Which loops the agent runs, and for each one that is off, why.
Optional freezeFreezes self-change, persisted in the learning store: no new candidate is made and the memory tools refuse
(frozen), while reverts, operatorSet and forgetUser still run. With userId, that user's learner; without,
the whole agent (every learner of the layer, for an agent with no key).
Optional target: { Optional userOptional by: stringOptional unfreezeLifts a freeze AgentLearning.freeze set on the same target.
Optional target: { Optional userOptional suspendOptional unsuspendLifts a suspension AgentLearning.suspend set on the same target.
Optional target: { Optional userOptional operatorSets a user's traits, honesty-humility and emotionality included, through the ledger; 'baseline' resets them.
Optional note?: stringOptional revertTakes back a kept change by its ledger id: an operator's trait change, an import, or a recorded memory write.
Optional exportA user's learned state as one JSON document.
Optional importTakes a document in: its traits, clamped to this agent's persona; what this release cannot admit is refused. With
target, into that user's learner of this agent; without, into the learner the document names.
Optional target: { Optional forgetDeletes every learned row and memory trace of a user in the agent's namespace that the layer reaches. With
memory.store, the agent's memory is opened for the call when no session holds it, so the deletion reaches the
store after agent.close() too.
What a host reads of an agent's learning layer, and the operator's calls on it (Agent.learning), with the agent's key (
learning.agentKey, else a named agent's persona id) and namespace (learning.namespace) filled in. A call that names a user acts on that user's learner of this agent, and rejects, naminglearning.agentKey, on an agent with no key. Every call runs in every mode, a freeze included.gmi()sets every member.