Roadmap

The MVP is complete. Phases 0-2 are shipped. The system is fully usable today.

What’s Built

Phase	Name	Status	Highlights
0	Foundation	✅	Git repo, Ralph loop, setup skill, docs
1	Vault Engine	✅	25 skills: vault management, 7 ingest types, query, lint, promote
1.5	Import & Private	✅	/vault-import, .private/ overlay
2	Tooling	✅	qmd search, Marp slides, image handling, Dataview, install script

Phase 0: Foundation

Git init, tooling, Ralph loop, docs.

• Public repo at github.com/RonanCodes/llm-wiki
• /ralph skill + ralph.sh for autonomous builds
• /setup for first-time bootstrap
• Workflow and dependency docs

Phase 1: Vault Engine

Core vault system — 25 skills.

• /vault-create, /vault-status, /vault-import, /vault-archive
• wiki-templates — 5 page types with frontmatter specs
• /ingest router + 7 sub-skills (web, PDF, Office, YouTube, tweet, gist, text)
• /query with --save to compound knowledge
• /lint with 8 checks and --fix auto-repair
• /promote for cross-vault knowledge transfer
• .private/ overlay for private skills

Phase 2: Tooling (MVP Complete)

Ecosystem tools from Karpathy’s recommendations.

• /search — qmd hybrid BM25/vector search + grep fallback + MCP server
• /slides — Marp presentation generation from wiki content
• Image handling — download images locally during ingest
• Dataview compatibility — queries, dashboard template, field reference
• install.sh — interactive macOS installer
• Astro landing site + Starlight docs

What’s Next

Community feedback will drive priorities. Some ideas being explored:

• Local web app — Next.js dashboard with chat interface for non-CLI users
• More source types — Slack threads, meeting transcripts, Kindle highlights
• Better search — deeper qmd integration, semantic queries
• Collaboration — shared vaults, team workflows

Key Metrics

Metric	Current
Skills built	25
Source types	7
Core operations	4 (ingest, query, lint, promote)
Page types	5 (source-note, entity, concept, comparison, summary)
Core external deps	0 (curl only)
Optional external deps	5 (yt-dlp, poppler, pandoc, qmd, marp)