Overview
title: Project Overview author: Suhas Devmane date: 2025-09-20 category: Talking Buildings layout: post order: 1
Traditional conversational agents struggle with heterogeneous smart buildings. Devices, vocabularies, and schemas vary widely, so fixed responses don’t generalize. OntoBot tackles this with a modular, standards-aware stack:
- Natural language interface (Rasa) with custom actions
- Analytics microservice that accepts standardized payloads (flat/nested) and returns unit-aware insights with UK defaults
- Knowledge stores (SQL + SPARQL/Fuseki)
- Optional AI helpers (NL2SPARQL via T5, local LLM)
Users can ask questions like “What’s the CO2 in Room 2.11?” or “Any anomalies in HVAC today?” and receive clear, explainable responses, plus artifacts hosted by a file server for convenient sharing.
Explore the docs: