| --- |
| license: mit |
| title: Hybrid RAG API |
| sdk: docker |
| colorTo: blue |
| pinned: false |
| --- |
| |
|
|
| # Hybrid RAG Retrieval API |
|
|
| **BM25 + E5 Embeddings | 93% Recall@10 | MRR=1.0 | MSc Data Science Thesis 2026** |
|
|
| ## Live Endpoints |
|
|
| | Endpoint | Description | |
| |----------|-------------| |
| | `GET /` | API info and metrics | |
| | `POST /search` | Hybrid search | |
| | `GET /health` | Health check | |
| | `GET /docs` | Swagger UI | |
|
|
| ## Example Request |
| ```json |
| POST /search |
| { |
| "query": "What is machine learning?", |
| "method": "hybrid", |
| "top_k": 5, |
| "alpha": 0.7 |
| } |
| ``` |
|
|
| ## Key Results |
|
|
| - **93.0%** Recall@10 on MS MARCO |
| - **+11.4%** improvement over baseline |
| - **MRR = 1.0** (Perfect) |
| - **Optimal α = 0.70** (statistically proven, p=0.002) |
|
|
| ## Links |
|
|
| - [Live Demo](https://rohith2026-hybrid-rag-demo.hf.space) |
| - [arXiv Paper](https://arxiv.org/abs/2507.18910) |
| - [GitHub](https://github.com/RohithkumarReddipogula/AI-Powered-Rag-System) |
|
|
| **Built by Rohith Kumar Reddipogula | University of Europe for Applied Sciences, Potsdam** |
