Operator Dashboard

The ROBIN Dashboard is a React-based operator interface at http://localhost:5174. It connects to the Alert Engine API and provides real-time monitoring, robot control, AI model management, and deviation alerting - all configurable per domain profile.

Live Ops

The Live Ops tab is the main cockpit.

Measurement KPIs

The top row shows the latest values for the five core telemetry fields (height, width, speed, current, voltage). Labels and units adapt automatically to the active profile.

Robot Control

The left panel provides per-robot controls:

• Start / Pause / Resume / Abort - manage robot state

• Parameter Freeze - lock current parameters so the AI cannot override them

• Active Model - which AI checkpoint is loaded for this position

• Trust Score - runtime confidence with OK / Warning / Stop gate

3D Visualisation

The centre panel embeds a Viser viewer (when a Viser server is running) with toggleable layers:

• Robot model

• Tool path

• Workpiece

• Profile segments

• Coordinate frames

Camera presets (Isometric, Top, Front, Side) and a timeline scrubber are available below the viewport.

Deviation Monitor

The right panel compares measured geometry against the AI prediction (parameter-driven mode) or AI-guided geometry reference (geometry-driven mode):

• Mode - shows the active operation mode

• Tolerance - configurable percentage threshold

• Latest Measurement - last received height and width

• Source - shows whether the expected geometry comes from AI prediction or target fallback

• Recommended actions when a deviation fires: manual adjust, request new AI recommendation, add data for fine-tuning, or start a new DOE

Process Controls

Below the deviation monitor, the Process Controls panel lets the operator:

• Switch between parameter-driven and geometry-driven modes

• Set tolerance percentage

• Enter process parameters (speed, current, voltage) or target geometry (height, width) depending on mode

• In geometry-driven mode, clicking Apply Settings stores target geometry and fetches AI-suggested parameters for the process

Telemetry Charts

The bottom section shows live time-series powered by Recharts:

• Select metric: height, width, speed, current, or voltage

• Data source badge shows whether data is coming from Mintaka stored data or fallback source

• Poll cadence is 1-2 seconds depending on session mode

• X-axis uses elapsed time from measurement timestamps (stored temporal data)

• Live toggle for auto-scrolling and Freeze Charts to inspect windows

Models & Trust

The Models & Trust tab manages AI model checkpoints and trust settings.

See AI Models & Trust for a detailed walkthrough.

Other Tabs

Tab	Purpose
Robots	Per-robot detail cards with individual telemetry and controls
History / Reports	Download CSV with all raw telemetry rows and persisted warning events (including raw JSON payload per row)
Settings	Dark mode, trust thresholds, chart freeze, connection status

History CSV Workflow

1. Open History / Reports.

2. Select the process ID in the dropdown.

3. Click Refresh to pull current stored telemetry + warnings.

4. Click Download CSV.

5. Verify CSV content: * record_type=measurement rows are raw telemetry samples. * record_type=warning rows are persisted deviation alerts. * raw_payload_json stores the exact backend payload for each row.

Top Bar

The top bar (visible on every tab) provides:

• Process selector - switch between active processes

• Session mode - Active Run / Demo Mode * Active Run: process-control actions write to backend * Demo Mode: process-control actions are simulated locally (no backend write)

• Connection indicators - Backend API, Orion-LD, Mintaka from live GET /health probes (green / yellow / red)

• Service latency chips - Orion and Mintaka round-trip times in milliseconds

• Backend uptime - service uptime in hours from the Alert Engine health API

• Dark mode toggle