HEXi's Domain system organises knowledge by area of expertise and assigns a council agent as the authoritative curator of each domain. Domains are the bridge between the governance layer and the knowledge graph — they determine which agent speaks with the most authority on which topics.
HEXi distinguishes between two domain layers, each serving a different purpose in the governance architecture.
Actor Domains
Each of HEXi's 21 council agents is the expert for one Actor Domain. This domain defines the body of knowledge that agent is responsible for curating, maintaining, and expanding. When a decision falls within an agent's domain, their vote carries additional weight in Tier 2 governance. Actor Domains are the governance-layer concept of expertise.
Entity Type Domains
Entity types in the knowledge graph are grouped into ontology domains — logical clusters like "Project Execution", "Risk & Governance", or "People & Organisation". These groupings help the Librarian agent and human administrators understand the structure of the knowledge graph at a glance, and guide LLM extraction when episodes are ingested.
Each council agent is named after a character from Terry Pratchett's Discworld and assigned a domain that reflects their character's essence. This is not merely thematic decoration — each agent's instructions, bias declarations, and pre-mortem signals are drawn from deep character analysis.
Domains are active participants in the governance process — not just labels.
Agents maintain private notes in their actor workspace — observations, patterns, and heuristics they've developed through work. When a note matures into a reusable insight, the agent curates it and publishes it to their domain via the Episode Queue.
Published episodes are tagged with the agent's domain, framed as entity/relationship observations, queued for human review, and — once approved — ingested into HEXi permanently. This makes domain knowledge cumulative: the longer an agent operates, the richer their domain becomes.