Supervise ROV commands using multi-agent OpenAI planners and Telegram

## ROV multi-agent intelligent supervisory control ## How It Works This workflow is for marine robotics teams and researchers who run remotely operated vehicles and need an AI supervisor that never bypasses safety. ROV operators handle constant telemetry, tight reaction times and the risk of unsafe commands. This template receives live telemetry, checks hard safety limits first, then lets a planner agent propose the next action using mission history, navigation, perception and hazard specialists. A second, independent validator agent reviews the plan before anything is sent. The result is routed by authority level: low risk commands are auto approved, medium risk commands wait for human approval on Telegram, and critical situations trigger a safe command and an alert. Every decision is logged and sent to the ROV control API. Two side flows add a 30 minute mission debrief and a chat assistant for operators. Expected outcome: faster, auditable and safer ROV supervision. ## Setup Steps 1. Add OpenAI and Telegram credentials. 2. Create the rov_mission_log data table and select it in all Data Table nodes. 3. Replace every Telegram chat ID placeholder. 4. Set the ROV control URL in Send Command to ROV Control. 5. POST a test payload to the rov-telemetry webhook and check each route. ## Customization Swap models per agent, change authority rules in Decision Package, add Slack or email alerts, connect another vehicle API, adjust safety limits in the telemetry body, or change the debrief interval.