KJP Phase 2 — AI Infrastructure (Deferred)
Parked specs for AI-driven tagging, descriptions, and LLM discoverability work. Not active.
- Status: Deferred to fall/winter 2026 (per Tori, 4/17)
- Reason: Johnsons aren't ready for AI yet + big mid-May client presentation absorbing bandwidth
- Original quote: $5,625 / 25 hrs (now superseded by Phase 1 split)
Files
| File | Description |
|---|---|
| KJP-AI-Feature-Brief.md | Client-facing AI brief — what they asked for, what we'd build |
| KJP-AI-Tagging-PRD.md | Full technical spec — Next.js/Vercel/Neon/Claude Vision architecture |
| KJP-AI-Cost-Estimate.md | Per-image token math, model comparison, OpenRouter strategy |
| KJP-AI-LLM-Strategy.md | S/M/L tiered AI options + LLM discoverability strategy |
| KJP-AI-Description-Tool-Plan.md | Pre-Casey-shift description-focused plan |
| KJP-PIPELINE-AND-CRON-PLAN.md | Pipeline + cron migration plan |
| claude-bundle-full/ | Full PRD, architecture, API specs (research bundle) |
Casey's priority order (from 3/10 follow-up)
- AI-generated tags — mood (uplifting, calm, soothing), specific flower/plant species, visual descriptors. #1 ask.
- AI color classification — dominant or multi-color tagging
- AI descriptions — open if reasonably priced and good for SEO; not the priority
- Region — NO. Handles manually (overlapping regions)
Voice/tone constraints
- Match Casey's existing description style
- Healthcare-focused client base — avoid negative, dark, threatening words
- Never use "magical" — Kurt hates it
Cross-references
- ../README.md — top-level project hub
- ../KJP-Project-Hub.md — canonical decisions hub
- ../phase-1/README.md — active build