API Reference
ReaderHierarchicalChunkingOptions
Options for bounded, token-aware hierarchical chunking.
Inheritance
- Object
- ReaderHierarchicalChunkingOptions
Usage
This type appears in these public API surfaces even when no hand-authored example is attached directly to the page.
Accepted by parameters
- Method DocumentReader.ChunkDocument
- Method DocumentReader.ReadHierarchical
- Method DocumentReader.ReadHierarchical
- Method DocumentReader.ReadHierarchical
- Method DocumentReader.ReadHierarchicalAsync
- Method DocumentReader.ReadHierarchicalAsync
- Method DocumentReader.ReadHierarchicalAsync
- Method OfficeDocumentReader.ChunkDocument
- Method OfficeDocumentReader.ReadHierarchical
- Method OfficeDocumentReader.ReadHierarchical
- Method OfficeDocumentReader.ReadHierarchical
- Method OfficeDocumentReader.ReadHierarchicalAsync
- Method OfficeDocumentReader.ReadHierarchicalAsync
- Method OfficeDocumentReader.ReadHierarchicalAsync
- Method ReaderHierarchicalChunker.Chunk
- Method ReaderHierarchicalChunker.Chunk
Constructors
public ReaderHierarchicalChunkingOptions() #Inherited Methods
public override Boolean Equals(Object obj) #BooleanParameters
- obj Object
Properties
public Int32 MaxTokens { get; set; } #Maximum tokens in each output chunk, including optional context. Default: 800.
public Int32 OverlapTokens { get; set; } #Maximum content tokens repeated from the previous segment. Default: 80.
public Int32 MaxInputChunks { get; set; } #Maximum source chunks inspected. Default: 10,000.
public Int32 MaxOutputChunks { get; set; } #Maximum output chunks emitted. Default: 50,000.
public Int32 MaxHierarchyDepth { get; set; } #Maximum container and heading hierarchy depth below the document root. Default: 32.
public Int32 MaxContextCharacters { get; set; } #Maximum characters retained in one hierarchy breadcrumb or title. Default: 4,096.
public Boolean PreferMarkdown { get; set; } #Prefer a source chunk's Markdown over plain text when available. Default: true.
public Boolean IncludeContextInText { get; set; } #Prefix output text with its container and heading breadcrumb when it fits the token budget. Default: true.
public IReaderTokenCounter TokenCounter { get; set; } #Token counter. Defaults to the dependency-free OfficeIMO heuristic.