How to get the contents of the nodes

[sacsac944]

Does Pageindex get the contents of the nodes as well or should it be done using some other framework? In the examples, the node content is not present in the tree and we might not have access to the source document during runtime.

Also does it support multiple modalities like tables and images apart from text in the documents?

[zmtomorrow]

Hi, currently you could do that by using the page number to extract the content from the original pdf. We also plan to add that function from our API end.

[KylinMountain]

Update since @zmtomorrow's reply — node content is built in now, so you don't need to re-extract from the PDF yourself.

Enable it when indexing:

• CLI: python run_pageindex.py --pdf_path doc.pdf --if-add-node-text yes
• config.yaml: if_add_node_text: "yes"
• SDK / page_index(...): if_add_node_text='yes'

Each node then carries its text, a summary (on by default), and start_index / end_index page anchors — so the source document isn't needed at query time.

On modalities: the open-source pipeline is text-based (page text via PyMuPDF), so tables come through as extracted text and images aren't captured as such. For richer table / image / scanned-document handling, the hosted service uses PageIndex OCR, and there are vision / OCR cookbooks at https://docs.pageindex.ai/cookbook

Closing as answered — reopen if anything's unclear!