Architect AI Workbench

What It Addresses

In early architectural design, much of the time is spent reconciling planning conditions, functional requirements, code indicators, and audience-specific explanations. A design lead needs to know quickly what the site allows, how a concept can be justified, whether indicators are at risk, and how the same proposal can be framed for clients, reviewers, and permit teams. This skill targets the schematic design stage, turning planning notices, redline drawings, task briefs, site photos, and other inputs into structured site analysis, concept narratives, typology strategies, indicator risks, and presentation-ready outputs.

How It Works

• Site and planning reading: capture plot ratio, height limit, setbacks, density, and other constraints in planning_conditions, producing a design brief that distinguishes what is allowed, what needs attention, and what cannot be optimized away. Optional site_analyze.py can derive site_brief.json from a DXF redline and coordinates.
• Concept generation: reason across location, culture, function, and form to shape theme, massing, circulation, facade direction, and a coherent concept narrative rather than disconnected ideas.
• Typology-aware design: apply different checklists for residential, commercial, office, educational, medical, industrial, urban renewal, and interior projects.
• Code and indicator support: flag risks related to fire safety, accessibility, parking, prefabrication, green building, and other common requirements, then produce technical indicator tables, code checklists, and missing-or-over-limit warnings.
• Scenario-based communication: use --scenario scheme|review|bid|client to adjust emphasis, writing style, and slide logic, supporting Word briefs, PPT decks, indicator tables, and compliance checklists.

Boundaries and Caveats

It is not a CAD, BIM, or rendering tool, and it does not replace registered architect sign-off, structural safety judgment, or final compliance decisions. Planning conditions should be treated as hard upper bounds, not parameters to be "optimized" past their limits. Indicators should retain both value and limit for traceability. If inputs are missing, ambiguous, or locally specific, the workflow should ask for clarification instead of inventing assumptions. It is best used for early condition structuring, design reasoning, review communication, and permit-preparation support.