3ad61b984d
Skills (new): - skills/obsidian-markdown/ — full Obsidian Flavored Markdown syntax reference (wikilinks, embeds, callouts, properties, math, Mermaid) - skills/obsidian-bases/ — Obsidian Bases (.base files) with correct filters/views/ formulas syntax (sourced from kepano/obsidian-skills authoritative spec) - skills/defuddle/ — web page cleaner; strips ads/nav before URL ingestion, saves 40-60% tokens on web articles wiki-ingest upgrades: - URL ingestion: pass https:// directly, auto-fetches + runs defuddle if available - Image/vision ingestion: .png/.jpg/.gif etc → Claude reads → description saved to .raw/ → standard ingest pipeline - Delta tracking: .raw/.manifest.json tracks hash per source, skips unchanged files wiki-query upgrades: - Quick mode (query quick:) — hot.md + index only, ~1500 tokens - Standard mode — existing behaviour, 3-5 pages - Deep mode (query deep:) — full wiki + optional web search supplement hooks: - PostToolUse auto-commit: every Write/Edit to wiki/ or .raw/ triggers git add + commit automatically, vault always versioned fixes: - Removed invalid allowed-tools field from all 10 SKILL.md files (not a valid skill frontmatter attribute per spec; was silently ignored) - Canvas SKILL.md now references json-canvas open standard and kepano/obsidian-skills wiki research: - Ecosystem research: 16+ Claude+Obsidian projects mapped and filed - New pages: comparisons/claude-obsidian-ecosystem, concepts/cherry-picks, entities/ (6 new), sources/claude-obsidian-ecosystem-research - Cherry-picks roadmap filed at wiki/concepts/cherry-picks.md Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
1.9 KiB
type, title, created, updated, tags, status, related, sources
| type | title | created | updated | tags | status | related | sources | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| overview | Wiki Overview | 2026-04-07 | 2026-04-07 |
|
developing |
|
Wiki Overview
Navigation: index | hot | log | dashboard
Purpose
This is the claude-obsidian demo vault. It demonstrates the LLM Wiki Pattern — a system for building persistent, compounding knowledge bases using Claude and Obsidian.
Run /wiki to scaffold this vault for your own domain and replace this overview.
Current Seed Content
Concepts seeded:
- LLM Wiki Pattern — the core architecture
- Hot Cache — session context mechanism
- Compounding Knowledge — why the pattern works
Entities seeded:
- Andrej Karpathy — originated the pattern
Sources seeded:
- claude-obsidian-ecosystem-research — 16+ projects, 13 cherry-picks identified (2026-04-08)
Current State
- Sources ingested: 2
- Wiki pages: 25
- Last activity: 2026-04-08 (ecosystem research ingest — 16+ projects mapped)
Canvases
- claude-obsidian-presentation — Full presentation: hero, overview, skills, architecture, Wiki vs RAG, visual demos (2026-04-07)
- AI Marketing Hub Cover Images Canvas — Cover image library for AI Marketing Hub brand assets
Key Themes
Knowledge compounds. Unlike RAG, the wiki pre-compiles synthesis. Cross-references are already there. Contradictions are flagged. Every ingest enriches existing pages rather than adding isolated chunks.
The hot cache is the force multiplier. A ~500-word file captures recent context. New sessions start with full context at minimal token cost.
Obsidian is the IDE, Claude is the programmer. The graph view shows what's connected. The human curates sources and asks questions. Claude writes and maintains everything else.