v0.3.1

Add @latest tag to all install commands (Claude Code, Claude Desktop, Cursor) so users always get the newest version automatically.

Full Changelog: v0.3.0...v0.3.1

v0.3.0 — Major quality release

44 tools, async I/O, search index with TF-IDF, concurrent write safety, 152 tests.

New tools

• extract_links — All wikilinks and URLs from a note
• delete_note — Permanently delete a note
• bulk_tag — Add/remove tags across multiple notes
• daily_notes — List all daily notes with stats
• export_note — Export as clean markdown or JSON

New Vault category

Bulk operations and vault management tools are now grouped under a dedicated Vault category (10 categories total).

Performance

• All filesystem I/O converted to async (fs/promises) — non-blocking on large vaults
• In-memory search index with incremental updates via file watcher
• TF-IDF scoring — rare terms rank higher than common ones for better search relevance
• Index pre-filters search candidates before reading files from disk

Reliability

• Per-file concurrent write locking — prevents data corruption when multiple operations target the same note
• Proper YAML frontmatter handling via yaml library (no more data corruption)
• Collision detection on note creation (no silent overwrites)
• Backlink updates on rename_note and move_note
• Graceful shutdown with proper resource cleanup

Search

• Query parser with path:, tag:#, file:, "exact phrase" operators
• find_by_date uses actual file modification time (not string search)
• searchBatch() reads each file once for multi-query operations

Quality

• 152 tests across 8 files (up from 45 across 4)
• MCP protocol integration tests
• Concurrent write safety tests
• TF-IDF ranking tests
• Tool integration tests with real vault operations

Fixes

• Fixed journal heading duplication bug
• Fixed biased shuffle with Fisher-Yates
• Fixed find_gaps/find_orphans subdirectory matching
• Fixed config boolean coercion
• Fixed patchNote heading level detection
• Fixed unused tags parameter in remember_idea and remember_reference
• Removed dead get_active_note tool
• Auto .md extension normalization on all file operations

Full Changelog: v0.2.1...v0.3.0

v0.2.1

Updated README with upgrade instructions and @latest tag recommendation.

v0.2.0

What's new

Smarter search

• Multi-word fuzzy matching — "auth decision" finds notes containing both words even if they're not adjacent
• Smart scoring — title matches, exact phrases, and recently modified notes rank higher
• Token-based fallback — still returns results when no exact phrase match exists

Recency sorting

• find_recent now sorts by actual file modification time
• Shows relative timestamps ("2h ago", "3d ago")

File watching

• Vault changes are detected instantly via fs.watch
• No more waiting for the 5-second cache TTL to expire

Fixed

• OG image rendering error (duplicate font-family attribute)

v0.1.1

Fix: Include README in npm package so it displays on npmjs.com.

v0.1.0

MindCache v0.1.0

Your Obsidian vault, connected to AI.

What's included

• 40 MCP tools across 9 categories: Find, Read, Remember, Journal, Write, Connect, Organize, Understand, Tasks
• 4 MCP resources for automatic context injection
• 10 note templates: decision, meeting, learning, idea, person, project, book, article, weekly-review, session
• CLI: init, doctor, config commands
• Direct filesystem access — no Obsidian plugins required
• Landing page at usemindcache.com

Get started

npx @augmnt-sh/mindcache init
claude mcp add --scope user mindcache -- npx @augmnt-sh/mindcache