🌿 Sowa-Rigpa RAG

An Indigenous Knowledge–Centered AI Stack for Sustainable Health Futures

---

📖 Overview

Sowa-Rigpa (Gso-ba Rig-pa), the ancient Tibetan/Himalayan medical system, balances the three nyes-pa (rLung – wind, mKhris-pa – bile, and Bad-kan – phlegm).
This repository implements a Retrieval-Augmented Generation (RAG) framework to digitally preserve, retrieve, and extend indigenous knowledge in a transparent, community-first, and offline-friendly way.

Core stack:

• Docling → Document parsing + hybrid chunking (fallback: hierarchical)
• Ollama → Local, laptop-friendly LLM inference
• LangChain → Embedding integration with Ollama
• Milvus Lite / Milvus Server → Vector database for knowledge storage
• Gradio → Interactive web-based UI

---

⚡ Features

• Upload PDF / DOCX / MD / TXT
• Automatic chunking with Docling (hybrid or hierarchical)
• Vector embeddings with Ollama embeddings
• Store/retrieve in Milvus (Lite or remote server)
• Top-K retrieval with transparent citations
• Runs fully offline, designed for eco-conscious laptop/edge use
• Logging:
  • JSONL query history
  • CSV metrics (ollama_metrics.csv)

---

🏗️ System Workflow

flowchart TD
    A[User Uploads Document / Question] --> B[Docling Parser + Chunking]
    B --> C[Ollama Embeddings]
    C --> D[Milvus Vector DB]
    D -->|Top-K Retrieval| E[Context Builder]
    E --> F[Ollama LLM - Generate Answer]
    F --> G[Answer + Metrics + Sources via Gradio UI]

Loading

---

📦 Installation

1. Clone repository

git clone https://github.com/your-username/sowa-rigpa-rag.git
cd sowa-rigpa-rag

2. Create virtual environment

python3 -m venv venv
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

---

🚀 Usage

Run locally

python sowarigpa.py

Access the Gradio UI at: 👉 http://localhost:7860

Upload Documents

• PDF, DOCX, MD, TXT supported
• TXT bypasses Docling (direct splitter)

Ask Questions

• Enter question → retrieves Top-K passages from Milvus
• Builds context prompt → generates answer with Ollama

---

📊 Logging

• logs/history.jsonl → Q&A with contexts and metrics
• logs/ollama_metrics.csv → performance metrics (latency, tokens/sec, etc.)

---

🌍 Applications

• Assist Amchi practitioners with diagnostics & formulations
• Herb–drug interaction alerts
• Climate-aware substitution for endangered herbs
• Educational & research support for Indian Knowledge Systems (IKS)

---

🔮 Future Directions

• Expand multilingual support (Hindi, English and more)
• IoT integration for climate/ecology-aware advice
• Patient-centered health advisory tools
• Cross-domain adaptation for Ayurveda, Siddha, Unani

---

🧑‍💻 Author

Partha Pratim Ray Assistant Professor, Department of Computer Applications Sikkim University (Central University), Gangtok, India

---

📜 License

This project is licensed under the MIT License.

---

⚡ I kept it professional + research-oriented while making it developer-friendly with badges, mermaid diagram, and clear steps.  

Do you want me to also prepare a `requirements.txt` file (with the correct package versions: `docling`, `langchain`, `pymilvus`, `gradio`, etc.) so your repo is immediately runnable?