What is ChatWithPDF?
ChatWithPDF is a ChatGPT plugin that enables users to search through small or large PDF documents directly within ChatGPT.
How to use ChatWithPDF
To use ChatWithPDF, simply provide a publicly accessible PDF URL for loading and processing. You can use a publicly accessible Google Drive link to a PDF document (no photocopies) or upload your documents to a service like tmpfiles.org and use that while chatting with ChatWithPDF.
Features of ChatWithPDF
ChatWithPDF offers various features to enhance your experience with PDF documents. These features include:
- Searching PDFs directly within ChatGPT
- Searching PDFs of any size
- Searching PDFs in any language
- Searching PDFs on any topic
- Searching PDFs of any length
- Searching PDFs of any quality
How ChatWithPDF works
Users provide a publicly accessible PDF URL for loading and processing (should be accessible from Incognito mode, for example). The plugin downloads and processes the PDF document, extracting relevant information. User queries are compared with the processed information from the PDF, and the most relevant matches are returned and displayed to the user.
Privacy of ChatWithPDF
ChatWithPDF does not store data permanently. All PDFs are embedded and immediately removed. Embeddings are stored with ChromaDB on the same deployment server and are wiped with each new deployment. Usually, embeddings are manually purged every 12-24 hours due to size and memory constraints. In the future, all embeddings will be deleted 1 hour after the user requests them. If someone has your link, they can fetch embeddings from it since it is cached in the vector database, but users cannot see what other users have searched or embedded.