Autonomous RAG Pipeline Builder & Optimizer

AutoRag is an intelligent framework that automates the creation, evaluation, and optimization of Retrieval-Augmented Generation (RAG) systems.

Instead of manually tuning embeddings, chunking, retrievers, and prompts — AutoRag experiments, evaluates, and finds the best pipeline for your data automatically.

🧠 Why AutoRag?

Building a good RAG system is hard:

Which embedding model works best?

What chunk size should you use?

Which retriever + reranker combination is optimal?

How do you evaluate performance objectively?

AutoRag solves this by:

⚙️ Automatically testing multiple pipeline configurations

📊 Evaluating performance using custom datasets

🧩 Selecting the best-performing architecture

🚀 Deploying optimized pipelines instantly

✨ Features

🔄 Automated RAG Pipeline Search

Tries multiple combinations of:

Embeddings

Chunking strategies

Retrievers

Prompt templates

📊 Evaluation-Driven Optimization

Uses QA datasets to score pipelines

Selects best pipeline based on metrics

🧱 Modular Architecture

Easily plug in:

Custom LLMs

Local models

APIs (OpenAI, etc.)

⚡ End-to-End Workflow

Data → Chunking → Retrieval → Generation → Evaluation → Deployment

🖥️ Interactive Execution

Raw Documents ↓ Parsing & Cleaning ↓ Chunking Strategies ↓ Embedding + Indexing ↓ Retrieval + Reranking ↓ LLM Generation ↓ Evaluation Engine ↓ Best Pipeline Selection

git clone https://github.com/VivekArgSharma/AutoRag.git cd AutoRag pip install -r requirements.txt python main.py

Run pipelines and visualize outputs

Debug and improve results in real-time