The Float chat

An innovative solution, meticulously crafted to transform challenges into opportunities and simplify everyday experiences.

Explore Our Prototype

Visualizing Our Vision

Main data dashboard view

Dashboard

A high-level overview displaying key operational metrics and real-time BGC data summaries for quick analysis. This central hub provides actionable insights into the status of all tracked assets.

AI chat interface for natural language query

Float Chat AI

The core interaction point where users input natural language questions to generate data queries or factual answers. It leverages RAG technology to ensure accurate, context-grounded responses from the database.

About Us information screen

About Us

Details the mission and team behind the project, ensuring transparency and building user trust in the system. Discover our commitment to innovative data solutions and shaping a streamlined future.

Geospatial data map visualization

Map View

Visual representation of geospatial data, allowing users to track and analyze BGC float deployments across oceans. This interactive feature is essential for monitoring asset locations and historical routes.

User and application settings configuration

Settings

Allows users to manage account preferences, configure data display options, and adjust notification settings. Control the look and feel of the interface and tailor data views to your specific needs.

The Idea in Motion

Watch our detailed video demonstration to fully grasp the potential and impact of this groundbreaking idea.

Our Technology Stack 🚀

💻

Frontend UI

Modern user interface and data visualization built with HTML, JavaScript, and Tailwind CSS.

⚡

Backend API

High-performance orchestration of the RAG pipeline and request management using FastAPI.

🧠

LLM Router

Determines user intent (SQL Query vs. Text Answer) and guides generation using Llama3.

☁️

LLM Server

Locally hosts and serves the Llama3 model and embedding models via Ollama.

🐘

Relational Data

Stores all high-volume, structured numerical and time-series data using PostgreSQL.

🗃️

Vector Data (RAG)

Manages contextual definitions and metadata for grounded answers in ChromaDB.

🔗

Drivers & Embeddings

Uses psycopg2 to connect to PostgreSQL and nomic-embed-text for vector creation.

🛠️

Utilities

Handles asynchronous requests (httpx), data validation (Pydantic), and SQL parsing (re).

Current Prototype Status

The FloatChat Data Viewer is a working proof-of-concept. This prototype successfully integrates a complex RAG and LLM pipeline to enable natural language interaction with structured data.

✅

Completion

The end-to-end pipeline is functional, demonstrating successful data ingestion, RAG-powered SQL generation, execution, and final data retrieval.

🛡️

Validation

We have successfully implemented strict prompt engineering to prevent model hallucination and ensure the AI accurately discriminates between factual questions (returning text) and data questions (returning SQL).

📈

Next Steps

The core framework is massively scalable and ready to be extended with more complex BGC (Biogeochemical) datasets and advanced interactive plotting features.

Back to Top