Glossary
Core concepts used across ARDH research notes and product documentation.
- Architectural Intelligence Layer (AIL) — A structured layer of architectural and planning knowledge—buildability rules, typology constraints, spatial heuristics—that sits between raw geospatial data and AI recommendations.
- Knowledge-Driven Land Intelligence (KDLI) — A methodology for land evaluation that grounds AI outputs in curated architectural, regulatory, and GIS knowledge rather than map similarity or market scores alone.
- Architectural Reasoning Engine (ARE) — The reasoning component that applies typology playbooks and constraint templates to a structured site model to produce program-conditioned feasibility briefs.
- Constraint-Aware AI — AI systems whose outputs are filtered or shaped by explicit planning, design, and access constraints rather than unconstrained language generation.
- Planning Intelligence Engine — A subsystem that maps zoning designations, overlays, and municipal practice into structured rule packs usable by AI and architects.
- Land Intelligence Graph — A knowledge graph of parcels, zones, typologies, constraints, and precedents used to ground retrieval and reasoning for land evaluation.
- Knowledge Graph — A network of entities and relations (nodes and edges) that represents domain structure for retrieval, reasoning, and explanation.
- GIS — Geographic Information Systems: layers of spatial data such as parcels, roads, zoning polygons, and amenities used for location context.
- Urban Planning — The practice of regulating land use, density, access, and public realm outcomes through codes, plans, and approval processes.
- Buildability — The practical capacity of a parcel to host a given program under zoning, access, topography, setbacks, and construction constraints.
- Constraint Graph — A structured representation linking site entities to rules (height, setbacks, access, use) that bound feasible design outcomes.
- Vector Search — Retrieval of documents or entities by embedding similarity in a high-dimensional vector space.
- LLM — Large Language Model: a neural model trained to predict and generate text, often used for synthesis and explanation.
- Agentic AI — AI systems that plan and execute multi-step workflows—tool calls, retrieval, checks—toward a goal rather than a single completion.
- RAG — Retrieval-Augmented Generation: combining document or graph retrieval with an LLM so answers cite external knowledge.