Parse Image

Convert text and visual content in images into document records for extraction, text processing, and knowledge base indexing.

Use this node for receipt screenshots, product images, or scans. Captions describe images in text. To index the images themselves, use Image Embedding.

Configure image parsing

Bind image sources from Read MOI Volume. File references must point to actual images.

Setting

Description

Image processing methods

ocr extracts text; caption generates descriptions. Both are enabled by default

Caption language

Language of the description; configuration default: zh

VLM OCR model

Vision-language model used for OCR and captions

Choose OCR for receipt text, captions for visual descriptions, or both to include both kinds of content.

Use the output

The main output is documents, with type image and source metadata. Use text for plain text; this field is omitted when empty.

Pass the full document array to extraction or indexing to retain source information.

Example: extract receipt fields

Flow: Read MOI Volume → Parse Image → Information Extraction.

Choose OCR, bind documents to Information Extraction, and define receipt number, date, and amount in its schema. Parse Image supplies the text; Information Extraction produces the structured fields.

Last updated on