The most accurate Markdown engine. Also the fastest.

import base64, requests

doc = base64.b64encode(open("contract.pdf", "rb").read()).decode()

r = requests.post(
    "http://localhost:5189/lmkit/v1/document-to-markdown",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={"input": doc, "strategy": "TextExtraction", "enable_ocr": True},
)
print(r.json())

{
  "mcpServers": {
    "lmkit": {
      "url": "http://localhost:5189/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}
// Agents ingest a file, then call document_convert_to_markdown.

using LMKit.Document.Conversion;
using LMKit.Extraction.Ocr;

var converter = new DocumentToMarkdown();
var options = new DocumentToMarkdownOptions
{
    Strategy  = DocumentToMarkdownStrategy.TextExtraction,
    OcrEngine = new LMKitOcr()
};

var result = await converter.ConvertAsync("report.pdf", options);
Console.WriteLine(result.Markdown);
Console.WriteLine($"certainty {result.Certainty:F2}");

var review = new List<int>();

converter.PageCompleted += (_, e) =>
{
    var page = e.PageResult;
    if (page != null && page.Certainty < 0.70)
        review.Add(page.PageNumber);     // only these pages need a person
};

var result = await converter.ConvertAsync("scanned-batch.pdf");

var options = new DocumentToMarkdownOptions
{
    EmitFrontMatter                  = true,   // provenance on every file
    PreferMarkdownTablesForNonNested = true
};

foreach (var path in Directory.EnumerateFiles("corpus"))
{
    var r = await converter.ConvertAsync(path, options);
    if (r.Certainty < 0.70) { reviewQueue.Add(path); continue; }
    rag.ImportText(r.Markdown, chunking, "corpus", path);
}