Chunking

Understand how Orka JS splits documents into optimally-sized chunks for embedding andretrieval.

How Chunking Works

Orka JS uses a recursive text splitter that intelligently cuts text at natural boundaries:

\n\n

Paragraph Breaks

Preserves natural structure first.

Step 1

\n

Line Breaks

Split by lines if paragraphs are too long.

Step 2

.

Sentences

Targets logical thought boundaries.

Step 3

Words

Last resort to avoid cutting a word in half.

Step 4

char

Characters

Strict limit enforcement (Emergency).

Step 5

Configuration

await orka.knowledge.create({
  name: 'docs',
  source: myContent,
  chunkSize: 1000,     // Max characters per chunk
  chunkOverlap: 200,   // Overlap between consecutive chunks
});

Recommended Sizes

Content Profile	Dimensions (Size / Overlap)	Strategic Goal
FAQ / Q&A	Size: 300Ovp: 50	Atomic precision
Technical Docs	Size: 1000Ovp: 200	Block Integrity
Long Articles	Size: 1200Ovp: 250	Narrative Flow
Legal / Contracts	Size: 600Ovp: 200	Clause context

💡 Why Overlap?

Overlap ensures information at chunk boundaries isn't lost. When a chunk ends mid-sentence, the next chunk starts a few hundred characters earlier, capturing the full context.