Split text

Split text into chunks.

Example
The example above illustrates a Flow that stores a document in SQL Server. It uses the Split text action to divide a large document into smaller chunks. Then, for each chunk, an embedding is generated, and the text chunk along with its corresponding embedding is stored in SQL Server as a vector record.

Properties

Name	Required	Description
Title	No	The name of the action or operation.
Input text	Yes	The text to split into chunks.
Splitter type	Yes	Read documentation below.
Text chunk variable name	No	The variable name used to store the output chunks.
Description	No	A field to describe the action or its purpose.

Splitter type

Required. Allows selection of a splitting type (algorithm). Based on the selected type, a number of additional properties will be shown.

Recursive character text splitter:

Splits text by characters into chunks of a specified size, optionally allowing overlap for better context retention.

Additional properties

Name	Required	Description
Chunk size	Yes	Maximum number of characters in each chunk.
Chunk overlap	No	Number of overlapping characters between chunks.

Token text spiltter:

Divides text based on token count using a chosen encoding, useful for models with token limits.

Additional properties

Name	Required	Description
EncodingName	Yes	The name of the token encoding to be used.
Max tokens	Yes	The maximum number of tokens per chunk.
Chunk overlap	No	Number of overlapping tokens between chunks.

Markdown header text splitter:

Breaks down documents at specific Markdown headers, ideal for structured texts like articles or reports.

Additional properties

Name	Required	Description
Headers to split on	No	Comma-separated list of Markdown headers to use as split points.
Include headers in output	No	If enabled, includes the header text in each chunk.

Returns

Split text returns string textChunk.