash-based cache manager with JSON index
Public Class DocumentCacheManager

    Private ReadOnly _cacheFolder As String
    Private ReadOnly _indexPath As String
    Private _index As Dictionary(Of String, CacheEntry)

    Public Sub New(cacheFolder As String)
        _cacheFolder = cacheFolder
        _indexPath = Path.Combine(cacheFolder, "cache-index.json")
        _index = LoadIndex()
    End Sub

    Private Function LoadIndex() As Dictionary(Of String, CacheEntry)
        If File.Exists(_indexPath) Then
            Dim json As String = File.ReadAllText(_indexPath)
            Return JsonSerializer.Deserialize(Of Dictionary(Of String, CacheEntry))(json) OrElse New Dictionary(Of String, CacheEntry)()
        End If
        Return New Dictionary(Of String, CacheEntry)()
    End Function

    Private Sub SaveIndex()
        Dim json As String = JsonSerializer.Serialize(_index, New JsonSerializerOptions With {.WriteIndented = True})
        File.WriteAllText(_indexPath, json)
    End Sub

    ' SHA256 hash to detect file changes
    Public Function ComputeFileHash(filePath As String) As String
        Using sha256 = SHA256.Create()
            Using stream = File.OpenRead(filePath)
                Dim hash As Byte() = sha256.ComputeHash(stream)
                Return Convert.ToHexString(hash)
            End Using
        End Using
    End Function

    Public Function GetCachedResult(fileName As String, currentHash As String) As String
        If _index.TryGetValue(fileName, entry) Then
            If entry.FileHash = currentHash AndAlso File.Exists(entry.CachePath) Then
                entry.LastAccessed = DateTime.UtcNow
                SaveIndex()
                Return File.ReadAllText(entry.CachePath)
            End If
        End If
        Return Nothing
    End Function

    Public Sub CacheResult(fileName As String, fileHash As String, result As String)
        Dim cachePath As String = Path.Combine(_cacheFolder, $"{Path.GetFileNameWithoutExtension(fileName)}-{fileHash.Substring(0, 8)}.txt")
        File.WriteAllText(cachePath, result)

        _index(fileName) = New CacheEntry With {
            .FileHash = fileHash,
            .CachePath = cachePath,
            .CreatedAt = DateTime.UtcNow,
            .LastAccessed = DateTime.UtcNow
        }

        SaveIndex()
    End Sub

End Class

Public Class CacheEntry

    Public Property FileHash As String = ""
    Public Property CachePath As String = ""
    Public Property CreatedAt As DateTime
    Public Property LastAccessed As DateTime

End Class

using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using System.Collections.Generic;
using System.Text.Json;
using System.Text;

// Compare financial metrics across multiple company filings for sector analysis

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

// Analyze company filings
string[] companyFilings = {
    "filings/company-a-10k.pdf",
    "filings/company-b-10k.pdf",
    "filings/company-c-10k.pdf"
};

var sectorData = new List<CompanyFinancials>();

foreach (string filing in companyFilings)
{
    Console.WriteLine($"Analyzing: {Path.GetFileName(filing)}");

    var pdf = PdfDocument.FromFile(filing);

    // Define JSON schema for 10-K extraction (numbers in millions USD)
    string extractionQuery = @"Extract key financial metrics from this 10-K filing. Return JSON:

mpanyName"": ""string"",
scalYear"": ""string"",
venue"": number,
venueGrowth"": number,
ossMargin"": number,
eratingMargin"": number,
tIncome"": number,
s"": number,
talDebt"": number,
shPosition"": number,
ployeeCount"": number,
yRisks"": [""string""],
idance"": ""string""


in millions USD. Growth/margins as percentages.
NLY valid JSON.";

    string result = await pdf.Query(extractionQuery);

    try
    {
        var financials = JsonSerializer.Deserialize<CompanyFinancials>(result);
        if (financials != null)
            sectorData.Add(financials);
    }
    catch
    {
        Console.WriteLine($"  Warning: Could not parse financials for {filing}");
    }
}

// Generate sector comparison report
var report = new StringBuilder();
report.AppendLine("=== Sector Analysis Report ===\n");

report.AppendLine("Revenue Comparison (millions USD):");
foreach (var company in sectorData.OrderByDescending(c => c.Revenue))
    report.AppendLine($"  {company.CompanyName}: ${company.Revenue:N0} ({company.RevenueGrowth:+0.0;-0.0}% YoY)");

report.AppendLine("\nProfitability Margins:");
foreach (var company in sectorData.OrderByDescending(c => c.OperatingMargin))
    report.AppendLine($"  {company.CompanyName}: {company.GrossMargin:F1}% gross, {company.OperatingMargin:F1}% operating");

report.AppendLine("\nFinancial Health (Debt vs Cash):");
foreach (var company in sectorData)
{
    double netDebt = company.TotalDebt - company.CashPosition;
    string status = netDebt < 0 ? "Net Cash" : "Net Debt";
    report.AppendLine($"  {company.CompanyName}: {status} ${Math.Abs(netDebt):N0}M");
}

string reportText = report.ToString();
Console.WriteLine($"\n{reportText}");
File.WriteAllText("sector-analysis-report.txt", reportText);

// Save full JSON data
string outputJson = JsonSerializer.Serialize(sectorData, new JsonSerializerOptions { WriteIndented = true });
File.WriteAllText("sector-analysis.json", outputJson);

Console.WriteLine("Analysis saved to sector-analysis.json and sector-analysis-report.txt");


s CompanyFinancials

public string CompanyName { get; set; } = "";
public string FiscalYear { get; set; } = "";
public double Revenue { get; set; }
public double RevenueGrowth { get; set; }
public double GrossMargin { get; set; }
public double OperatingMargin { get; set; }
public double NetIncome { get; set; }
public double Eps { get; set; }
public double TotalDebt { get; set; }
public double CashPosition { get; set; }
public int EmployeeCount { get; set; }
public List<string> KeyRisks { get; set; } = new();
public string Guidance { get; set; } = "";

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.Collections.Generic
Imports System.Text.Json
Imports System.Text
Imports System.IO

' Compare financial metrics across multiple company filings for sector analysis

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

' Analyze company filings
Dim companyFilings As String() = {
    "filings/company-a-10k.pdf",
    "filings/company-b-10k.pdf",
    "filings/company-c-10k.pdf"
}

Dim sectorData = New List(Of CompanyFinancials)()

For Each filing As String In companyFilings
    Console.WriteLine($"Analyzing: {Path.GetFileName(filing)}")

    Dim pdf = PdfDocument.FromFile(filing)

    ' Define JSON schema for 10-K extraction (numbers in millions USD)
    Dim extractionQuery As String = "Extract key financial metrics from this 10-K filing. Return JSON:

mpanyName"": ""string"",
scalYear"": ""string"",
venue"": number,
venueGrowth"": number,
ossMargin"": number,
eratingMargin"": number,
tIncome"": number,
s"": number,
talDebt"": number,
shPosition"": number,
ployeeCount"": number,
yRisks"": [""string""],
idance"": ""string""


in millions USD. Growth/margins as percentages.
NLY valid JSON."

    Dim result As String = Await pdf.Query(extractionQuery)

    Try
        Dim financials = JsonSerializer.Deserialize(Of CompanyFinancials)(result)
        If financials IsNot Nothing Then
            sectorData.Add(financials)
        End If
    Catch
        Console.WriteLine($"  Warning: Could not parse financials for {filing}")
    End Try
Next

' Generate sector comparison report
Dim report = New StringBuilder()
report.AppendLine("=== Sector Analysis Report ===" & vbCrLf)

report.AppendLine("Revenue Comparison (millions USD):")
For Each company In sectorData.OrderByDescending(Function(c) c.Revenue)
    report.AppendLine($"  {company.CompanyName}: ${company.Revenue:N0} ({company.RevenueGrowth:+0.0;-0.0}% YoY)")
Next

report.AppendLine(vbCrLf & "Profitability Margins:")
For Each company In sectorData.OrderByDescending(Function(c) c.OperatingMargin)
    report.AppendLine($"  {company.CompanyName}: {company.GrossMargin:F1}% gross, {company.OperatingMargin:F1}% operating")
Next

report.AppendLine(vbCrLf & "Financial Health (Debt vs Cash):")
For Each company In sectorData
    Dim netDebt As Double = company.TotalDebt - company.CashPosition
    Dim status As String = If(netDebt < 0, "Net Cash", "Net Debt")
    report.AppendLine($"  {company.CompanyName}: {status} ${Math.Abs(netDebt):N0}M")
Next

Dim reportText As String = report.ToString()
Console.WriteLine(vbCrLf & reportText)
File.WriteAllText("sector-analysis-report.txt", reportText)

' Save full JSON data
Dim outputJson As String = JsonSerializer.Serialize(sectorData, New JsonSerializerOptions With {.WriteIndented = True})
File.WriteAllText("sector-analysis.json", outputJson)

Console.WriteLine("Analysis saved to sector-analysis.json and sector-analysis-report.txt")

Public Class CompanyFinancials
    Public Property CompanyName As String = ""
    Public Property FiscalYear As String = ""
    Public Property Revenue As Double
    Public Property RevenueGrowth As Double
    Public Property GrossMargin As Double
    Public Property OperatingMargin As Double
    Public Property NetIncome As Double
    Public Property Eps As Double
    Public Property TotalDebt As Double
    Public Property CashPosition As Double
    Public Property EmployeeCount As Integer
    Public Property KeyRisks As List(Of String) = New List(Of String)()
    Public Property Guidance As String = ""
End Class