Table of Contents

Class DocumentToMarkdownPageResult

Namespace
LMKit.Document.Conversion
Assembly
LM-Kit.NET.dll

Represents the Markdown conversion outcome for a single page of a document.

public sealed class DocumentToMarkdownPageResult
Inheritance
DocumentToMarkdownPageResult
Inherited Members

Examples

Flag low-quality VLM pages for re-processing.

using LMKit.Document.Conversion;

var converter = new DocumentToMarkdown();
var result = converter.Convert("report.pdf");


foreach (var page in result.Pages)
{
if (page.StrategyUsed == DocumentToMarkdownStrategy.VlmOcr &&
page.QualityScore.HasValue && page.QualityScore.Value < 0.6)
{
Console.WriteLine($"Review page {page.PageNumber} (quality {page.QualityScore:F2}).");
}
}

Remarks

Instances of this class are created by DocumentToMarkdown and exposed through Pages. They carry both the per-page Markdown body and diagnostics about which strategy handled the page.

GeneratedTokenCount and QualityScore are populated only when the page was transcribed by the vision-language model. Use them to detect pages that hit the completion-token cap or flag pages whose quality score suggests the conversion should be re-run at higher fidelity.

Properties

Certainty

Gets a confidence score in the [0, 1] range that the page's Markdown faithfully represents its source. A value of 1.0 means the converter is very confident the output is a correct and complete rendering; values below 0.70 are worth reviewing or routing to a more thorough pipeline.

Elapsed

Gets the wall-clock time spent processing this page.

GeneratedTokenCount

Gets the number of tokens emitted by the vision model for this page, or 0 when the page was handled by the text-extraction strategy.

HasExtractableText

Gets a value indicating whether the source page exposed an extractable text layer at the time of conversion.

Markdown

Gets the Markdown content produced for this page. May be empty when the page contains no textual content or when the conversion could not recover any text.

PageIndex

Gets the zero-based index of the page within the source document.

PageNumber

Gets the 1-based page number, convenient for user-facing output.

QualityScore

Gets the quality score reported by the vision model for this page, or null when the page was handled by the text-extraction strategy.

StrategyUsed

Gets the strategy that was actually applied to this page. When the converter runs in Hybrid, this value reflects the per-page decision (either TextExtraction or VlmOcr).

Warning

Gets an optional warning message associated with this page (for example, a notice that a page was empty because no vision model was available).

Share