🌍 GeoInsight AI

Intelligent Geospatial Data Platform

A production-style geospatial data pipeline that ingests satellite datasets, extracts spatial features, and enables natural-language exploration through an AI-powered query agent.

---

🏗️ System Architecture

              Satellite / Environmental Data
                          │
                          ▼
            ┌─────────────────────────┐
            │    Apache Airflow DAG    │
            │  ┌───────────────────┐  │
            │  │  Check Raw Data   │  │
            │  └────────┬──────────┘  │
            │           ▼             │
            │  ┌───────────────────┐  │
            │  │  Validate Schema  │  │
            │  └────────┬──────────┘  │
            │           ▼             │
            │  ┌────────┴──────────┐  │
            │  │     PARALLEL      │  │
            │  │ ┌──────┐┌──────┐  │  │
            │  │ │ NDVI ││Slope │  │  │
            │  │ └──┬───┘└──┬───┘  │  │
            │  └────┼───────┼──────┘  │
            │       ▼       ▼         │
            │  ┌───────────────────┐  │
            │  │ Feature Extract   │  │
            │  └────────┬──────────┘  │
            │           ▼             │
            │  ┌───────────────────┐  │
            │  │  Load to PostGIS  │  │
            │  └───────────────────┘  │
            └─────────────────────────┘
                          │
                          ▼
            ┌─────────────────────────┐
            │   PostGIS / SQLite DB    │
            │  ┌───────────────────┐  │
            │  │  spatial_features │  │
            │  │  regions          │  │
            │  │  ndvi_data        │  │
            │  │  terrain_data     │  │
            │  │  land_cover       │  │
            │  └───────────────────┘  │
            └────────────┬────────────┘
                         │
              ┌──────────┴──────────┐
              ▼                     ▼
    ┌──────────────────┐  ┌──────────────────┐
    │    dbt Models     │  │  AI Query Agent  │
    │  ┌────────────┐  │  │  ┌────────────┐  │
    │  │  Staging   │  │  │  │ LangChain  │  │
    │  │  Marts     │  │  │  │ Text→SQL   │  │
    │  └────────────┘  │  │  │ RAG        │  │
    └──────────────────┘  │  └────────────┘  │
                          └────────┬─────────┘
                                   │
                                   ▼
                      ┌────────────────────────┐
                      │   Streamlit Dashboard    │
                      │  ┌────────┐ ┌────────┐  │
                      │  │  Map   │ │  Chat  │  │
                      │  │Explorer│ │ Agent  │  │
                      │  └────────┘ └────────┘  │
                      │  ┌────────┐ ┌────────┐  │
                      │  │Analytics│ │ Status │  │
                      │  └────────┘ └────────┘  │
                      └────────────────────────┘

---

✨ Key Features

Feature	Description
🌿 NDVI Processing	Compute vegetation index from multi-band satellite data
⛰️ Terrain Analysis	Slope, aspect, elevation gradients, and roughness
☀️ Solar Suitability	Score regions for solar panel installation potential
🔄 Airflow Pipeline	Automated DAG with parallel processing and validation
🗄️ PostGIS Schema	Spatial indexing with 7 tables and 3 analytics views
📐 dbt Models	3 staging + 3 mart models with schema tests
🤖 AI Query Agent	Natural language → SQL with LangChain + Gemini/OpenAI
📚 RAG Pipeline	Domain knowledge retrieval for contextual responses
🗺️ Interactive Maps	Folium maps with choropleth layers and heatmaps
📊 Analytics Charts	Plotly visualizations with correlation analysis

---

🚀 Quick Start

1. Clone the Repository

git clone https://github.com/yourusername/geoinsight-ai.git
cd geoinsight-ai

2. Set Up Virtual Environment

python -m venv venv
# Windows
venv\Scripts\activate
# macOS/Linux
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment

cp .env.example .env
# Edit .env with your API key:
#   GEMINI_API_KEY=your_key_here

5. Generate Sample Data & Run Pipeline

# Generate synthetic geospatial data
python scripts/generate_sample_data.py

# Process NDVI and terrain
python -m geoinsight.processing.ndvi
python -m geoinsight.processing.terrain

# Extract spatial features
python -m geoinsight.processing.features

# Load into database
python -m pipeline.ingestion.ingest

6. Launch Dashboard

streamlit run dashboard/app.py

Open http://localhost:8501 and start exploring!

💡 Tip: The dashboard has a one-click "Generate & Load Sample Data" button if you skip step 5.

---

🤖 AI Agent – Example Queries

The agent converts natural language to geospatial SQL:

Query	What It Does
"Show areas with high vegetation density and low terrain slope"	Filters NDVI > 0.5 and slope < 15°
"Find regions with tree canopy coverage greater than 60%"	Queries canopy_coverage_pct > 60
"Which regions have the highest NDVI values?"	Ranks by ndvi_mean DESC
"Identify locations suitable for solar panel installation"	Filters slope < 15° and sunlight > 6 hrs
"Compare vegetation across different land types"	GROUP BY land_type aggregation

The agent works in two modes:

• LLM Mode (with API key): Uses Gemini/OpenAI for intelligent SQL generation
• Rule-Based Mode (no API key): Uses keyword matching for common query patterns

---

📁 Project Structure

geoinsight-ai/
├── geoinsight/                  # Core processing package
│   ├── processing/
│   │   ├── ndvi.py              # NDVI computation & classification
│   │   ├── terrain.py           # Slope, aspect, solar analysis
│   │   ├── features.py          # Zonal stats & feature extraction
│   │   └── utils.py             # Coordinate & I/O utilities
│   └── agent/
│       ├── query_agent.py       # LangChain text-to-SQL agent
│       ├── prompts.py           # System prompts & few-shot examples
│       ├── rag.py               # RAG knowledge retrieval
│       └── tools.py             # Custom geospatial tools
│
├── pipeline/                    # Data pipeline
│   ├── dags/
│   │   └── geospatial_ingestion_dag.py  # Airflow DAG
│   ├── ingestion/
│   │   └── ingest.py            # Data loading logic
│   └── validation/
│       └── validators.py        # Quality checks
│
├── database/
│   ├── schema.sql               # PostGIS DDL (7 tables, 3 views)
│   └── connection.py            # DB manager (PostGIS + SQLite)
│
├── dbt_models/
│   ├── dbt_project.yml
│   └── models/
│       ├── staging/             # stg_regions, stg_ndvi, stg_terrain
│       └── marts/               # vegetation, terrain, region summary
│
├── dashboard/
│   ├── app.py                   # Streamlit main app
│   ├── components/
│   │   ├── map_view.py          # Folium interactive maps
│   │   ├── chat_interface.py    # AI agent chat UI
│   │   └── analytics.py         # Plotly charts
│   └── static/
│       └── style.css            # Premium dark theme
│
├── scripts/
│   └── generate_sample_data.py  # Synthetic data generator
│
├── tests/                       # 36 unit tests
│   ├── test_processing.py
│   ├── test_pipeline.py
│   └── test_agent.py
│
├── data/
│   ├── raw/                     # Input data (generated)
│   └── processed/               # Processed outputs
│
├── docker-compose.yml           # PostGIS + Airflow
├── requirements.txt
├── pyproject.toml
├── Makefile
└── README.md

---

🛠️ Technology Stack

Layer	Technologies
Geospatial	GeoPandas, Rasterio, Shapely, NumPy, SciPy
Data Pipeline	Apache Airflow, Python, dbt
Database	PostgreSQL + PostGIS (production) / SQLite (dev)
AI Agent	LangChain, Google Gemini / OpenAI, RAG
Dashboard	Streamlit, Folium, Plotly
Testing	pytest

---

🐳 Docker Setup (Production)

For full PostGIS + Airflow deployment:

docker-compose up -d

Services:

• PostGIS: localhost:5432 (user: geoinsight, pass: geoinsight)
• Airflow: localhost:8080 (user: admin, pass: admin)

---

🧪 Running Tests

python -m pytest tests/ -v

---

📊 Data Pipeline Stages

1. Ingestion – Raw satellite/terrain data (GeoTIFF, GeoJSON)
2. Validation – Schema, CRS, and quality checks
3. NDVI Processing – Vegetation index from NIR/RED bands
4. Terrain Analysis – Slope, aspect, solar suitability
5. Feature Extraction – Zonal statistics per region
6. Database Loading – PostGIS with spatial indexes
7. dbt Transformation – Analytics-ready mart tables

---

🎯 Resume Entry

GeoInsight AI – Intelligent Geospatial Data Platform
Python, GeoPandas, Airflow, PostGIS, dbt, LangChain

• Built a production-style geospatial pipeline that ingests satellite datasets
  and extracts spatial features including NDVI vegetation index and terrain slope.

• Developed an Airflow-orchestrated ELT pipeline with PostGIS storage and
  modular dbt models for spatial analytics.

• Implemented an AI-powered data agent using LangChain to convert natural-language
  queries into geospatial SQL queries and visualize results on interactive maps.

---

📄 License

MIT License – See LICENSE file for details.

---

Built with ❤️ for Geospatial Intelligence