@adonis-agora/agent/dist/src/spi/roles-policy.d.ts

Governed, durable-backed AI agent (chat + tool-calling + governance) for AdonisJS — part of the Agora ecosystem.

import type { Actor, ToolSpec } from '../types.js';
/**
 * Decides whether an actor may invoke a tool. The default impl checks the actor's role
 * against `spec.roles` (defaulting to an ADMIN-only set). Apps can plug `nestjs-authz`
 * or any custom gate here.
 */
export interface RolesPolicy {
    /** May return a promise — an authz Gate (`gate.forUser(actor).allows(...)`) is async. */
    can(actor: Actor, tool: ToolSpec): boolean | Promise<boolean>;
    /**
     * Whether to OFFER the tool to the model (list it), when that differs from {@link can}. Omitted →
     * `can`. Invoking always re-checks `can`, so offering a tool an actor cannot run only lets the
     * model ask for it — `personalAgentGate` uses this so a personal agent's call to a tool behind a
     * scope it was not delegated becomes a permission request instead of a silence.
     */
    canOffer?(actor: Actor, tool: ToolSpec): boolean | Promise<boolean>;
}
//# sourceMappingURL=roles-policy.d.ts.map