A simple python script to rename research PDF files based on their content.

Leverages pdfminer to extract text and GPT to generate the filename.

Why note use the Title from the PDF metadata? Because it's often missing: in my personal collection of research papers, only 44% of the PDFs have a Title metadata field.

Isn't this expensive? In my personal collection of research papers, renaming each PDF uses ~2.1K tokens on average. At gpt-3.5-turbo-0125 current cost of $0.0005 / 1K tokens, that means renaming each PDF costs ~$0.001 (1 tenth of a penny). I think it's worth it!

Set your OpenAI API key in pdfrename.py:

openai.api_key = "YOUROPENAIKEY"

Then run:

pip install -r requirements.txt
python pdfrename.py filetorename.pdf

To run recursively on a directory:

find whichdirectory -name "*.pdf" | parallel -j 10 python pdfrename.py 

In the folderaction.workflow directory there is a MacOS Automator workflow that can be used to monitor a folder for new pdfs and automatically rename them.

Change PATHTOCODE in folderaction.workflow/document.wflow to point to the location of the pdfrename.py script:

<key>ActionParameters</key>
				<dict>
					<key>COMMAND_STRING</key>
					<string>for f in "$@"
do
	if [[ -d "$f" ]]; then
		continue
	fi
	python PATHTOCODE/pdfrename.py "$f"
done</string>

To install, copy the folderaction.workflow to ~/Library/Workflows/Applications/Folder Actions and then right click on a folder in Finder and select Services -> Folder Actions Setup... and then select the folderaction.workflow from the list of available workflows.

Copyright (c) 2024 Salle, Alexandre [email protected]. All work in this package is distributed under the MIT License.