Social Network Module

Overview

The Social Network module maintains a graph of entities and their relationships as perceived by each agent. This enables agents to understand social dynamics, recall relationship history, and tailor their behavior based on interpersonal context. The module builds and updates the graph automatically as the agent interacts with users and learns about their connections.

Entity Graph

At the core of this module is a directed graph where nodes represent entities (people, organizations, concepts) and edges represent relationships between them. Each relationship edge carries attributes including type (colleague, friend, reports-to), strength, sentiment, and a history of interactions that informed the relationship assessment.

Context Contribution

During the Context Building step, the Social Network module identifies entities mentioned in the current interaction (using output from the Awareness module) and retrieves their relationship context. If a user mentions a colleague, the agent can recall previous discussions involving that person, the nature of the relationship, and any relevant social dynamics.

MCP Tools

The module exposes tools for the LLM to explicitly query or update the social graph. social_lookup retrieves the relationship context for a given entity. social_connect creates or updates a relationship between two entities. social_graph returns a subgraph centered on a specified entity, useful for understanding complex social networks the agent has observed over time.