Class IngestionColumns

Namespace

Qavren.Edge.Ingestion

Assembly

Qavren.Edge.Ingestion.dll

Every storage name as a const, plus the breadcrumb separator and the sanitiser that makes the join/split round-trip (spec 8.3). The sanitiser lives beside the separator on purpose: the round-trip is a property of the PAIR, and splitting them across two files is how the two drift.

public static class IngestionColumns

Inheritance

object

IngestionColumns

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Fields

BlockKind

The chunk's structural kind (paragraph, heading, list item, and so on).

CharEnd

The chunk's ending character offset within the document.

CharStart

The chunk's starting character offset within the document.

ContentHash

The document's content hash, used to detect unchanged documents between runs.

CreatedUtc

When the chunk was written, in UTC.

DocumentId

The document's id within its source.

Embedding

The chunk's embedding vector column.

ExtractorId

The id of the IDocumentExtractor that produced the chunk.

HeadingPath

The chunk's rendered heading-path breadcrumb.

HeadingPathSeparator

" \u203A " — U+203A between two spaces. Joins and splits heading_path; headings are sanitised against it at chunk time so the split round-trips (spec 8.3).

Key

The vector store's primary key column.

MediaType

The document's media type.

Ordinal

The chunk's position within its document.

Page

The page number the chunk came from, for paginated formats such as PDF.

SourceId

The id of the IngestionSource the document came from.

Text

The chunk's text.

TokenCount

The chunk's token count.

Methods

RenderBreadcrumb(IReadOnlyList<string>)

The rendered breadcrumb for a structural path, or null when the path is empty.

SanitizeHeading(string)

Makes one heading safe to join: control characters go, whitespace runs collapse to a single space, every occurrence of HeadingPathSeparator collapses to a single space, and the result is trimmed. Heading text is display text, so this is lossless in every way a reader would notice and lossy in exactly the way that would otherwise corrupt a boundary. Idempotent, and culture-invariant throughout.

SplitBreadcrumb(string?)

The structural path a stored breadcrumb splits back into. Empty when null.