How UltraRAG Delivers One‑Click, No‑Code RAG Deployment and Boosts Retrieval Accuracy

UltraRAG is an open‑source Retrieval‑Augmented Generation (RAG) framework that unifies the entire RAG pipeline—knowledge‑base construction, data processing, model fine‑tuning, and performance evaluation—through a zero‑code Web UI.

Key Technical Features

Zero‑code Web UI : All stages can be operated via visual controls without writing code.

One‑click data construction & fine‑tuning : Built‑in KBAlign automatically slices and annotates domain documents; RAG‑DDR optimizes retrieval and supports fine‑tuning strategies such as SFT and LoRA.

Multi‑dimensional evaluation : The proprietary RAGEval metric reports ROUGE‑L together with effective information recall and key‑knowledge coverage. In a legal benchmark, ROUGE‑L improved from 40.75 % to 53.14 % and key‑clause recall rose by ~30 %.

Multimodal support : The VisRAG module parses mixed text‑image PDFs (tables, figures, formulas) and integrates visual content into the knowledge base.

Modular architecture : Retrieval, generation, and embedding components are isolated, allowing users to replace any module with custom implementations.

Open‑source deployment : Licensed under Apache‑2.0; can be launched with a single Docker command or via a Conda environment.

Representative Use Cases

1. Enterprise Legal Question‑Answering

Start the service with docker-compose up --build -d and open the Web UI at http://localhost:8843.

Import legal books; KBAlign automatically slices and annotates the documents.

Select the “RAG‑DDR” strategy to trigger retrieval optimization and one‑click fine‑tuning.

Run evaluation with RAGEval: ROUGE‑L rises from 40.75 % to 53.14 %; key‑clause recall improves by 30 %.

Legal professionals receive AI‑generated answers up to five times faster than manual lookup.

2. Multimodal Medical Document Knowledge Base

Upload a medical PDF (containing text, CT images, and reports) via the “Knowledge Base Management” module.

VisRAG extracts image content (e.g., lesion locations) and merges it with textual data into structured slices.

Enable “multimodal retrieval” to query both textual and visual evidence.

Generated answers combine text and image information, providing richer diagnostic suggestions.

3. Researcher Rapid Validation of New RAG Methods

Replace the default retrieval component with a custom algorithm using UltraRAG’s modular interface.

Launch built‑in multi‑dimensional evaluation on provided or custom datasets with a single click.

Compare metrics such as effective recall and knowledge‑point coverage between the new method and baseline.

Focus on algorithmic innovation without re‑implementing data pipelines, reducing validation time from a week to a few days.

Getting Started in 5 Minutes

Option 1: Docker Deployment (Recommended)

# Clone the repository (optional)
git clone https://github.com/OpenBMB/UltraRAG.git
cd UltraRAG
# Build and start containers
docker-compose up --build -d

After the containers are running, open http://localhost:8843 to access the Web UI.

Option 2: Conda Environment (Custom Development)

# Create conda environment
conda create -n ultrarag python=3.10
conda activate ultrarag
# Install dependencies
pip install -r requirements.txt
# Download default models
python scripts/download_model.py
# Launch Web UI
streamlit run ultrarag/webui/webui.py --server.fileWatcherType none

Visit http://localhost:8843 to begin.

Repository

https://github.com/OpenBMB/UltraRAG