Building an AI Map Agent with Self-Hosted Planet-Scale OSM and Tiny LLMs
Yui Matsumura
How far can a fully open geospatial stack take a natural-language request like “Show me cafes in Hiroshima City”? This talk presents TRIDENT, text2geoql-dataset, and a self-hosted planet-scale OpenStreetMap stack for building a grounded, interactive AI map agent.
Conference Management Room1