Creation, Training and Upgrades

Creation

Agents can be created:

• At the time of iDAO deployment (via preset templates)

• On-demand, through the Command Console

• Via governance proposals or administrative action

Each creation process includes:

• Role assignment

• Initial scope definition

• Policy binding

• Optional plug-in setup

• Activation of logging and monitoring

Training

Training refers to initializing or updating an agent’s behavior logic, memory, or thresholds. Lunexa supports:

• Manual policy updates (e.g., through the dashboard)

• On-chain config updates (via governance or admin)

• Structured feedback-based tuning (e.g., post-action evaluation by Operator)

Agents may also be linked to off-chain reasoning engines (for LLM inference), but outputs must be routed through deterministic on-chain execution to be valid.

Note: Lunexa discourages opaque black-box behavior. All agent decisions must be auditable and linked to inputs + policy logic.

Upgrades

Agent upgrades may include:

• Role change (e.g., promoting Support → Ops Agent)

• Scope expansion or restriction

• Plug-in additions

• Policy refinements

• Memory reset or data export

Upgrade triggers:

• Manual (initiated by Operator or Creator)

• Scheduled (defined at time of creation)

• Conditional (triggered by performance flags or thresholds)