A data intelligence system that uses Autonomous AI Agents to monitor, classify, and cross-reference geopolitical, military, and economic events worldwide in real time. The system extracts data from unstructured sources, structures it into JSON via LLMs, and plots events on an interactive dashboard with geolocation filters.
In the future, the project will include an experimental Natural Language Processing (NLP) module for comparative analysis between contemporary events and symbolic/historical texts.
The project is designed using microservices principles and will be fully containerized via Docker:
- Ingestion Engine: Python scripts using
Pydanticfor strict validation of LLM-extracted data. - Database: PostgreSQL with
PostGISextension for spatial and geolocation queries. - Automation Agents:
CrewAIimplementation for asynchronous news and treaty search and analysis. - API & Backend:
FastAPIto serve consolidated data. - Frontend: Interactive dashboard rendering maps using
Leaflet.js.
To run the project on your machine, use the automated script that sets up the environment, starts the API and Dashboard, and cleans up all state and metadata on exit (avoiding issues from past runs).
Open a terminal and navigate to the project folder:
cd /path/to/geosentryGrant execution permission to the script and start the environment:
chmod +x start.sh
./start.shThis will set up the geosentry.db database, start the API (Backend), and open the Frontend.
The Dashboard will open automatically (or visit http://localhost:5173).
When you stop the script by pressing Ctrl+C, all temporary files, database files, and logs created during the local session will be deleted, ensuring the project starts fresh on the next run.
├── /api # FastAPI Backend
├── /dashboard # Frontend (HTML, JS, Leaflet)
├── /ingestion # Extraction scripts and Pydantic models
├── /agents # CrewAI agent configuration
└── docker-compose.yml
GeoSentry is a project conceived, designed, and developed by Pablo Dias.
© 2026 Pablo Dias. All rights reserved.
This software is distributed under the MIT License. See the LICENSE file for details.