This loader extracts the text from an image that has text in it, e.g. a receipt (key-value pairs) or a plain text image. If the image has plain text, the loader uses pytesseract. If image has text in key-value pairs like an invoice, the Donut transformer model is used. The file extensions .png, .jpg, and .jpeg are preferred. A single local file is passed in each time you call load_data.

To use this loader, you need to pass in a Path to a local file.

from pathlib import Path
from llama_index import download_loader

ImageReader = download_loader("ImageReader")

# If the Image has key-value pairs text, use text_type = "key_value"
loader = ImageReader(text_type = "key_value")
documents = loader.load_data(file=Path('./receipt.png'))

# If the Image has plain text, use text_type = "plain_text"
loader = ImageReader(text_type = "plain_text")
documents = loader.load_data(file=Path('./image.png'))

This loader is designed to be used as a way to load data into LlamaIndex and/or subsequently used as a Tool in a LangChain Agent. See here for examples.