metadata

title: RAGnosis
emoji: 👁
colorFrom: red
colorTo: indigo
sdk: gradio
sdk_version: 5.35.0
app_file: app.py
pinned: false
license: mit
short_description: Clinical Query Answering with RAG + MIMIC-IV Notes.

🩺 RAGnosis – Clinical Reasoning via Retrieval-Augmented Generation

⚕️ A fully offline-capable, Gradio-powered RAG assistant trained on annotated clinical notes from the MIMIC-IV-Ext-DiReCT dataset to perform explainable diagnostic reasoning.

🖼️ Demo

Try it live on Hugging Face Spaces 👉
🔗 https://huggingface.co/spaces/asadsandhu/RAGnosis

Demo

⚙️ Tech Stack

Layer	Details
🧠 Model	`Nous-Hermes-2-Mistral-7B-DPO`
🏥 Dataset	`MIMIC-IV-Ext-DiReCT`
🔍 Retriever	FAISS + SentenceTransformers (`all-MiniLM-L6-v2`)
💻 Frontend	Gradio (Hugging Face Spaces)
🧠 Backend	PyTorch + Transformers + BitsAndBytes

🚀 Features

🔎 Top-k document retrieval from real annotated clinical notes
📋 Reasoning based on structured diagnostic chains
🧠 GPT-style generation from LLM (Mistral 7B) without internet dependency
🧾 Clean Gradio interface for natural medical queries
🧠 Answers explained like a clinical reasoning expert

⚡ Example Prompt

Patient presents with fatigue, orthopnea, and lower extremity edema.

💬 Model response:

Based on the patient's symptoms and context, the most likely diagnosis is congestive heart failure (CHF)...

🛠 How It Works

✅ Step 1: Preprocessing

Extract chains from samples/ and diagnostic_kg/
Build retrievable clinical observations + diagnoses

✅ Step 2: Retrieval (FAISS)

Embed notes using MiniLM-L6-v2
Save as FAISS index → faiss_index.bin
Paired with → retrieval_corpus.csv

✅ Step 3: Generation

Format prompt in [INST] syntax
Generate diagnosis using Nous-Hermes-2-Mistral-7B-DPO

🧪 Run Locally

# 1. Clone the repository
git clone https://github.com/asadsandhu/RAG-Diagnostic-Assistant.git
cd RAG-Diagnostic-Assistant

# 2. Install dependencies
pip install -r requirements.txt

# 3. Run the app
python app.py

✔️ Required files:

📁 Folder Structure

RAG-Diagnostic-Assistant/
├── app.py
├── faiss_index.bin
├── retrieval_corpus.csv
├── requirements.txt
├── assets/
│   └── demo.png
└── README.md

👤 Author

Built with ❤️ by Asad Ali 🔗 LinkedIn – @asadsandhu0 🔗 Hugging Face – RAGnosis

📄 License

This project is under the MIT License.

🙏 Acknowledgments

MIMIC-IV-Ext-DiReCT: Annotated diagnostic data
Hugging Face Transformers + Gradio
Facebook Research – FAISS
Nous Research – Instruction-tuned Mistral model

⚠️ Disclaimer: This project is for research/demo use only. Not intended for clinical decision-making.