ACC-RAG is a simple Streamlit-based RAG app for asking questions over uploaded documents. It is inspired by the ACC-RAG paper, which proposes adaptive context compression to keep only the most useful information for answering a query efficiently.
- Upload and index documents.
- Select the active document.
- Ask questions in chat format.
- View a logic trace for answer reasoning.
- Manage documents from the sidebar.
This project is based on “Enhancing RAG Efficiency with Adaptive Context Compression” by Shuyu Guo, Shuo Zhang, and Zhaochun Ren. The paper introduces Adaptive Context Compression for RAG (ACC-RAG), a framework that dynamically adjusts compression rates based on query complexity and reports more than 4 times faster inference than standard RAG while maintaining or improving accuracy.
Paper link: ACC-RAG
git clone https://github.com/pranjalparmar/Adaptive-Contextual-Compression-RAG.git
cd Adaptive-Contextual-Compression-RAG
pip install -r requirements.txt
streamlit run main.py