llm-pdf-to-images

LLM fragment plugin to load a PDF as a sequence of images

Installation

Install this plugin in the same environment as LLM.

llm install llm-pdf-to-images

The llm-pdf-to-images plugin provides a fragment loader that converts each page of a PDF document into an image attachment.

You can use the pdf-to-images: fragment prefix to convert a PDF file into a series of image attachments which can be sent to a model.

Example usage:

llm -f pdf-to-images:path/to/document.pdf 'Summarize this document'

Fragment syntax

pdf-to-images:<path>?dpi=N&format=jpg|png&quality=Q

• <path>: Path to the PDF file accessible to the environment where LLM runs.
• dpi=N: (optional) Dots per inch to use when rendering the PDF pages, which affects the resolution of the output images. Defaults to 300 if omitted.
• format=jpg|png: (optional) Image format to use for the output. Can be either jpg (default) or png.
• quality=Q: (optional) JPEG quality factor between 1 and 100. Only applies when using JPG format. Defaults to 30 if omitted. Higher values produce better quality but larger file sizes.

More examples

Convert a PDF file to images with default settings (300 DPI, JPG format, quality 30):

llm -f pdf-to-images:document.pdf 'summarize this document'

Convert a PDF with higher resolution (600 DPI):

llm -f 'pdf-to-images:document.pdf?dpi=600' 'summarize'

Convert a PDF to PNG format:

llm -f 'pdf-to-images:document.pdf?format=png' 'describe all figures'

Convert a PDF with high-quality JPG images:

llm -f 'pdf-to-images:document.pdf?quality=90' 'extract all visible text'

Combine multiple parameters:

llm -f 'pdf-to-images:document.pdf?dpi=450&format=jpg&quality=75' 'OCR'

Development

To set up this plugin locally, first checkout the code. Then create a new virtual environment:

cd llm-pdf-to-images
python -m venv venv
source venv/bin/activate

Now install the dependencies and test dependencies:

python -m pip install -e '.[test]'

To run the tests:

python -m pytest