🟣 PÚRPURA Climate OS

Climate risk assessment and IFRS S2 compliance platform for Brazilian organizations, utilities and municipalities

Powered by OS-Climate open-source tools + custom AI extraction pipeline.

---

✅ MVP Status: Ready for Testing!

Week 5-6 Complete — Physical Risk Dashboard is live!

🎯 What's Working:

• ✅ Backend API: Physical risk assessment for 10 Brazilian municipalities
• ✅ Frontend Dashboard: React + TypeScript with advanced visualizations
• ✅ Data Integration: INPE + Cemaden + INMET + Geographic Heuristics
• ✅ 5 Hazard Types: Flood, Drought, Heat Stress, Landslide, Coastal Inundation
• ✅ 3 Climate Scenarios: RCP 2.6, 4.5, 8.5
• ✅ Temporal Projections: Current → 2030 → 2050

📊 Dashboard Features:

• StatsOverview (6 key risk metrics)
• RiskChart (temporal evolution line chart)
• ScenarioComparison (RCP scenario bar chart)
• RiskCard grid (individual hazard cards)

🚀 Quick Demo: See Quick Start below to run locally in 5 minutes!

---

🎯 Mission

Enable Brazilian organizations to:

• Assess physical climate risks (floods, droughts, heat stress)
• Extract climate data from sustainability reports (AI-powered)
• Comply with Lei 14.904/2024, IFRS S2, and TSB taxonomy
• Report actionable insights for climate adaptation

---

🏗️ Architecture

┌─────────────────────────────────────────────────────────────┐
│                  Frontend (React + TypeScript)               │
│     📊 Municipal Dashboard ✅ │ 📄 Reports (Planned)         │
│   • Risk visualizations • Scenario comparison • Charts       │
└──────────────────────────┬──────────────────────────────────┘
                           │ REST API (FastAPI)
┌──────────────────────────┴──────────────────────────────────┐
│                  Backend (FastAPI + Python)                  │
│  ┌─────────────┐  ┌──────────────┐  ┌──────────────┐       │
│  │ Extraction  │  │ Risk Engine  │  │  Compliance  │       │
│  │ (Hybrid) ✅ │  │ (physrisk) ✅│  │  (Planned)   │       │
│  └─────────────┘  └──────────────┘  └──────────────┘       │
│  INPE • Cemaden • INMET • IBGE                              │
└──────────────────────────┬──────────────────────────────────┘
                           │
┌──────────────────────────┴──────────────────────────────────┐
│             Data Layer (Trino + Iceberg + MinIO)             │
│  📦 Documents ✅ │ 📊 Extractions ✅ │ 🗺️ Geospatial ✅     │
└─────────────────────────────────────────────────────────────┘

Tech Stack

Backend:

• FastAPI (async API framework)
• OS-Climate tools:
  • osc-transformer-based-extractor (BERT-based KPI extraction)
  • physrisk-lib (physical climate risk modeling)
• OpenAI GPT (hybrid extraction)
• ChromaDB (RAG vector store)
• H3 geospatial indexing

Data Platform:

• Trino (distributed SQL)
• Apache Iceberg (table format)
• MinIO (S3-compatible storage)
• PostgreSQL (metadata)

Frontend:

• React 18 + TypeScript
• Vite (build tool with HMR)
• Axios (HTTP client)
• Recharts (data visualizations)
• Tailwind CSS (utility-first styling)
• Lucide React (icons)

---

🚀 Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+
• Docker + Docker Compose
• 8GB RAM minimum

1. Clone & Setup

git clone https://github.com/brunoccteixeira/PURPURA.git
cd PURPURA

# Create Python environment
python -m venv .venv
source .venv/bin/activate  # or .venv\Scripts\activate on Windows

# Install dependencies
pip install -r requirements.txt

2. Configure Environment

cp .env.example .env
# Edit .env with your settings:
# - OPENAI_API_KEY (for LLM extraction)
# - Database credentials
# - Trino connection

3. Start Infrastructure

# Start MinIO, Hive, Trino
docker-compose up -d

# Wait for services to be healthy (~30 seconds)
docker-compose ps

# Create Iceberg tables
python scripts/create_iceberg_tables.py

4. Start API Server

cd backend
uvicorn api.main:app --reload --port 8000

# API docs available at http://localhost:8000/docs

5. Start Frontend

cd frontend
npm install
npm run dev

# Dashboard at http://localhost:3000
# Features:
# - 10 Brazilian municipalities (São Paulo, Rio, Fortaleza, etc.)
# - 5 hazard types (flood, drought, heat stress, landslide, coastal inundation)
# - 3 RCP scenarios (2.6, 4.5, 8.5)
# - Temporal projections (Current → 2030 → 2050)
# - Advanced visualizations (charts, stats, scenario comparison)

---

📖 Documentation

• CLAUDE.md — Developer guide for AI assistants
• docs/MVP_PLAN.md — 12-week implementation roadmap
• docs/API.md — API reference
• docs/OS_CLIMATE_INTEGRATION.md — OS-Climate tools usage

---

🧪 Testing

# Run all tests
pytest

# With coverage
pytest --cov=backend --cov-report=html

# Specific test suite
pytest backend/tests/test_extraction.py

---

📊 Data Pipeline

Ingest PDF → Extract → Store

# 1. Ingest PDF to chunks
python scripts/ingest_pdf.py data/sample.pdf \
  --out data/chunks.jsonl \
  --doc-id sample_001

# 2. Extract KPIs (hybrid: transformer + LLM)
python scripts/extract_from_chunks.py data/chunks.jsonl \
  --out data/extracted.jsonl \
  --method hybrid

# 3. Publish to Trino
python scripts/publish_jsonl_to_trino.py data/chunks.jsonl
python scripts/publish_jsonl_to_trino.py data/extracted.jsonl

Query with Trino

-- View extracted KPIs
SELECT * FROM lake.ifrs.extract_results LIMIT 10;

-- Search by municipality
SELECT * FROM lake.ifrs.municipal_risks
WHERE ibge_code = '3550308';  -- São Paulo

---

🌍 OS-Climate Integration

PÚRPURA leverages these open-source components:

Component	Purpose	Status
osc-transformer-presteps	PDF → JSON conversion	✅ Integrated
osc-transformer-based-extractor	BERT KPI extraction	✅ Integrated
physrisk-lib	Climate risk calculations	✅ Integrated
physrisk-ui	Dashboard patterns	✅ Integrated
H3 geospatial indexing	Municipal risk mapping	✅ Integrated

Brazilian Data Sources

Source	Purpose	Status
INPE (PCBr API)	Climate projections (temperature, precipitation)	✅ Integrated
Cemaden	Historical hazard frequency (floods, landslides)	✅ Mock data ready
INMET (BDMEP)	Climate normals (1961-2023), station catalog	✅ Mock data ready
IBGE	Municipality codes, population, coordinates	✅ Integrated

License Compliance: All OS-Climate code is Apache 2.0 licensed. PÚRPURA maintains attribution and contributes improvements upstream.

---

🎨 Branding

• Name: PÚRPURA (purple in Portuguese, symbolizing innovation + sustainability)
• Target: Brazilian mid-market (municipalities 50-500k, regional utilities, SMEs)
• Language: 100% Portuguese (pt-BR)
• Differentiator: Partnership-first approach (Colab, Cemaden, TNC) + tropicalized models

---

🛣️ Roadmap

Phase 1: Municipal MVP (Weeks 1-12) — Current

• Data lakehouse infrastructure
• LLM extraction pipeline (OpenAI)
• FastAPI backend scaffold
• Transformer-based extraction integration
• Physical risk engine (physrisk + Brazilian data sources)
• Municipal dashboard UI (React + TypeScript)
• 2 pilot deployments

Phase 2: Enterprise TSB (Months 4-6)

• TSB taxonomy classifier
• IFRS S2 reporting module
• Enterprise dashboard
• 10-15 customer deployments

Phase 3: Scale (Months 7-12)

• Agro module (AQUASENSE)
• Smart city monitoring (MONITOR)
• PIM-PAM tools (RMPT, CCS, PIA)
• 100+ municipalities, 20+ enterprises

---

📈 Success Metrics (Week 12 Target)

• ✅ 2 pilot customers deployed
• ✅ 10+ documents processed end-to-end
• ✅ <5min extraction time per document
• ✅ >85% pilot user satisfaction
• ✅ 1 Colab partnership MoU signed

---

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md.

Key areas:

• Brazilian climate data sources integration (Cemaden, INPE, ANA)
• IFRS S2 / TSB schema validation
• Portuguese NLP improvements
• Municipal dashboard UX

---

📄 License

PÚRPURA Climate OS is licensed under the Apache License 2.0.

This project incorporates components from OS-Climate, also Apache 2.0 licensed.

---

🙏 Acknowledgments

• OS-Climate — Open-source climate tools foundation
• Cemaden — Brazilian climate monitoring data
• Linux Foundation / FINOS — Governance and community support

---

📞 Contact

• Website: purpura.climate (coming soon)
• GitHub: github.com/brunoccteixeira/PURPURA
• Email: bruno@purpura.climate

---

Made with 🟣 in Brazil