How UltraRAG Turns RAG Development into a Zero‑Code, One‑Click Process

Overview

UltraRAG is an open‑source Retrieval‑Augmented Generation (RAG) framework (Apache‑2.0) that integrates data construction, model fine‑tuning, inference, and multi‑dimensional evaluation into a single zero‑code WebUI. It targets four common pain points of traditional RAG pipelines: fragmented tooling, high technical threshold, insufficient evaluation, and limited multimodal support.

Core Technical Features

Zero‑code WebUI : All stages—knowledge‑base creation, data processing, model fine‑tuning, and evaluation—are performed through a visual interface, eliminating the need for custom scripts.

One‑click data construction & fine‑tuning : Built‑in methods such as KBAlign and RAG‑DDR automatically slice, label, and align domain documents. Supports standard fine‑tuning strategies (SFT, LoRA) and can be triggered with a single button.

Multidimensional evaluation (RAGEval) : Extends ROUGE‑L with metrics for effective information recall and key‑knowledge coverage. Example: vanilla RAG achieved 40.75 % ROUGE‑L on a legal benchmark, while UltraRAG‑DDR raised it to 53.14 % and increased key‑law recall by ~30 %.

Multimodal support (VisRAG) : Parses mixed‑content PDFs (tables, figures, formulas, medical images) and incorporates extracted visual tokens into the vector store, enabling text‑image joint retrieval.

Modular architecture : Separate layers for modules, workflows, and functions allow swapping of retrieval models, generators, or vector databases without rewriting pipelines.

Micro‑service deployment mode : Embedding models, LLMs, and vector stores can be run as independent services, facilitating high‑concurrency enterprise scenarios.

Real‑World Scenarios

1. Enterprise Legal Question‑Answering

Deploy UltraRAG with Docker‑Compose and open the WebUI at http://localhost:8843.

Import legal textbooks; the system automatically slices and labels data using KBAlign.

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

Run RAGEval: ROUGE‑L improves from 40.75 % to 53.14 % and key‑law recall rises by 30 %.

Lawyers query the system and receive AI‑generated answers up to five times faster than manual lookup.

2. Multimodal Medical Knowledge Base

Upload a medical PDF containing text and CT‑scan screenshots via the WebUI.

VisRAG extracts image regions (e.g., lesion locations, numeric indicators) and merges them with textual content into structured knowledge slices.

Enable “multimodal search” during retrieval; queries return both textual reports and relevant imaging cases.

The generated answer combines visual and textual evidence, improving diagnostic completeness.

3. Researcher Rapid Method Validation

Replace the default retrieval module with a custom algorithm using UltraRAG’s modular API.

Load a built‑in evaluation dataset or import a custom one; launch multidimensional evaluation with a single click.

Compare effective information recall and key‑knowledge coverage between the new method and baseline RAG.

Focus on algorithmic innovation without re‑implementing data pipelines, roughly halving experiment setup time.

Evaluation Metric – RAGEval

RAGEval reports three complementary scores:

ROUGE‑L : Traditional n‑gram overlap.

Effective Information Recall (EIR) : Fraction of ground‑truth facts correctly retrieved.

Key‑Knowledge Coverage (KKC) : Proportion of domain‑specific concepts present in the generated answer.

These metrics are computed automatically after each fine‑tuning run and displayed in the WebUI.

Getting Started (5‑Minute Deployment)

Option 1 – Docker (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 WebUI.

Option 2 – Conda Environment (custom development)

# Create a conda environment
conda create -n ultrarag python=3.10
conda activate ultrarag

# Install Python dependencies
pip install -r requirements.txt

# Download default models (saved under resources/models)
python scripts/download_model.py

# Launch the WebUI
streamlit run ultrarag/webui/webui.py --server.fileWatcherType none

Visit http://localhost:8843 to begin using UltraRAG.

Repository

Project source code: https://github.com/OpenBMB/UltraRAG

Key Requirements

CUDA ≥ 12.2

Python ≥ 3.10

First launch downloads pretrained models; a stable internet connection is recommended.