NNyquest docs

Search and Tags

Two ways to find an artifact you saved a while back: search (text or semantic) and tags (manual categorization). Both work; combining them is best.

The search box at the top of the vault has two modes:

Fuzzy mode (default)

Substring match against title, content, and tags. Fast, predictable, no embeddings involved.

  • Case-insensitive
  • Substring — typing "auth" matches "authentication", "Auth class", "OAuth flow"
  • Boolean operators not supported — multi-word queries are fuzzy-OR'd
  • Best for: when you remember any text from the artifact

Semantic mode

Click the toggle to switch to semantic search (the AI/⚡ icon). Embedding-based similarity search using your saved artifact embeddings.

  • Slower (~200-300ms vs <50ms for fuzzy)
  • Finds artifacts conceptually related, not just text-matching
  • Example: searching "user login" semantic-mode finds an artifact titled "Auth flow with JWT" even without "login" in the text
  • Best for: when you remember the topic but not the exact wording

Switching modes is per-search; the next search defaults back to fuzzy.

Tags

Tags are short labels you attach to artifacts for filtering. Examples:

  • python, rust, ts — language
  • auth, db, frontend — domain
  • client-acme, project-q4 — work scope
  • keepers, gen — provenance flag

Add tags:

  • During save (in the Save dialog: Tags field, comma-separated)
  • After save (Detail panel → Tags → click + → type)

Tag chips at the top

The vault header shows the top 10 most-used tags as clickable chips. Clicking a chip filters the list to artifacts with that tag.

Multiple tag selection: click multiple chips. The vault filters to artifacts that match all selected tags (intersection, not union).

Click a selected chip again to deselect.

Combining search + tags

Filters compose:

  • Search "rust" + tag auth → all artifacts where text matches "rust" AND tag is auth
  • Search "" + tag auth + tag python → empty search but filtered by both tags

The list updates live as you type or click chips.

Type filter

The type chips (text, markdown, code, JSON, list) are another dimension of filter. Click "Code" to limit results to code artifacts.

Type filter persists across searches until you click a different type or "All".

Quick patterns

  • "Where's the Rust auth thing I saved last month?" — search "rust auth" with tag auth. Done.
  • "Show me all code I've saved" — type filter Code, no search.
  • "Anything tagged with my client name" — click client tag chip.
  • "Find that one snippet about JWT but I don't remember the title" — semantic mode, search "JWT".

When search misses

If you're sure you saved something but can't find it:

  1. Switch search modes. Fuzzy → Semantic might find what fuzzy missed (and vice versa).
  2. Try different keywords. Recall what you wrote in the actual content, not just what you remember the topic was.
  3. Check the type filter. Maybe filter is set to "Code" but the artifact is "JSON".
  4. Check tags. Multiple tags AND-filter; deselect to broaden.
  5. Check pinned section. Pinned items appear first; you might have pinned and forgotten.
  6. Sort by oldest. Sometimes the artifact is just on a deeper scroll, not missing.

If you still can't find it after the above, the artifact might:

  • Be on a different device's IndexedDB (vault is local-first per device)
  • Have been cleared by browser data wipe
  • Have been accidentally deleted

Keep a separate vault export or file backup. An embedding index is not a content backup and cannot be relied on to restore a deleted artifact.

Best-practice tagging

A few recommendations from heavy users:

  • 2-4 tags per artifact — more is unwieldy, fewer hurts findability
  • Use a hierarchy — broader tags first (code), then narrower (auth)
  • Be consistent with naming — client-acme and acme-client shouldn't both exist
  • Tag generated artifacts — they're harder to find later by text alone

Where to next