ROADMAP — Multi-Library & AI Optimization

Status (Aug 2026)

Phase Status
1 — Editable title ✅ Done
2 — libraries schema + switcher MVP ✅ Done (SQL applied; app reads/writes libraries)
3a — buildLibraryContext ✅ Done (src/lib/library-ai-context.mjs)
3b — Public JSON-LD ✅ Done (/l/{token})
3c — Gemini auto-fill title/description ✅ Done (library-ai Function + UI)
4 — AI export / recommend / README ⏳ Partial — recommend empty-library examples ✅
5 — Multi-library UX ✅ Done (shipped with Phase 2)
Cleanup — drop dormant user_profiles.library_* ⏳ Optional later

user_profiles note: Library title/description/share live on libraries only. user_profiles is still used for author fields (public_display_name, public_avatar_url, public_show_author). Legacy library_* / share_* columns on user_profiles are unused by the app after migration — safe to leave until a dedicated drop migration.

Public discovery (read): MCP list_public_libraries / search_public_libraries / get_public_library, site search type: "library", sitemap, and chatbot proxy (parallel public-library search) all read share-enabled libraries.

Context

PATTTTERNS supports N libraries per user with editable title, share token, and AI context description. Architecture: Supabase REST + RLS from the client (library-cloud.ts / bookmark-cloud.ts) — no Prisma.

Constraints:

  • Patterns may belong to one or many libraries (M2M via bookmark_on_library).
  • Pricing / plan gating is not in scope.
  • Token share URLs (/l/{token}) are canonical; username slugs stay later (Public Libraries PRD).

Goals

  • Editable library title
  • Library context (description) for AI
  • Multiple libraries per user + switcher
  • Patterns assignable to multiple libraries
  • Gemini auto-fill when title is default and/or description empty
  • AI recommend examples for empty libraries with context
  • AI export / README (rest of Phase 4)

Non-Goals (this roadmap)

  • Pricing tiers or isPro enforcement.
  • Username SEO slug URLs (tracked in Public Libraries PRD).
  • Full CMS / Notion replacement.

Data model (current) 

auth.users
    │ 1:N
    ▼
libraries  (title, description, share_token, share_enabled, is_default, position)
    │
    │ M:N via bookmark_on_library
    ▼
bookmarks  (pattern payload)

user_profiles  — author display fields only (not library chrome)

SQL: docs/queries/DB_LIBRARIES_MULTILIBRARY.sql

Phased Roadmap

Phase 1 — Editable Library Title ✅

Shipped: inline rename (now via library tabs), persistence on libraries.title.

Phase 2 — Schema + multi-folder MVP ✅

  • libraries + bookmark_on_library + RLS + backfill
  • library-cloud.ts / bookmark-cloud.ts cut over
  • Share via /l/{token} + /library?token=
  • Switcher, create, rename, delete (non-default), ?lib=
  • Save picker when user has multiple libraries
  • Drawer filter + per-library counts
  • Public shared library: save-to picker + duplicate as new owner library
  • Drop dormant user_profiles.library_* / share_* columns (optional)

Phase 3 — AI context ✅ (core)

3a — Structured context object ✅

src/lib/library-ai-context.mjs — buildLibraryContext() grounds title, curator, description, top tags, and up to 20 patterns.

3b — JSON-LD for public libraries ✅

Canonical /l/{token} injects ItemList JSON-LD (Public Libraries PRD).

3c — Gemini auto-fill ✅

  • Netlify Function: POST /.netlify/functions/library-ai (alias /api/v1/library-ai)
  • Auth: owner Bearer token
  • Fills title only if still default (My Pattern Library / empty)
  • Fills description only if empty
  • Requires ≥1 pattern in the library
  • UI: Auto-fill with AI on owner /library context field

Phase 4 — AI Features: Export & Build From Patterns ⏳

  • Pattern recommender (empty library → Auto-load examples with AI from title/context)
  • Code export from library context
  • Shareable README / summary copy

library-ai action: "recommend" ranks search-index candidates, asks Gemini to pick 5–8, client upserts into the active library.

Phase 5 — Multiple Libraries Per User ✅

Merged into Phase 2 ship (client Supabase REST, not Next /api/libraries routes).

Exit criteria:

  • Create / rename / delete additional libraries
  • Patterns in multiple libraries (join table + save picker)
  • Default library keeps working
  • Switcher on /library

Public listing (MCP / search / bot)

Surface Status
MCP list_public_libraries ✅ reads libraries
MCP search_public_libraries ✅
MCP get_public_library ✅
Topbar search (type: "library") ✅ listNonEmptyPublicLibraries
Chatbot proxy ✅ parallel search_public_libraries in prompt
Sitemap /sitemap-libraries.xml ✅

Critical path doc

Close stage and multilibrary path — Stage close decisions (no public_library_index table; branded OG deferred).

Migration safety

  1. Do not drop user_profiles.library_* in the same deploy that still reads them — app no longer reads them; drop can be a separate SQL change when ready.
  2. Default library is created lazily via ensureDefaultLibrary() (internal) / SQL backfill.
  3. Deleting a library detaches join rows; bookmarks remain unless removed from all libraries.