PDF Search Engine with Jina

This search engine will index a folder of PDF files, break them down into chunks, and then let you search using text or image for relevant chunks. In the frontend you'll see the returned chunks with a link to their associated PDF file.

Bear in mind this is a work in progress

Run the example

1. Create a virtual environment
2. pip install -r requirements.txt

Backend

1. cd backend
2. python app.py -t index to index your documents
3. python app.py -t search to open a RESTful search interface

Frontend

1. cd frontend
2. streamlit run frontend.py

Configure

• Settings are in config.yml
• Additional frontend settings are in frontend/.streamlit/config.toml

Deploy

docker-compose up

Test

If you're planning to use "Print to PDF" from your web browser for testing, I recommend using Chrome over Firefox. Firefox converts characters strangely (for example fi becomes fi) which could affect search results depending on what the encoder recognizes as a meaningful unit.

TODO

Feel free to make a PR to help out with these!

• [ ] Test image display in frontend
• [ ] Embed original image as datauri when indexing images
• [ ] Single restful endpoint for both indexing and search
• [ ] Working docker-compose.yml