🎯 What is this?

FloatChart turns the ARGO ocean observation network — 4,000+ autonomous floats, 46 million+ profiles of temperature, salinity, and depth — into something you can just talk to. No SQL, no NetCDF wrangling, no oceanography degree. Ask a question in plain English, get back a chart, a map, or a clean export.

"Show temperature trends in the Bay of Bengal for 2024" → done in under a second.

🔓 Open	🤖 Agent-Ready	📴 Offline-First	🔒 Privacy-First	🛡️ Safety-First
All data & code publicly accessible	Versioned REST API built for LLMs	Runs fully on-device after setup	Zero telemetry, zero tracking	7-layer SQL sanitizer on every query

---

🏗️ How it fits together

---

✨ What you get

💬 AI Chat Interface Natural language → SQL → chart, automatically. 7 query types: statistics, proximity, profiles & more Instant AI-written summaries One-click CSV export	🗺️ Live Ocean Map Click anywhere on Earth's oceans. Nearby floats surface instantly Real-time trajectory playback Date-range filtering + hover details
📊 Analytics Dashboard Pre-built views for real analysis. Temperature time series Salinity distributions Vertical depth profiles	🤖 Developer API Everything above, callable from code. POST /api/v1/query GET /api/v1/tools — agent manifest Sub-second responses, no auth needed locally

---

⚡ Quick Start

git clone <your-repo-url>
cd FloatChart

python local_setup.py   # installs deps, configures DB, launches app

Opens automatically at http://localhost:5000 🎉

🔑 Prerequisites

Service	Cost	Get a key
NVIDIA NIM	Varies	build.nvidia.com
DeepSeek (optional)	Cheap	platform.deepseek.com

⚙️ Configuration (.env)

DATABASE_URL=postgresql://postgres:YOUR_PASSWORD@localhost:5432/floatchart
NVIDIA_API_KEY=nvapi-xxxxxxxxxxxxxxxxxxxxxxxx
NVIDIA_MODEL=meta/llama-3.3-70b-instruct

# Optional providers
# ANTHROPIC_API_KEY=sk-ant-...
# OPENAI_API_KEY=sk-...
# DEEPSEEK_API_KEY=...

---

🤖 Developer API

curl -X POST http://localhost:5000/api/v1/query \
  -H "Content-Type: application/json" \
  -d '{"query": "Show average temperature in Bay of Bengal for 2024"}'

{
  "success":      true,
  "answer":       "Average temperature in Bay of Bengal (2024): 28.4°C based on 3,201 measurements.",
  "chart_type":   "line",
  "query_type":   "Time-Series",
  "record_count": 365,
  "elapsed_ms":   218.4
}

Method	Endpoint	Description
POST	/api/v1/query	Natural language → structured ocean data
GET	/api/v1/tools	Machine-readable tool manifest for agents
POST	/api/v1/validate-sql	Run the safety sanitizer manually
GET	/api/health	Health check + DB status
GET	/api/stats	Database statistics

From Python:

import requests

def ask_ocean(question: str) -> dict:
    r = requests.post("http://localhost:5000/api/v1/query",
                       json={"query": question, "max_rows": 500})
    return r.json()

print(ask_ocean("Find the 5 nearest floats to Chennai")["answer"])

Built for agents: every tool is auto-discoverable via curl http://localhost:5000/api/v1/tools, so LLM agents (including anything speaking MCP) can call query_ocean_data, get_temperature_trend, get_depth_profile, and more without hand-written glue code.

---

🛡️ Safety Layer

Every AI-generated SQL query is checked before it ever reaches the database:

#	Layer	Blocks
1	Allowlist	Anything that isn't SELECT / WITH
2	Keyword scan	DROP, DELETE, INSERT, UPDATE, TRUNCATE, ALTER, GRANT
3	No stacking	;-chained injection attempts
4	No comment tricks	-- and /* */ bypasses
5	LIMIT cap	Runaway million-row queries
6	No system abuse	pg_read_file, pg_sleep, COPY
7	Clear rejections	Every block returns a human-readable reason

from sql_sanitizer import SQLSanitizer

SQLSanitizer.validate("SELECT AVG(temperature) FROM argo_data")
# → {"safe": True}

SQLSanitizer.validate("DROP TABLE argo_data;")
# → {"safe": False, "reason": "Blocked keyword detected: DROP"}

---

📁 Project Layout

FloatChart/
├── ARGO_CHATBOT/          # Main app — Flask server, LLM brain, SQL builder + sanitizer
│   └── static/            # Chat UI · Map UI · Dashboard UI
├── DATA_GENERATOR/        # Local ERDDAP data ingestion tool
├── local_setup.py         # One-click setup wizard
├── requirements.txt
├── .env.example
└── CONTRIBUTING.md

---

🛠️ Tech Stack

Layer	Technology
Backend	Python 3.9+ · Flask 2.0+
AI / LLM	NVIDIA Llama 3.3 70B · DeepSeek · OpenAI · Claude · Gemini
Database	PostgreSQL 15+ · SQLAlchemy
Frontend	HTML5 · CSS3 · JavaScript
Visualization	Chart.js · Leaflet.js

---

📚 Resources

Resource	Link
🌊 ARGO Program	argo.ucsd.edu
📡 ERDDAP Server	erddap.ifremer.fr
🧠 NVIDIA NIM	build.nvidia.com
🤝 Contributing	CONTRIBUTING.md

---

📄 License

MIT — free to use, modify, and distribute. See LICENSE.

The open intelligence layer for ARGO ocean data 🌊