What's Changed

feat：

• Add RAG API
• Integration of RAG into llama_index project
• Update Dockerfile
• Fine grained model singleton, reducing memory usage and accelerating initialization speed
• CLI and API add parsing range parameters, allowing customization of start and end pages
• Support image footnotes

bugfix：

• When removing the smaller overlapping block, retain the boundary information of that block
• Fill in the threshold of 0.6->0.3 for the span block
• The problem of losing low score lines in OCR DET stage
• Merge multiple spans of a single line in the OCR DET stage
• Optimization of English Adhesive Word Segmentation Logic
• Inaccurate layout box issue
• The problem of merging words after being broken by line breaks
• The final output result contains certain special characters

Full Changelog: https://github.com/opendatalab/MinerU/compare/magic_pdf-0.7.1-released...magic_pdf-0.8.0-released