Armadillo

Overview

Armadillo Code Compliance is an automated reasoning engine designed to integrate Building Information Models (BIM) with formal knowledge graphs and LLM-powered semantic reasoning to check compliance with local codes.

By leveraging semantic web technologies (RDF/OWL, SHACL) ontologies, knowledge graphs, LLM agents, and geometric tools, Armadillo automatically extracts spatial topologies, structural metadata, and functional attributes from BIM formats. The raw geometric data is transformed into raw geometric data into queryable, rule-evaluable schemas.

The core of Armadillo is a retrieval and reasoning engine built to translate building codes into rigorous queries, and execute the queries against BIM data formats such as Industry Foundation Classes (IFC). The engine leverages AI agents to build ontologies and queries, and deterministic python with robust geometry tools to execute the queries, taking advantage of both the semantic power of LLMs and the reliability of pre-built reasoning tools.

When running automated code compliance checks, the engine flags non-compliant building components directly within 3D views to accelerate safety and structural audits.

Example of a building code requirement processed into rigorous logic via ontology and formal grammar.

Example of robust python code generated by the pipeline to check the requirement given a .ifc building plan.

Example of Armadillo engine detecting building elements failing code compliance rules in 3D view.

Key Features

• Automated Ontology Generation: Converts code logic into entities and relationships, with automated schema mapping to RDF triples and OWL ontologies.
• Semantic Resolution: Designs rigorous logic implemented as qraph queries which faithfully reflect complex building code constraints expressed.
• Spatial Topology Graphing: Computes room adjacencies, circulation paths, and structural containment topologies directly from spatial geometry. Uses a robust library of pre-built geometry functions.
• Verdict and Provenance: Executes the queries produced by the engine and tracks all logical steps, providing an automatically generated summary of results with non-compliant elements singled out and the reasons for failure highlighted.

System Architecture

  +---------------------+
  |    Requirements     | (Building Codes)
  +----------+----------+
             |
             v
  +---------------------+
  |    Formal Rules     | (Codes -> Ontology and Formal Grammar)
  +----------+----------+
             |
             v
  +---------------------+
  |  Semantic Reasoner  | (Formal Rules -> Deterministic IFC Queries)
  +----------+----------+
             |
             v
  +---------------------+
  |   Query Execution   | (Uses IFC, Python, and Geometry Library)
  +----------+----------+
             |
             v
  +---------------------+
  |  Compliance Report  | (Automated Audit Results, With Provenance)
  +---------------------+

Technical Stack

• Ontology & Knowledge Graphs: OWL, IFC4 / IFC4.3 Schema Definitions
• BIM and Geometry Parsers: Python (IfcOpenShell, Shapely), Neo4j
• AI Agents: Gemini API, Claude API