PROJECT OVERVIEW :
Introduction:
->What is MOSDAC?
MOSDAC Portal ([Link]) is a website that provides satellite data,
services, and scientific information to users such as researchers, students, or
professionals.
->Why is this bot is created (Purpose)?
To answer user queries instantly and accurately by understanding their questions
and retrieving the right information from the portal.
Key Objectives :
1. A virtual assistant for query understanding and precise information
retrieval.
2. Extract structured/unstructured content into a dynamic knowledge graph.
3. Support geospatial intelligence to answer location-specific queries.
4. Ensure modularity for easy reuse across similar portals.
System Architecture :
A. Frontend (User Interface)
Built using [Link] / HTML / CSS / JavaScript or Streamlit framework.
Provides chat interface, menus, and user-friendly dashboards.
Sends user queries to backend via API calls.
B. Backend (Server & APIs)
->Developed using [Link] (or Python Flask if used)
->Acts as a middleware between frontend, NLP models, and external APIs
->Handles:
->Routing requests
->Calling AI/NLP models
->Fetching data from MOSDAC APIs
->Sending processed response back to frontend
C. NLP / AI Layer
->Uses spaCy / NLTK for Natural Language Processing tasks
->Dialogflow / Rasa / LangChain for conversational flow and prompt chaining
->NVIDIA RAG (Retrieval-Augmented Generation) for enhanced, context-based
answers (if integrated)
->Processes user input → understands intent → generates appropriate response
D. External APIs
MOSDAC Data APIs:
Accesses satellite mission data, weather data, and related services to provide
real-time information to users.
E. Database (if used)
Stores FAQs, user data, chat logs, or predefined query mappings.
(Specify MongoDB, MySQL, or None as per your implementation)
Complete Architecture :
Workflow Summary
[Link] enters a query in the chat interface.
[Link] sends the query to the backend via API call.
[Link] routes the query to NLP models for understanding.
[Link] required, backend fetches data from MOSDAC APIs.
[Link] processes the final response and sends it to frontend.
[Link] sees the response in the bot chat
How will the MOSDAC Help-Bot solve the problem?
The MOSDAC Help Bot is designed to solve the challenges faced by users while
accessing satellite data and information in the following ways:
[Link] Query Resolution
Users often struggle to find specific satellite data or mission details on the
MOSDAC portal. The bot provides quick, direct answers without navigating
multiple pages.
2. Natural Language Interaction
Instead of using technical keywords or filters, users can ask questions in simple
natural language, which the bot understands and processes efficiently.
3. 24/7 Availability
The bot is always active, ensuring round-the-clock assistance without human
dependency.
[Link] with MOSDAC APIs
It fetches real-time data directly from MOSDAC APIs, ensuring accurate and
updated information delivery.
[Link] Manual Load
Reduces the workload of support teams by automating repetitive queries,
allowing them to focus on complex tasks.
6. Improved User Experience
Enhances the overall user experience by providing an interactive, conversational,
and intuitive platform for learning and data retrieval.
7. Scalability for Future Enhancements
Its modular design allows adding voice support, multilingual responses, or data
visualiations in the future, making it a scalable solution for diverse user needs.
In summary, the MOSDAC Help Bot ensures that accessing satellite data becomes
easy, fast, and efficient, thus bridging the gap between complex scientific data
and user-friendly information delivery.
Uniqueness of the MOSDAC Help Bot Project
1. Domain-Specific AI Integration
Unlike generic chatbots, this project is tailored specifically for satellite data,
weather information, and MOSDAC mission datasets, addressing a niche and
technical user domain.
2. Retrieval-Augmented Generation (RAG) Implementation
Uses advanced AI techniques like NVIDIA RAG (if integrated) for providing
contextual, knowledge-based answers, which most standard bots do not offer.
3. No Dependency on Static FAQs
Rather than only using predefined FAQs, it dynamically fetches updated data
from MOSDAC APIs, ensuring fresh and real-time information to the users.
4. Scalable & Modular Architecture
Its modular system design allows easy integration with other ISRO or data
portals, making it a future-proof and extendable solution.
Folder Structure :
mosdac-helpbot/
│
├── data/
│ ├── raw/ # Downloaded PDFs, DOCXs, XLSXs, web pages
│ └── processed/ # Cleaned and extracted data for Knowledge
Graph
│
├── knowledge_graph/
│ ├── entity_extraction/ # Scripts for NLP entity and relation extraction
│ ├── graph_construction/ # Scripts to build and update Knowledge
Graph
│ └── models/ # Trained models for entity/relation extraction
│
├── models/
│ ├── intent_recognition/ # AI models for understanding user queries
│ └── rag/ # Retrieval-Augmented Generation components
(NVIDIA RAG)
│
├── chatbot/
│ ├── rasa_dialogflow/ # Rasa/Dialogflow configurations and training
data
│ ├── langchain/ # LangChain scripts and workflows
│ └── scripts/ # Chatbot integration scripts
│
├── frontend/
│ ├── react_app/ # React-based chatbot interface
│ └── streamlit_app/ # Optional Streamlit interface for testing/demo
│
├── backend/
│ ├── api/ # [Link] APIs to connect models and frontend
│ └── utils/ # Helper scripts for data access, cleaning, etc.
│
├── tests/ # Unit and integration tests
│
├── deployment/
│ ├── docker/ # (If used later) Dockerfiles and configs
│ └── scripts/ # Deployment scripts for different environments
│
├── docs/ # Design documents, architecture diagrams, reports
│
├── [Link] # Python dependencies
├── [Link] # [Link] dependencies
├── [Link] # Project overview and setup instructions
└── .gitignore # Git ignore file