DocumentBlock carries a document as bytes or extracted text. It also keeps the MIME type and an optional display name so codecs can choose the provider’s document representation.

API surface

type DocumentBlock struct {
	MediaType MediaType
	Name      string
	Data      []byte
	Text      string
}

Field	Use
`MediaType`	`MediaTypeDocumentPDF`, `MediaTypeDocumentText`, `MediaTypeDocumentHTML`, `MediaTypeDocumentCSV`, `MediaTypeDocumentMarkdown`, `MediaTypeDocumentDOCX`, or `MediaTypeDocumentXLSX`
`Name`	Filename or user-facing label
`Data`	Binary document bytes
`Text`	Extracted text when the caller already decoded the document

Either Data or Text may be populated depending on how the document arrived. Core does not impose a mutual-exclusion validator; the codec decides what its wire dialect supports.

Example

package main

import "github.com/looprig/core/content"

func main() {
	block := &content.DocumentBlock{
		MediaType: content.MediaTypeDocumentText,
		Name:      "report.txt",
		Text:      "Revenue increased.",
	}
	wire, err := content.MarshalBlock(block)
	if err != nil {
		panic(err)
	}
	decoded, err := content.UnmarshalBlock(wire)
	if err != nil {
		panic(err)
	}
	_ = decoded.(*content.DocumentBlock)
}

Nested document blocks inside a tool result are found by the same recursive block codec used for top-level blocks.

Proof

Source: core/content/block.go, core/content/media_type.go
Tests: core/content/block_json_test.go
Example: core/examples/content/example_test.go