Roadmap

The plan is a 7-week arc from greenfield to a publishable, pipeline-driven PLM knowledge base. Current status:

Week	Scope	Status
1	Foundations — repo scaffold, ontology code, validator, CI plumbing	✔ Done
2	Seed ontology — 44 hand-authored notes across the 4 layers	✔ Done
3	Custom Quartz — LayerBreadcrumb, RelatedByRelation, layer color tokens	✔ Core done
4	Pipeline skeleton — Wikidata + Wikipedia sources, Markdown writer, PR opener	✔ Done
5	LLM integration — classify + normalize + aliases + relate (Haiku + Sonnet)	✔ Done
6	RDF export — Turtle + JSON-LD + per-layer shards + SHACL in CI	✔ Done
7	Polish — workflows un-gated, changelog, runbook, docs	✔ Done

Deferred (documented in CHANGELOG)

• Graph view colored by layer — needs forking Quartz’s Graph.tsx.
• Layer-faceted search chips — extend ContentIndex + preact shim.
• Wikidata redirect/merge handling — follow 301 to canonical QID.
• Analytics (Plausible) and private-notes support (visibility field already in schema).

What’s next after the plan

1. Run LLM enrichment at scale. Point plm enrich --terms at ~500 candidate concepts culled from Wikidata PLM/CAD/MBSE categories; review PRs in batches.
2. Deploy. Set CF_API_TOKEN / CF_ACCOUNT_ID repo secrets; the existing deploy.yml will take over.
3. Attach the weekly cron. Set ANTHROPIC_API_KEY; the enrich.yml Monday cron will start opening review-able PRs.