Skip to content

[submmision] Bantu jyothi-Temple Culture Explorer RAG chatbot

Jyothi Bantu requested to merge Bantu_jyothi13/temple-culture-explorer-rag-chatbot:main into main

Temple Culture Explorer

Temple Culture Explorer is a Retrieval-Augmented Generation (RAG) chatbot built with Streamlit. It helps users explore famous Indian temples, sacred sites, architecture, rituals, festivals, pilgrimage etiquette, and cultural heritage using a local knowledge base.

The app works even without an API key by retrieving the most relevant local content. If an OpenAI API key is available, it can generate a more polished answer using the retrieved context.

Features

  • Streamlit chatbot interface
  • Local RAG-style retrieval from temple knowledge files
  • Supports .md and .txt files inside the data/ folder
  • Separate text files for famous temples and sacred sites in India
  • Sidebar list of loaded knowledge files
  • Reload button to refresh the knowledge base after adding new files
  • Optional OpenAI-powered answer generation
  • Works in fallback mode without an LLM API key

Tech Stack

  • Python
  • Streamlit
  • scikit-learn
  • TF-IDF vectorization
  • Cosine similarity retrieval
  • Optional OpenAI API integration

Project Structure

global-cultural-explorer-rag-chatbot/
├── data/
│   ├── akshardham_delhi.txt
│   ├── brihadeeswara_temple.txt
│   ├── dwarkadhish_temple.txt
│   ├── jagannath_temple_puri.txt
│   ├── kashi_vishwanath_temple.txt
│   ├── kedarnath_temple.txt
│   ├── meenakshi_amman_temple.txt
│   ├── temple_culture_notes.md
│   └── ...
├── streamlit_app.py
├── requirements.txt
├── README.md
└── .gitignore

Included Knowledge Base

The app includes information about several famous temples and sacred sites, including:

  • Kashi Vishwanath Temple, Varanasi
  • Sri Venkateswara Temple, Tirumala
  • Jagannath Temple, Puri
  • Somnath Temple, Gujarat
  • Kedarnath Temple, Uttarakhand
  • Ramanathaswamy Temple, Rameswaram
  • Meenakshi Amman Temple, Madurai
  • Brihadeeswara Temple, Thanjavur
  • Konark Sun Temple, Odisha
  • Kamakhya Temple, Guwahati
  • Vaishno Devi Temple, Katra
  • Dwarkadhish Temple, Dwarka
  • Siddhivinayak Temple, Mumbai
  • Lingaraja Temple, Bhubaneswar
  • Virupaksha Temple, Hampi
  • Kandariya Mahadeva Temple, Khajuraho
  • Swaminarayan Akshardham, Delhi
  • Mahabodhi Temple, Bodh Gaya
  • Golden Temple, Amritsar

Installation

Clone the repository:

git clone https://code.swecha.org/Charvi_Gandla/global-cultural-explorer-rag-chatbot.git
cd global-cultural-explorer-rag-chatbot

Create and activate a virtual environment:

python -m venv .venv

On Windows:

.venv\Scripts\activate

On macOS/Linux:

source .venv/bin/activate

Install dependencies:

pip install -r requirements.txt

Run the App

Activate your virtual environment, then run Streamlit. Examples below assume you created the .venv environment as shown earlier.

On Windows PowerShell:

.venv\Scripts\activate
streamlit run streamlit_app.py

On macOS/Linux:

source .venv/bin/activate
streamlit run streamlit_app.py

Either streamlit run streamlit_app.py or streamlit run app.py will work; a small wrapper streamlit_app.py is included to call main() from app.py.

Then open the local URL shown in the terminal, usually:

http://localhost:8501

How It Works

  1. The app reads .md and .txt files from the data/ folder.
  2. Markdown files are split by ## Section Title headings.
  3. Text files use the first non-empty line as the temple title.
  4. The app creates TF-IDF vectors for all knowledge chunks.
  5. When a user asks a question, cosine similarity finds the most relevant chunks.
  6. The answer is generated from retrieved context or displayed directly when no API key is set.

Add More Temple Files

Add a new .txt file inside the data/ folder.

Use this format:

Temple Name, Location

Write temple history, architecture, rituals, festivals, cultural importance, visitor etiquette, and travel notes here.

After adding or editing files, click Reload knowledge base in the app sidebar or restart Streamlit.

Optional OpenAI Generation

The app works without OpenAI. To enable generated answers, set an API key before running the app.

On Windows PowerShell:

$env:OPENAI_API_KEY="your_api_key_here"
streamlit run streamlit_app.py

On macOS/Linux:

export OPENAI_API_KEY="your_api_key_here"
streamlit run streamlit_app.py

You can optionally set a model:

export OPENAI_MODEL="gpt-4.1-mini"

Example Questions

  • Tell me about Tirumala Venkateswara Temple.
  • Which temple is famous for Ratha Yatra and Mahaprasad?
  • Explain Dravidian temple architecture.
  • What should I know before visiting Kedarnath?
  • Which temples are important in Odisha?
  • Tell me about festivals at Meenakshi Amman Temple.

Notes

  • Do not commit API keys or secrets.
  • Keep custom temple data in the data/ folder.
  • The current retrieval system is lightweight and local. It can later be upgraded to embeddings with FAISS, ChromaDB, or another vector database.

Team members

  • charvi
  • akshaya
  • laya
  • jyothi
  • yavaneetha

Team

Developed for an internship project on cultural heritage exploration using RAG.

Merge request reports