Skip to main content
Data Ingestion

Chunking and Vectorization

Understand how chunking strategies and embedding models affect retrieval quality

Chunking and vectorization are performed automatically during data import. Chunking determines the smallest unit for retrieval, while vectorization determines the precision of semantic matching. Both are configured in the index settings step of Importing Data and cannot be modified after creation.

Chunking Methods

The following chunking methods are available when importing data:
Chunking MethodDescriptionUse Cases
Smart Chunking (default)Automatically detects section and paragraph boundariesMost scenarios
Fixed-Length ChunkingSplits text into equal-length chunks based on a fixed token countScenarios with strict token-count requirements
Page-Based ChunkingSplits by PDF page boundariesDocuments where each page conveys an independent topic
Heading-Based ChunkingSplits by Markdown / HTML heading levelsDocuments where headings demarcate independent topics
Regex-Based ChunkingSplits at positions matched by a custom regular expressionText with consistent delimiters
Delimiter-Based ChunkingSplits using specified delimiter charactersDocuments where content is separated by specific identifiers
Maximum Chunk Length: Default 600; range 10–6000 tokens.

Chunking Selection Guidelines

ScenarioRecommended ConfigurationReason
FAQ DocumentsFixed-Length Chunking, 256 tokensQ&A pairs are short and self-contained
Product ManualsSmart Chunking, 600 tokens (default)Clear chapter structure
Long ReportsSmart Chunking, 2048 tokensPreserves more contextual information
Legal ContractsHeading-Based Chunking, 512 tokensClauses are clearly organized by heading hierarchy

Vectorization

After chunking, the system uses an embedding model to convert text chunks into vectors and writes them to the index. The embedding model is selected when creating the knowledge base and cannot be changed afterward. During data import, the cost details panel on the right displays the currently used embedding model and reranking model.

Viewing and Managing Chunks

Once chunking is complete, view and manage chunks in the Chunk Details tab on the knowledge base detail page. For more information, see Document Management — Chunk Management.

API Management

Chunk-related APIs:
After configuring chunking and vectorization, create a Retrieval Service to begin hybrid retrieval.