search_brain

Benchmarked against: Anthropic — Web fetch tool MCP Tool: Available on both Cloud UB and Local UBI servers Cost: Free (internal search)

search_brain is the primary knowledge retrieval tool — a hybrid search combining semantic vectors and keyword matching to find relevant entries in Universal Brain.

---

Usage

search_brain(query="LangGraph checkpoint API", limit=10)

Parameters

Parameter	Type	Default	Description
query	string	required	Search query (supports Chinese and English)
limit	number	10	Maximum results to return

Response

Returns a JSON array of matching entries, ranked by combined relevance score:

[
  {
    "entry_id": "ub-abc123",
    "title": "LangGraph Checkpoint Architecture",
    "category": "knowledge",
    "score": 0.87,
    "snippet": "..."
  }
]

How it works

Hybrid search combines two strategies:

Strategy	Technology	Strength
Semantic	Gemini embedding-001 (768d vectors)	Finds conceptually related content
Keyword	SQLite FTS5 full-text search	Finds exact term matches

Results from both strategies are merged, deduplicated, and ranked by combined score.

Best practices

1. Search UB first — before any external search or decision
2. Use English queries — embedding model works best with English
3. Be specific — "LangGraph checkpoint sqlite" beats "checkpoint"
4. Check before ingesting — avoid duplicates by searching first

Related tools

Tool	Purpose
intel_search	External web search + auto-ingest
search_by_category	Browse by category
get_entry	Fetch single entry by ID
get_recent	Latest ingested entries