Senital-FOSS
An open-source, multi-agent intelligence engine that transforms chaotic, high-volume social media feeds into actionable, real-time crisis maps for first responders.
The Problem: During large-scale disasters—such as floods, earthquakes, or industrial accidents—first response agencies and local NGOs are paralyzed by information overload. Genuine, life-or-death distress signals are scattered across social media platforms and buried under a mountain of news updates, misinformation, and duplicate reports. Currently, responders lack an open-source, real-time tool to instantly ingest this unstructured noise, extract precise locations, assess emotional severity, and cluster duplicate requests into verified "Crisis Events."
The Solution: Sentinel FOSS bridges the gap between digital chaos and on-the-ground action. It is an end-to-end data ingestion and triage system that continuously monitors open feeds. By leveraging local Large Language Models (LLMs) and a multi-agent orchestration layer, the system automatically reads incoming text or transcribed audio, detects the urgency and emotional distress of the message, extracts key entities (location, resources needed, number of people), and maps them on a live dashboard.
Key Features:
High-Throughput Data Ingestion: Safely queues massive spikes in data during a crisis without bottlenecking the main server.
Multi-Modal Emotion & Urgency Analysis: Differentiates between a critical SOS ("Water is rising, trapped on roof") and general news ("Reports say water is rising") using an entirely local, privacy-first AI pipeline.
Agentic Deduplication: A verification agent queries historical and geographical context to cluster similar reports. If 50 people tweet about the same collapsed bridge, the map shows one high-priority event, not 50 overlapping markers.
Zero-Latency Command Center: A live, WebSocket-powered map dashboard designed for high-stress environments, ensuring responders see critical updates the second they are verified.
Technical Architecture:
Frontend (The Command Center): React.js with Leaflet/Mapbox for a responsive, live-updating geographic interface.
Backend & API: Node.js and Express to manage webhooks, rate limits, and client connections, backed by MongoDB for structuring processed crisis events.
Data Pipeline: Redis Streams/ kafka to handle real-time queuing of massive, asynchronous incoming data.
AI & Orchestration Layer: LangGraph to manage multi-agent workflows (Extraction Agent, Verification Agent, Clustering Agent), utilizing Qdrant for vector storage and local LLMs (like Llama 3) for rapid, cost-free NLP classification.
Social Impact: Sentinel FOSS democratizes crisis response technology. By relying entirely on open-source infrastructure and local models, it removes the dependency on expensive, proprietary APIs, allowing underfunded NGOs and local government bodies to deploy enterprise-grade intelligence during their most critical hours.