README.md

---
title: Rag Transparency Lab
emoji: 💻
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 6.11.0
app_file: app.py
pinned: false
license: mit
short_description: 'See inside every step of a RAG pipeline '
---

# 🔬 RAG Transparency Lab

**See inside every step of a RAG pipeline — applied to scientific papers.**

Upload any scientific PDF, ask a question, and watch the entire pipeline unfold — chunking strategies, hybrid retrieval scores, reranking decisions, filtering logic, and the exact prompt sent to the LLM.

Built by **Zalina Dezhina, PhD** — AI Evaluation Scientist  
*RAG Education Series · Project 1 of 3*

---

## What this tool shows

| Stage | What you see |
|---|---|
| 📄 Chunking | All chunks with word count, sentence count, strategy comparison |
| 🔍 Retrieval | Dense score, BM25 score, hybrid score per chunk |
| ⚖️ Reranking | Before/after rank, filter decisions with reasons |
| 💬 Generation | Answer with citations + full prompt transparency |

## Stack

`sentence-transformers` · `rank-bm25` · `pypdf` · `Anthropic API` · `Gradio`

## RAG Education Series

1. 🔬 **RAG Transparency Lab** ← this Space
2. ⚡ Classic vs Advanced RAG — side-by-side
3. 🧪 Scientific Claim Verifier

---

[LinkedIn](https://linkedin.com/in/zalina-dezhina) · [GitHub](https://github.com/Mioulin)