Skip to main content

HAL (Human Assist Layer) — Master Manual

Overview​

HAL (Human Assist Layer) is the human interaction, oversight, and responsibility-routing layer within RABS.

It connects humans to live system state, ensures accountability, and enables real-time intervention in an otherwise autonomous system.

HAL is not a UI. It is a control plane for human involvement.

Core Purpose​

HAL ensures:

  • Humans remain informed
  • Responsibility is always assigned
  • Decisions are visible and explainable
  • Intervention is possible at the right moment

System Position​

[ Brainframe / Agents ] ↓ HAL ↓ [ Humans / Operators ]

Core Capabilities​

Event Awareness​

  • Real-time system events
  • Alerts and anomalies

Approval & Intervention​

  • Hard approvals
  • Soft approvals
  • Overrides

Responsibility Routing​

  • Determines who should act
  • Uses active sessions, schedules, escalation

Reasoning Visibility​

  • Human-readable Brainframe output

Audit & Traceability​

  • Full history of actions and decisions

HAL/LINK — Human Entry System​

HAL/LINK allows a human to “plug into” the system.

Entry Modes​

  • Full HAL Session (dashboard)
  • Token HAL View (link-based)
  • Observer Mode

Role & Shift Model​

Roles​

  • On-Call Operator
  • Billing Supervisor
  • Transport Coordinator

Resolution Logic​

IF active session exists → route to operator ELSE IF scheduled → notify + prompt ELSE → escalate

Concept​

Short-lived, user-specific dashboards delivered via link.

Properties​

  • No login required
  • Scoped access
  • Real-time updates
  • Expiring tokens

Token Structure​

{ "token": "abc123", "user_id": "staff_001", "role": "oncall_operator", "variant": "incident_review", "scopes": ["approve"], "expires_at": "...", "requires_pin": true }

HAL PIN​

  • Step-up authentication
  • Required for sensitive actions
  • Protects token access

Variant Dashboard System​

Variants define layout + purpose.

Examples:

  • incident_review
  • approval_action
  • roster_shift
  • billing_exception
  • profile_update

Live Update System​

Initial Load​

Token resolves → HTML renders

Real-Time Updates​

Server-Sent Events (SSE)

Flow​

HAL → emit → SSE → update UI

Brainframe Integration​

HAL exposes:

  • Decision
  • Reason
  • Context

Allows:

  • Approval
  • Rejection
  • Modification

Session Model​

{ "session_id": "...", "user_id": "...", "role": "...", "status": "active" }

Event Routing​

HAL determines who receives events based on:

  • Active operator
  • Schedule
  • Fallback chain

Logging & Observability​

Tracks:

  • Event routing
  • Actions taken
  • Token usage
  • Operator activity

Architecture Components​

  • Scheduler
  • Event Bus
  • HAL Router
  • Token Service
  • SSE Layer
  • UI Layer

API Surface (High-Level)​

  • /hal/link/:token
  • /hal/verify-pin
  • /hal/resolve-role
  • /hal/route-event
  • /hal/emit

Design Principles​

HAL must be:

  • Real-time
  • Human-first
  • Deterministic
  • Explainable
  • Secure

Strategic Definition​

HAL is the human runtime layer of RABS.

Brainframe decides. HAL exposes and routes. Humans act.

Future Expansion​

  • Voice-based HAL sessions
  • Predictive alerts
  • Simulation/replay
  • Adaptive routing

Summary​

HAL enables humans to dynamically enter, understand, and influence system behaviour in real time, without requiring full system access.

It is the bridge between autonomy and accountability.