SYNAPSE: Synthesizing Narratives from Agentic Path Spatial Exploration

We introduce Synapse, a modular, privacy-preserving framework for generating multilingual summaries of logistics routes, a task distinct from traditional routing optimization or forecasting. Our approach transforms static route definitions by synthesizing them with a variety of signals, including pre-computed failure predictions, real-time traffic, and weather, through a network of specialized AI agents, orchestrated using the FastMCP (Multi-Agent Communication Protocol) framework. To ensure GDPR compliance, we apply random perturbation to GPS coordinates before querying external APIs, a method that obfuscates precise locations while preserving local context. The system simulates the route’s temporal progression via iterative calls to a traffic API, enabling time-aware queries for other agents gathering weather and urban context data. A generative agent fuses these multimodal signals into a core narrative. Parallel translation agents then render that narrative in multiple languages. This work presents a scalable pipeline for interpretable logistics intelligence, introducing an agentic architecture and a human-validated, multilingual evaluation benchmark.