exploring-knowledge-graph

Exploring the Knowledge Graph

Brain stores knowledge as an interconnected graph: notes link to other notes via wikilink relations, observations capture categorized facts, and semantic search with depth traversal reveals context that simple queries miss.

Triggers

Trigger Phrase	Operation
what do you know about X	Full knowledge graph traversal
how do I explore the knowledge graph	Graph exploration workflow
how are these concepts connected	Relation traversal via depth parameter
give me comprehensive context on X	Deep multi-phase exploration
map out related knowledge for X	Note discovery and relation linking

---

When to Explore

Explore the knowledge graph when:

• Starting complex work that spans multiple topics
• User asks "what do you know about X"
• Planning requires understanding existing decisions/patterns
• Investigating how concepts connect across projects
• Need comprehensive context, not just top search results

Exploration Phases

Track visited note identifiers to prevent cycles. Execute phases sequentially.

Phase 1: Semantic Entry Point

mcp__plugin_brain_brain__search({
  "query": "<topic>",
  "mode": "semantic",
  "limit": 10
})

Collect: primary note results with titles, permalinks, and content previews.

Phase 2: Expand Note Details

For key notes from Phase 1, get full content including observations and relations:

mcp__plugin_brain_brain__read_note({
  "identifier": "<note title or permalink>"
})

Extract: observations (categorized facts), relations (wikilinks to other notes), tags.

Phase 3: Graph Traversal via Depth Search

Use the depth parameter to traverse the knowledge graph through relations:

mcp__plugin_brain_brain__search({
  "query": "<topic>",
  "mode": "semantic",
  "depth": 2,
  "limit": 10
})

Depth levels:

• depth: 1 - Direct relations (1-hop connections)
• depth: 2 - Relations of relations (2-hop connections)
• depth: 3 - Deep traversal (3-hop connections, use sparingly)

Phase 4: Folder Discovery

Discover related notes by browsing semantic folders:

mcp__plugin_brain_brain__list_directory({
  "dir_name": "decisions",
  "depth": 2
})

Key folders to explore: decisions/, analysis/, sessions/, specs/, planning/, skills/

Phase 5: Synthesize

Group findings by type and present structured results.

Presenting Results

Group findings by type:

Notes: Primary (direct matches) -> Related (via depth traversal) -> Folder-discovered

Observations: Key facts, decisions, and insights extracted from notes

Relations: Wikilink connections between notes showing how concepts relate

Graph Summary: Total notes found, key themes, relation patterns, suggested follow-up queries

Depth Control

• Shallow (phases 1-2): Quick context, ~5-15 notes
• Medium (phases 1-4): Include graph traversal and folder discovery
• Deep (all phases): Full graph traversal with synthesis

Match depth to task complexity. Start shallow, go deeper if context insufficient.

When to Use

Use this skill when:

• Starting complex work spanning multiple topics
• User asks "what do you know about X"
• Planning requires understanding existing decisions and patterns
• Investigating how concepts connect across projects

Use curating-memories instead when:

• Updating, editing, or linking specific notes
• Cleaning up duplicate or stale content

---

Anti-Patterns

Avoid	Why	Instead
Running all 5 phases for simple queries	Wastes tokens on unnecessary traversal	Start shallow (phases 1-2), go deeper only if needed
Not tracking visited identifiers	Causes infinite cycles in graph traversal	Maintain a visited set, skip already-seen notes
Using depth:3 on broad queries	Exponential blowup on dense graphs	Start with depth:1, increase only if needed
Skipping folder discovery when notes reference unknown entities	Misses cross-domain connections	Check list_directory for related folders
Presenting raw results without grouping	Overwhelming and unstructured	Group by type: notes, observations, relations

---

Verification

After graph exploration:

• [ ] Entry point search returned relevant results
• [ ] No cycles encountered (visited identifiers tracked)
• [ ] Depth matched task complexity (shallow/medium/deep)
• [ ] Results grouped by type for readability
• [ ] Follow-up queries identified if context was insufficient

---

Efficiency Tips

• Start with depth: 1 and increase only if needed
• Use folder parameter to scope searches to specific domains
• Skip Phase 4 if Phase 3 already found sufficient related notes
• Stop expanding when hitting diminishing returns