Social Network Module
A persistent social graph of users, agents, and groups. Captures identity, expertise, contact details, and behavioral personas that evolve through conversation analysis.
Entity Types
The social graph tracks three types of entities. Concepts, platforms, and APIs are not entities— they belong as keywords.
userHuman users interacting with the agentagentOther AI agents in the systemgroupTeams or squads acting as a collectiveEntity Structure
entity_descriptionNatural language description, auto-updated via LLM after conversationsidentity_infoStructured fields — organization, position, backgroundcontact_infoMulti-channel contacts — email, matrix, slack (nested JSON, deep-merge friendly)keywords3–5 contextual tags — topics, domains, expertise levels (strictly enforced)personaCommunication persona auto-generated from conversation patternsembeddingSemantic vector for similarity search across the graphHow the Graph Updates
Entity data is updated through two complementary mechanisms. The module exposes tools that the agent uses to capture information as it flows through conversations:
The agent can also search the graph semantically using search_social_network to find people by expertise, domain, or role, look up contact details with get_contact_info, or reach out to someone through their best available channel via contact_agent. When duplicate entities are detected, they can be consolidated with merge_entities.
Tools
extract_entity_infoUpdate entity identity, contact, tags, keywordssearch_social_networkSemantic search by expertise, domain, or roleget_contact_infoQuick contact lookup for a known entityget_agent_social_statsRelationship summary — recent interactions, strongest tiescontact_agentSend message to entity through best available channelcheck_channel_updatesCheck for updates across all registered channelsmerge_entitiesConsolidate duplicate entity records