🌊 FloatChat – AI-Powered ARGO Data Explorer

A conversational AI application for exploring ARGO oceanographic float data using natural language. FloatChat combines a chatbot interface, semantic search, and interactive visualizations to help users query, analyze, and understand oceanographic datasets without writing SQL or complex data-processing code.

The application integrates Retrieval-Augmented Generation (RAG), vector search, and structured database queries to provide meaningful insights from ARGO float observations.

---

🌐 Overview

FloatChat enables users to:

• Explore ARGO float observations through natural language
• Visualize float locations on interactive maps
• Generate pressure and parameter time-series plots
• Retrieve relevant information using semantic search
• Interact with oceanographic datasets in a ChatGPT-style interface

---

✨ Features

• AI-powered conversational interface
• ChatGPT-style chat experience
• Floating chat input
• Interactive map visualization
• Pressure and ocean parameter plotting
• Retrieval-Augmented Generation (RAG)
• Semantic search using vector embeddings
• PostgreSQL-backed structured data retrieval
• Fast and intuitive data exploration

---

🏗️ System Architecture

flowchart TD

A[User] --> B[Streamlit Chat Interface]

B --> C[RAG Pipeline]

C --> D[Vector Search - ChromaDB]
C --> E[(PostgreSQL Database)]

D --> F[Relevant Float Metadata]
E --> G[ARGO Float Data]

F --> H[Local LLM]
G --> H

H --> I[Generated Response]

I --> J[Interactive Maps]
I --> K[Time-Series Charts]
I --> L[Chat Response]

L --> A
J --> A
K --> A

Loading

---

🛠️ Technology Stack

Frontend

• Streamlit
• HTML
• CSS

Backend

• Python 3.x

Database

• PostgreSQL
• ChromaDB

AI & Machine Learning

• Retrieval-Augmented Generation (RAG)
• Flan-T5 / Local Large Language Model
• Vector Embeddings

Data Visualization

• Plotly
• Leaflet

---

📊 Data Exploration

FloatChat enables users to analyze ARGO observations through conversational queries.

Users can:

• View ARGO float locations
• Explore pressure profiles
• Analyze temperature variations
• Compare salinity measurements
• Visualize parameter trends over time
• Search oceanographic metadata semantically

The combination of structured SQL retrieval and vector search provides accurate and context-aware responses.

---

💬 Conversational Interface

The application features a modern chatbot interface inspired by ChatGPT.

The assistant helps users:

• Search ARGO observations
• Generate visualizations
• Retrieve oceanographic information
• Understand float metadata
• Explore datasets using natural language

---

🧠 Retrieval-Augmented Generation (RAG)

FloatChat combines structured database queries with semantic search.

The workflow includes:

• User submits a natural language query
• ChromaDB retrieves semantically relevant metadata
• PostgreSQL provides structured ARGO observations
• Local LLM interprets retrieved context
• Final response is generated with supporting visualizations

This architecture enables accurate, context-aware exploration of oceanographic datasets.

---

🚀 Demo Queries

Try asking questions such as:

• Show the temperature and pressure profiles of ARGO floats near the equator.
• Compare salinity levels in the Arabian Sea and Bay of Bengal over the last six months.
• Find the nearest ARGO floats to latitude 10°N and longitude 75°E.
• Plot pressure versus depth for floats in the Indian Ocean.
• Show pressure trends of ARGO floats recorded during March 2023.

---

⚙️ Configuration

The application uses environment-based configuration for managing:

• PostgreSQL connection settings
• ChromaDB configuration
• LLM parameters
• Data paths
• Streamlit settings

This structure allows seamless deployment across development and production environments.

---

🚀 Running the Application

Activate the Virtual Environment

D:/my_project/venv/Scripts/Activate.ps1

Start the Application

cd D:\floatchat
streamlit run front_end/app.py

Open in Browser

http://localhost:8501/

---

📁 Project Workflow

1. User enters a natural language query.
2. Streamlit sends the query to the backend.
3. ChromaDB performs semantic retrieval.
4. PostgreSQL fetches relevant ARGO observations.
5. The Local LLM generates a contextual response.
6. Interactive maps and charts are created when applicable.
7. Results are displayed within the chat interface.

---

📌 Notes

• Demo uses a subset of approximately 500 ARGO floats for improved performance.
• Designed as a proof-of-concept for oceanographic data exploration.
• Supports conversational querying with inline visualizations.
• Focused on Indian Ocean ARGO datasets.

---

🔮 Future Work

• Full global ARGO dataset integration
• Biogeochemical (BGC) ARGO support
• Satellite data integration
• Improved Retrieval-Augmented Generation pipeline
• Advanced filtering and analytics
• Enhanced UI/UX
• Multi-modal visualizations
• Support for additional Large Language Models

---

🎯 Project Scope

The primary objective of FloatChat is to demonstrate how conversational AI can simplify scientific data exploration.

The project focuses on:

• Conversational AI
• Retrieval-Augmented Generation
• Semantic search
• Database integration
• Interactive visualization
• Oceanographic data analytics
• Natural language interfaces for scientific datasets

---

🎓 Academic Purpose

FloatChat was developed as an AI-powered data exploration project to demonstrate:

• Retrieval-Augmented Generation (RAG)
• Vector database integration
• Natural language querying
• PostgreSQL data retrieval
• Interactive scientific visualization
• Conversational AI for domain-specific datasets

The project showcases how modern AI technologies can transform complex scientific datasets into intuitive, conversational experiences.

---

📜 License

This project is intended for educational, research, and learning purposes. Contributions, experimentation, and further development are encouraged.