🇮🇳 Hindi RAG

Retrieval-Augmented Generation for Hindi Language

A production-style Hindi RAG system for document retrieval, contextual answering, and Hindi NLP workflows.

---

📌 Table of Contents

• Overview
• Features
• Architecture
• Project Structure
• Installation
• Environment Variables
• Usage
• API Usage
• Hindi RAG Pipeline
• Models Supported
• Evaluation
• Roadmap
• Contributing
• License
• Acknowledgements

---

🧠 Overview

Hindi RAG is a Retrieval-Augmented Generation system designed specifically for Hindi documents.
It allows users to upload Hindi text/documents, retrieve relevant chunks, and generate accurate answers in Hindi using LLMs.

यह project Hindi documents पर question-answering करने के लिए बनाया गया है।
इसमें retrieval, embedding, vector search, और LLM generation का उपयोग किया जाता है।

---

✨ Features

• ✅ Hindi document ingestion
• ✅ Text chunking optimized for Devanagari
• ✅ Vector database storage
• ✅ Semantic search
• ✅ LLM-based answer generation
• ✅ Hindi + English response support
• ✅ FastAPI-based REST API
• ✅ Optional Streamlit UI
• ✅ Modular and production-friendly structure
• ✅ Easy model swapping

---

🏗 Architecture

flowchart LR
    A[Hindi Documents] --> B[Document Loader]
    B --> C[Text Cleaning]
    C --> D[Chunking]
    D --> E[Embedding Model]
    E --> F[(Vector Database)]
    G[User Query] --> H[Query Embedding]
    H --> F
    F --> I[Top-K Retrieved Chunks]
    I --> J[Prompt Builder]
    J --> K[LLM]
    K --> L[Hindi Answer]

Loading