All Tools
V
RAGFreeOpen Source
VISRAG
Parsing-free visual RAG with vision-language models
Apache-2.0
ABOUT
Text RAG parses PDFs into tokens and drops charts, layout, and figures that often hold the answer. VisRAG embeds pages as images with a VLM retriever and generator so visual evidence survives retrieval, and EVisRAG reasons over multiple retrieved page images.
INTEGRATION GUIDE
1. Retrieve PDF pages as images for chart-heavy or scanned document QA
2. Benchmark visual document retrieval with VisRAG-Ret on SlideVQA or ChartQA
3. Run EVisRAG multi-image reasoning when answers span several document pages
4. Replace OCR-then-chunk pipelines with a VLM embedding of the original page
TAGS
pythonragvisionvlmmultimodaldocumentsopenbmb