Agents
Status: Designed, not yet implemented.
In the memory architecture, agents are the workers responsible for maintaining the store. The infrastructure layer provides content-addressing, atomic writes, search, and the event log. Higher-level operations such as synthesis, decay, and conflict resolution are performed by those workers according to policy documents.
Agent identity
Section titled “Agent identity”Each agent has a name (unique within the colony), a role (defines capabilities and authority), and an author identity (every write is attributed). Multiple agents work concurrently.
A minimal deployment has five roles. The names below are mnemonics, not required implementation terms:
| Role | Responsibility | Trigger |
|---|---|---|
| Scribe | Ingest and parse raw input into typed objects | New data arrives |
| Analyst | Synthesize patterns from accumulated raw material | Object count exceeds threshold |
| Archivist | Decrement confidence scores, archive stale objects | Periodic schedule |
| Editor | Detect and resolve contradictions | Conflicting objects surface |
| Steward | Monitor health metrics, propose convention changes | Metrics degrade |
New roles are added by writing a policy document that defines the role’s triggers, procedures, and authority.
Execution
Section titled “Execution”An agent operates by:
- Reading applicable policies from the store
- Checking if trigger conditions are met
- Executing the referenced procedure
- Writing the result with provenance (author, source objects, confidence)
- Appending the event to the log
The policy defines the trigger, the procedure defines the work, and the output schema defines the result. In this design, changing the documents changes the behavior.
Scaling the colony
Section titled “Scaling the colony”When the system encounters gaps — unanswered queries cluster in a domain, or health metrics degrade in a specific area — new agents can be created:
| Type | Trigger | Mechanism |
|---|---|---|
| Renewal | Context exhaustion | Fresh agent takes over with compressed context from predecessor |
| Specialization | Domain mismatch | Generalist spawns a domain specialist |
| Delegation | Recurring subtask | Repeated work pattern justifies a dedicated agent |
System size, inference budget, and spawn approval are human-configurable parameters.
Open questions: At what scale do coordination costs outweigh specialization benefits? How should the colony balance inference budget across agents? These are empirical questions the implementation needs to answer.