ai-concept-generation

name	ai-concept-generation
description	Use when preparing landscape AI concept prompts, generating multiple visual directions from a measured base plan, logging generation batches, selecting variants, and protecting geometry from AI drift.

AI Concept Generation

Goal

Generate managed concepts from measured plan, brief, site analysis, and designer logic instead of random beautiful images.

Before Generation

1. Check for AI-ready base in 02_base_plan/04_ai_ready_base.
2. Check 03_client_context/BRIEF.md and 03_client_context/SITE_ANALYSIS.md.
3. Create 3-5 written concepts in 04_ai_concepts/01_concept_variants.
4. Fill the structure from 04_ai_concepts/CONCEPT_BRIEF_TEMPLATE.md for each concept.
5. Prepare common inputs in 04_ai_concepts/00_common_inputs.

Generation

1. Run a separate batch for each concept.
2. Use prompts from 90_prompt_library as a base, but save the actual prompt.
3. Record each batch in 04_ai_concepts/GENERATION_LOG.md: source files, exact prompt, service/model, settings, result files, selection reason, geometry risks.
4. Check: boundaries, house, entry, permanent structures, utilities, proportions.
5. Reject options where AI broke geometry even if they look attractive.

Selection

In 04_ai_concepts/SHORTLIST.md, record:

• what is shown to the client;
• which elements can be used in the project;
• which elements are mood-only;
• risks and caveats.

Output

• 3-5 written concepts.
• Filled generation log.
• Shortlist of 5-7 client-facing options.