Search and Filter

Search and filter

One library, several ways into it. Every prompt and sequence you save is findable by text search, by tag, and by model, and the same finding tools repeat wherever the library surfaces: the main extension page, the sidebar's Insert view, and the // shortcut. Above them all sits universal search, one field across everything you have saved. This page is the canonical reference for finding things fast.

Prerequisites:

• Signed in: Your library requires a signed-in account. See sign in to the extension.
• Something saved: Search is only as useful as the library behind it. See save a prompt from the sidebar.

Where search lives#

Search the libraries#

The search field filters the list live, narrowing with every keystroke, and prompt search reads deep: it matches titles, prompt content, model, and tags, so a phrase you remember from inside a prompt is enough to surface it. An over-narrow query empties the list with a no-results notice such as No prompts found; shorten the query and the list returns. The Sequences view searches the same way over your sequences.

The // shortcut is the shallow, fast version: typing after //prompt filters by title as you type, built for reaching a prompt you already know by name. See insert a prompt with the shortcut.

Filter by tag, model, and library#

Filters narrow the list before or alongside search, and they stack:

• The tag filter: Shows only items carrying the selected tags. Filtering by a parent tag includes everything nested under it, which is what makes domain-level hierarchies pay off. Building the hierarchy is covered in manage tags.
• The model filter: Shows only items whose target model matches. The target model is organisational, so this filter is about grouping your library, never about where a prompt can insert.
• The library filter: Splits your prompts by origin: User library for prompts you created, and Linked Templates for prompts imported from the template library, which stay read-only until forked.

In the sidebar's Insert view, the tag and model filters appear as Select Tags… and Select Models, working the same way mid-chat. See insert a prompt from the sidebar.

Universal search#

Search Anything in the sidebar header is the everything field: one query across your prompts, sequences, conversations, and scratchpads, with a type label on every result telling you which is which. It works without signing in, searching whatever is already in local storage, synced or not.

• It is the one genuinely fuzzy search: Search Anything tolerates typos with real weighted-distance matching. Every other search on this page, the library search and the // shortcut alike, is plain substring matching.
• Results rank by relevance, not recency: title matches count most, then tags, then model, then content, with ties broken by type in this order: conversation, prompt, sequence, scratchpad, instruction profile.
• Opening a result jumps straight to the exact spot: the specific turn expanded in a conversation, or the exact item's edit modal, depending on the result type. A prompt or sequence result lands on the main extension page, where the item is managed.
• The keyboard works throughout: arrow keys move between results, Enter opens the highlighted one, and Escape closes the modal.

Universal search is the tool for the half-memory: when you know it exists but not which drawer it is in.

Related pages#

• Prompt and sequence interfaces
• Insert a prompt from the sidebar
• Insert a prompt with the shortcut
• Manage tags
• Manage the prompt library