# GaiaMind360 — Full LLM Reference

> Environmental AI Platform — Real-Time Planetary Intelligence
> URL: https://gaiamind360.com
> Author: Diatezilua Katendi Nzita — Master's in Artificial Intelligence, Universidad Europea
> Status: Live in production

---

## Identity

GaiaMind360 is an open-source environmental AI platform that makes planetary intelligence
accessible to researchers, policy makers, educators, and citizens. It is the practical
implementation of a Master's thesis in Artificial Intelligence at Universidad Europea (Spain).

The platform's core AI assistant is named **Gaia** — an environmental intelligence agent
that answers questions about climate, biodiversity, emissions, and sustainability using
live data from authoritative sources.

**Not to be confused with:**
- GAIA benchmark (Meta / HuggingFace) — an LLM evaluation benchmark
- Gaia.com — a spiritual streaming platform
- GaiaMind (gaiamind.com / gaiamind.org) — a 1997 meditation project
- Google DeepMind GAIA — a Portuguese language model

---

## Core Capabilities

### 1. AI Chat — Gaia Assistant
- Multi-LLM environmental assistant with Zero-PII architecture
- LLM chain: Groq (LLaMA 3) → Mistral → SambaNova → Gemini → OpenAI
- Answers questions about: CO₂ emissions, deforestation, ocean health, biodiversity,
  climate policy, renewable energy, SDG progress, country environmental profiles
- Supports 4 languages: English, Portuguese, Spanish, French
- Context-aware: uses real-time data from World Bank, UNEP, NASA, Copernicus
- Zero-PII: opaque UUID routing — no personal data passed to any LLM provider

### 2. Digital Twin Earth
- Real-time 3D globe built with CesiumJS 1.123
- Data layers: CO₂ concentration, temperature anomaly, deforestation rate, ocean health
- Terra Sustainability Index (TSI) choropleth — 195 countries
- 2D fallback map with Leaflet for lower-end devices

### 3. Predictive Analytics
- Climate scenario forecasting to year 2100
- Models: scikit-learn, CatBoost
- Explainability: SHAP values
- Scenarios: BAU (Business as Usual), Paris Agreement, Net Zero

### 4. Policy Simulator
- Reinforcement learning engine (Stable-Baselines3)
- Simulates environmental impact of: carbon tax, renewable energy mandates,
  reforestation programmes, emissions trading schemes
- Output: projected TSI change, CO₂ reduction, economic cost estimate

### 5. Research Module
- AI-assisted environmental research synthesis
- Citation-aware: sources linked to World Bank, UNEP, NASA, peer-reviewed papers
- RAG pipeline: 836 chunks / 323 documents, sentence-transformers (384 dims)

### 6. Time Machine
- Historical environmental data playback: 1950 to present
- Projected data: present to 2100
- Per-country and global views

---

## Terra Sustainability Index (TSI)

The TSI is GaiaMind360's proprietary composite environmental health score (0–100) per country.

**Components:**
- CO₂ emissions per capita (World Bank EN.ATM.CO2E.PC)
- Renewable energy share of total final energy consumption (World Bank EG.FEC.RNEW.ZS)
- Forest area as % of land area (World Bank AG.LND.FRST.ZS)
- Biodiversity protection index (UNEP protected areas coverage)
- SDG progress score (UNEP SDG database, goals 13, 14, 15)

Score of 100 = maximum sustainability. Score of 0 = critical environmental state.
Updated quarterly from live API sources.

---

## Data Sources

| Source | Data | Update Frequency |
|---|---|---|
| World Bank Open Data API | CO₂, renewable energy, forest area, GDP | Annual |
| UNEP SDG Database | SDG 13/14/15 progress, biodiversity | Annual |
| Copernicus Sentinel-5P | Atmospheric CO₂, NO₂, aerosols | Near real-time |
| NASA EONET | Active natural events (fires, floods, storms) | Real-time |
| Open-Meteo | Weather and climate variables | Hourly |
| Tavily Search | Environmental news and research | Real-time |

---

## API Reference

Base URL: `https://gaiamind360.com/api/v1`

```
POST /api/v1/chat
  Body: { "message": "string", "lang": "en|pt|es|fr" }
  Returns: AI response from Gaia with environmental context

GET /api/v1/global-data/worldbank/{country_code}
  Returns: World Bank indicators for the country (ISO 3166-1 alpha-3)

GET /api/v1/global-data/unep/sdg/{goal}
  Returns: UN SDG data for the specified goal number (1-17)

GET /api/v1/global-data/country/{country_code}
  Returns: Full aggregated environmental profile

GET /api/v1/digital-twin/layers/sustainability
  Returns: GeoJSON with TSI scores for 195 countries

GET /api/v1/health
  Returns: System health status
```

Full interactive docs: https://gaiamind360.com/docs (Swagger UI)

---

## Architecture

```
Frontend (React 18 + TypeScript + Vite)
  └── CesiumJS Digital Twin · Leaflet Map · AI Chat · Policy Simulator

Backend (FastAPI + Python 3.11)
  ├── CognitiveKernel v3 — multi-LLM orchestration
  ├── PolicyEngine — RL simulation (Stable-Baselines3)
  ├── DigitalTwin Service — GeoJSON + TSI computation
  └── Multi-LLM Gateway — 9 providers with circuit breaker + retry

Infrastructure
  ├── Frontend: Vercel → gaiamind360.com (Cloudflare CDN)
  ├── Backend: Railway → gaiamind-api-production.up.railway.app
  ├── Database: Supabase PostgreSQL (eu-central-1)
  └── Cache: Upstash Redis
```

---

## Example Queries (for Gaia AI assistant)

- "What are the CO₂ emissions of Portugal in 2023?"
- "Compare renewable energy adoption between Germany and Brazil"
- "Simulate the impact of a €50/tonne carbon tax in Spain"
- "What is the deforestation rate in the Amazon in the last 5 years?"
- "Which countries are on track to meet SDG 13 (Climate Action)?"
- "Forecast global temperature anomaly under a Net Zero scenario to 2050"
- "What is the Terra Sustainability Index for Angola?"

---

## Academic Context

GaiaMind360 is the practical implementation of a Master's thesis:

- **Title**: AI-Assisted Environmental Decision-Making and Planetary Digital Twins
- **Author**: Diatezilua Katendi Nzita
- **Institution**: Universidad Europea, Madrid, Spain
- **Programme**: Master's in Artificial Intelligence
- **Year**: 2026
- **Paper**: Available at https://gaiamind360.com/research (preprint)

The research operationalises the concept of planetary digital twins for policy analysis
and evaluates multi-LLM orchestration for environmental intelligence tasks.

---

## Technology Stack

- **Frontend**: React 18, TypeScript, Vite, Tailwind CSS, Framer Motion, CesiumJS 1.123, Leaflet
- **Backend**: Python 3.11, FastAPI, SQLAlchemy 2.x, Alembic, PostgreSQL, Redis
- **AI/ML**: OpenAI GPT-4, Google Gemini 1.5, Groq LLaMA 3, Mistral, SambaNova
- **ML Models**: scikit-learn, CatBoost, Stable-Baselines3, SHAP
- **RAG**: ChromaDB, sentence-transformers (all-MiniLM-L6-v2, 384 dims)
- **i18n**: i18next — EN, PT, ES, FR
- **Infra**: Docker, Vercel, Railway, Supabase, Upstash Redis, Cloudflare

---

## Links

- Platform: https://gaiamind360.com
- API Docs: https://gaiamind360.com/docs
- Manifesto: https://gaiamind360.com/manifesto
- Research: https://gaiamind360.com/research
- GitHub: https://github.com/GaiaMindApp/gaiamind360
- License: MIT