Structured Text Block

Structured blocks store data in structured formats (i.e., JSON Objects or Lists). They’re ideal for complex inputs, outputs, or intermediate data passed through processing nodes.

Key features:

• Stores nested structured data that can be navigated via keys or indexes.
• Supports creating multiple views for the same content, with different arrangements.

Examples:

• JSON Object: { "name": "Alice", "role": "Manager" }
• JSON Array: [ {"id": 1, "item": "pen"}, {"id": 2, "item": "book"} ]

All views follow a unified JSON format with two fields:

• content: The semantic core
• metadata: Extra info to guide downstream tasks

Loop Mode：

[Loop Mode Button Icon]

• What is Loop Mode?： Loop Mode enables concurrent processing of each element inside a structured block. It is ideal when the structured content contains multiple items (e.g., a list of records) and you want the connected edge to process each item individually. When enabled, the edge execution will iterate over each element of the block content.
• Why we use Loop Mode?：Useful for edge types (like LLM) that have context-length limitations or cannot handle large input sizes.
• This mode is disabled by default and must be toggled manually when needed.

Index Management：

[Index Management Button Icon]

• Used to prepare content for embedding by isolating the semantic content and attaching meaningful metadata.
• Clicking Add New Index button to create a new Index.
• Data Preview Contains content and metadata .
• Key Path: Press Add Path Segment and enter the fields（content or metadata）you want to be indexed.
  • You must add a path segment.
• Value Path: Press Add Path Segment and enter the fields and enter the fields（content or metadata）you want to be matched.
  • Default: Return whole chunk contains content and metadata .
• Tips: You can’t edit your index, but you can create a new index to replace it.