AI Content Pipeline

apps site-builder ai claude gemini

Hybrid AI approach: Claude Sonnet generates website copy, Gemini 2.5 Flash generates hero/about images. Config-driven, model-swappable.

Content Generation (Claude)

Module: site_generator.py (1039 lines)

• Input: BusinessData (scraped from Google Maps) + optional WebsiteData
• Output: SiteContent — structured JSON with all section content
• Prompt includes business name, category, rating, reviews, hours, services
• Generates: hero headline/sub, about text, services list, FAQ, testimonials, CTAs, SEO metadata
• Section-level regeneration via /api/generate-section endpoint

Image Generation (Gemini)

Module: image_generator.py (302 lines)

• Uses Gemini 2.5 Flash Image model
• Generates hero background and about section image
• Prompt derived from business category and generated content
• Falls back gracefully if GEMINI_API_KEY not set

Pattern: Config-Driven Provider

The pipeline doesn’t hardcode Claude or Gemini. The model is selected via env vars and could be swapped. This matches the config-driven-routing pattern from Clarity.

Related

• site-builder-overview — full pipeline context
• config-driven-routing — routing logic in config, not code

---

Source: raw/site-builder-architecture.md, raw/site-builder-readme.md | Ingested: 2026-04-08