Skip to content

knowledge-graph.md

Latest commit

 

History

History
135 lines (101 loc) · 3.33 KB

knowledge-graph.md

File metadata and controls

135 lines (101 loc) · 3.33 KB

Knowledge Graph

Gru builds a personal knowledge graph of your relationships, conversations, and history. Ask questions like "What did I discuss with John last month?" and get instant answers.

How It Works

Gru automatically extracts entities and relationships from:

  • Conversations with Gru
  • Emails (via Gmail)
  • Slack messages
  • Calendar events

Everything is stored locally in your database and indexed for fast retrieval.

Querying Your Knowledge

Natural Language Queries

"What did I discuss with John last month?"
"Who have I been meeting with recently?"
"What do I know about Project Alpha?"
"What topics have we talked about this week?"

Tools

Tool Description
ask_knowledge Answer natural language questions about your history
get_person_info Get details about a known person
get_recent_interactions Filter by entity, source, or time
get_known_people List all known contacts
get_recent_topics Topics discussed recently
add_person Manually add someone to the graph
add_relationship Add a relationship between entities
knowledge_stats View graph statistics

Entity Types

Gru tracks these entity types:

Type Examples
person John Smith, Sarah, @john
project Project Alpha, gru, website redesign
company Acme Corp, Google, Anthropic
topic API design, database migration, funding
tool React, PostgreSQL, Docker
place San Francisco, NYC office

Relationships

Entities are connected by relationships:

  • works_with - Professional relationship
  • knows - Personal acquaintance
  • works_on - Person works on project
  • discussed_with - Talked about topic with person
  • mentioned_with - Co-occurred in conversation

Time-Based Queries

All interactions are timestamped, enabling queries like:

  • "last month" - 30-60 days ago
  • "this week" - Past 7 days
  • "recently" - Past 14 days
  • "yesterday" - Past 2 days

Data Sources

Conversations

Every conversation with Gru is analyzed for entities and relationships.

Email

When Gmail is connected, Gru extracts:

  • Sender as person entity
  • Subject and topics discussed
  • Timestamp for time-based queries

Slack

When Slack is connected, Gru extracts:

  • Message senders
  • Channel context
  • Topics mentioned

Calendar

Calendar events provide:

  • Meeting attendees
  • Event topics
  • Time context

Privacy

  • All data stored locally in SQLite
  • No data sent to external services (except LLM for extraction)
  • Entity extraction uses Claude Haiku for efficiency
  • You control what gets indexed

Example Session

You: What did I discuss with Sarah last week?

Gru: Found 3 interactions:
  [2024-01-28] (conversation) Sarah: Discussion about API redesign
  [2024-01-27] (email) Sarah: Q1 planning meeting notes
  [2024-01-25] (slack) Sarah: Quick sync on deployment timeline

You: Who else is working on the API redesign?

Gru: About API redesign:
  - Sarah works_on API redesign
  - Mike works_on API redesign
  - discussed_with John (review meeting)

Statistics

View your knowledge graph stats:

You: knowledge stats

Gru: Knowledge Graph Statistics:
  Entities by type:
    person: 47
    project: 12
    topic: 89
    company: 8
  Total relationships: 234
  Total interactions: 1,847
  Interactions (last 7 days): 42