Demos
See UltraRAG in action.
Short walkthroughs on real corpora — ingestion, retrieval, sources, multilingual, vision search.
Ingestion
Create a workspace & ingest a corpus
Upload PDFs/DOCX/XLSX, watch the live pipeline (parse → chunk → embed → index → graph), and ask the first question.
Retrieval
Hybrid retrieval with sources & citations
Vector + full-text + graph fusion, inline [N] citations, and the source panel that opens the exact PDF page.
Multilingual
Arabic & French, same precision
RTL Arabic with broken-plural recall, alongside French — first-class multilingual retrieval.
Vision search
Catalogues — return the product image
Batched OCR on a 100 MB catalogue + vision search: ask for a product and get its spec text and image.
More demos coming soon.