Real-Time Kafka and Generative AI Pipeline for Vehicular Route Analysis and Control Point Recommendations in Public Safety Contexts

This dataset contains latency and performance metrics collected from a pilot deployment of a real-time vehicular route analysis system using Kafka, Google Maps API, and a local LLM (DeepSeek / Llama 3) executed through Ollama.The system, developed within the SITIA infrastructure of Carabineros de Chile, analyzes the fastest route between a detection point and a known vehicle abandonment site.Using AI-assisted reasoning, it identifies possible control or observation points along that route to support operational decision-making in vehicle theft investigations. The dataset includes 9,530 total records, of which 9,427 were valid (98.92%), and provides timing breakdowns for each system component (Kafka ingestion, Google API query, and Ollama inference). This dataset accompanies the paper “AI-Assisted Route Analysis for Vehicular Control and Public Safety: A Real-Time Kafka–LLM System” (submitted to MDPI Applied Sciences).All data were collected under controlled operational testing, fully anonymized, and released under a Creative Commons Attribution 4.0 International (CC-BY 4.0) license.