ChainReaction LogoFeatured Case Study

ChainReaction: Graph-Powered Predictive Risk Intelligence

An autonomous GraphRAG ecosystem that continuously monitors global news, extracts risk entities using DSPy, and traces cascading impacts across multi-hop supply chain networks.

40%Reduction in Risk Detection Latency
100+Events Processed Per Minute
99%NER Extraction Reliability via DSPy

The Visibility Gap: Reactive vs. Proactive

Global supply chains are opaque. When a geopolitical event occurs—a strike, a port closure, or a regional conflict—logistics teams often don't realize they are affected until a shipment fails to arrive. This "Domino Effect" is difficult to track because relationships between tier-1, tier-2, and tier-3 suppliers are rarely mapped to real-world events.

"Traditional risk management is a backward-looking spreadsheet. ChainReaction is a forward-looking intelligence engine."

The Autonomous Ecosystem (LangGraph)

Scout Agent

The "eyes" of the system. Continuously monitors RSS feeds, geopolitical news, and trade data. It identifies potential risk signals and passes them to the orchestration layer.

DSPy Analyst

Uses DSPy-based risk extraction to convert unstructured text into structured JSON. Includes confidence scoring and validation to ensure 99% accuracy in entity mapping.

GraphRAG Engine

Powered by Neo4j, it traces multi-hop impact paths. It doesn't just find the supplier; it finds every component and assembly downstream that will be stalled.

Event-Driven Architecture

IngestionScout Agent feeds live news to the pipeline.
ExtractionLLM (GPT-4o/Ollama) with DSPy structures data.
CorrelationNeo4j maps entities to known supply chain nodes.
AlertingNext.js Dashboard & Webhooks notify stakeholders.

v1.0 Enhancements & Enterprise Readiness

Advanced AlertingNEW

Multi-channel delivery (Email, Slack) with escalation rules and acknowledgment tracking. Integrates directly with existing enterprise ERPs via bidirectional APIs.

Query PerformanceOPTIMIZED

Implemented query caching with TTL and batch processing, allowing the system to handle 100+ events per minute without degrading performance.

Local-First AI

Support for Ollama allows for deployments in secure, air-gapped environments, ensuring mission-critical data never leaves the network.

WCAG 2.1 AA

A fully accessible Next.js dashboard with force-directed graph visualizations, ensuring all decision-makers have equal access to insights.

The Technical Stack

Python 3.11
LangGraph / DSPy
Neo4j (Cypher)
FastAPI (v2)
Next.js 14
Docker Compose

Explore the Architecture

The complete codebase, documentation, and deployment guides are available on GitHub.

GitHub RepositoryBack to Projects