Skip to content

/recall — Semantic Search

Searches the knowledge base using natural language queries and returns ranked results with full provenance.

Usage

/recall distributed caching strategies
/recall --type bookmark --limit 5 API design
/recall --author alice --project billing authentication

Trigger phrases: "what do we know about", "search knowledge", "find in knowledge base"

When to Use

  • Finding relevant entries on any topic
  • Searching for past decisions, discussions, or bookmarks
  • Filtering by type, author, or project

Options

Option Description Default
--type <type> Filter by entry type (session, bookmark, minutes, etc.) All types
--author <name> Filter by author All authors
--project <name> Filter by project All projects
--limit <n> Maximum results to return 10

If no query is provided, you'll be prompted for one.

Output

Each result includes:

  • Similarity score — percentage match (e.g., 92%)
  • Type badge — [session], [bookmark], [minutes], etc.
  • Full content — never truncated
  • Provenance — entry ID, author, project, creation date
  • Tags — if any are present
## Results (3 found)

### 92% [session] Caching Strategy Decision
We decided to use Redis for the session cache with a 15-minute TTL...

ID: a1b2c3d4 | Author: alice | Project: billing | 2026-03-15
Tags: domain/caching, project/billing/decisions

---

### 78% [bookmark] Redis Best Practices
Summary of Redis configuration recommendations from the official docs...

ID: e5f6g7h8 | Author: bob | Project: billing | 2026-03-10
Tags: source/bookmark/redis.io, domain/caching

Tips

  • Queries are semantic, not keyword-based — "how do we handle auth?" finds entries about authentication even if they don't contain the word "auth"
  • Use --type to narrow results when you know what you're looking for
  • If an invalid --type is provided, you'll see the list of valid types