AI Extract Document Node

A

AI/Processing

Extracts text and content from documents using AI for enhanced image descriptions and OCR.

ai_processing_extract_document_aiprocessingLong running

Inputs13

Outputs2

Security exposure5/10

Packageprocessing

Ratings

Scores range from 0 to 10. Higher values mean more impact, exposure, or operational weight.

SecurityAttack surface and exposure impact.

5/10Medium

PrivacyPotential sensitivity of processed data.

5/10Medium

PerformanceRuntime or resource pressure.

6/10Medium

GovernancePolicy, audit, or compliance impact.

5/10Medium

ReliabilityOperational stability considerations.

7/10High

CostExternal or compute cost impact.

4/10Medium

Input Pins

13

Input

Execution

exec_in

Execution trigger to start AI-powered document extraction.

File

Struct

file

Document file to extract (PDF, DOCX, XLSX, images, etc.).

SchemaFlowPathFlowPathFlowPath3 fields

pathstringrequired

store_refstringrequired

cache_store_refstring | null

Schema enforced

Model

Struct

model

Vision-capable AI model for image analysis and OCR.

SchemaBitBitBit19 fields

idstring

default ""

typeBitTypes

enum "Llm", "Vlm", "Tts", "Stt"...default "Other"

metaMap<string, Metadata>

default {}

*Metadatamap value

namestringrequired

descriptionstringrequired

long_descriptionstring | null

release_notesstring | null

tagsArray<string>required

itemsstringarray item

+11 more fields

authorsArray<string>

default []

itemsstringarray item

repositorystring | null

default null

download_linkstring | null

default null

file_namestring | null

default null

hashstring

default ""

sizeinteger | null

format uint64default nullmin 0

hubstring

default ""

parametersvalue

default null

versionstring | null

default null

licensestring | null

default null

dependenciesArray<string>

default []

itemsstringarray item

dependency_tree_hashstring

default ""

createdstring

default ""

updatedstring

default ""

model_slugstring | null

default null

+1 more fields

Schema enforced

Extract Images

Boolean

extract_images

Whether to extract and embed images from the document.

Default true

Images Per Message

Integer

images_per_message

Number of images to batch per LLM request (higher = faster but may hit token limits).

Default 1

Pages Per Batch

Integer

pages_per_batch

Number of PDF pages to process in parallel (higher = faster but uses more memory).

Default 4

Temperature

Float

temperature

LLM temperature (0.0 = deterministic, 1.0 = creative). Lower is better for extraction.

Default 0.1

Max Tokens

Integer

max_tokens

Maximum output tokens per LLM call. Leave at 0 for model default. Set lower for unreliable models.

Default 0

Prompt Preset

String

prompt_preset

Prompt contract of the selected model. Document-parsing models only answer to their own trained prompt: • Default — a tuned general prompt for vision models (GPT-4o, Claude, Gemini, Qwen-VL) • Unlimited-OCR — baidu/Unlimited-OCR, self-hosted via vLLM • DeepSeek-OCR — deepseek-ai/DeepSeek-OCR and -OCR-2 • olmOCR — allenai/olmOCR-2, emits YAML front matter • Nanonets-OCR — nanonets/Nanonets-OCR-s and -OCR2 • dots.ocr — plain text extraction; use Page Prompt for its JSON layout mode • Granite-Docling — IBM Granite-Docling and SmolDocling, emits DocTags • PaddleOCR-VL — PaddlePaddle/PaddleOCR-VL Every preset except Default forces full-page OCR and one image per request. Recommended temperature is 0.0 for all of them except olmOCR (0.1).

Default Default

DefaultUnlimited-OCRDeepSeek-OCRolmOCRNanonets-OCRdots.ocr+2 more

Page Prompt

String

page_prompt

Prompt for converting a rendered document page to text. Switching the preset rewrites this unless you have typed your own. Leave empty to fall back to the preset.

Image Prompt

String

image_prompt

Prompt for describing a standalone or embedded image. Switching the preset rewrites this unless you have typed your own. Leave empty to fall back to the preset.

Batch Image Prompt

String

batch_prompt

Prompt used when Images Per Message is greater than 1. Only the Default preset fills this in — the OCR presets never batch. Leave empty to fall back to the preset.

Force OCR

Boolean

force_ocr

Run every PDF page through the model instead of only pages whose extracted text looks poor. Presets other than Default turn this on regardless.

Default false

Output Pins

2

Output

Execution

exec_out

Execution output after extraction completes.

Pages

Struct Array

pages

Extracted document pages with AI-generated descriptions and images.

SchemaDocumentPageDocumentPageDocumentPage3 fields

page_numberinteger:uint32required

format uint32min 0

contentstringrequired

imagesArray<NodeImage>required

itemsNodeImagearray item

image_refstringrequired

Schema enforced

Node Info

Internal name

ai_processing_extract_document_ai

Category

AI/Processing

Version

3