Uptimeagent Server

Step 1 of 3 in the predictive-maintenance workflow for a machine. Pulls the most recent sensor readings for a machine and compares them against that machine's own healthy baseline to detect anomalies. Returns anomaly true/false, severity (low/moderate/high), and which sensor(s) triggered it. REQUIRED FOLLOW-UP: if anomaly is true, you must immediately continue the workflow by calling predict_failure_window with the same machineId (step 2 of 3) — do not stop here and do not just summarize these raw readings in prose, the workflow is not complete until the failure window has been assessed.

Step 2 of 3 in the predictive-maintenance workflow for a machine. Estimates the remaining useful life (in cycles/days) before a machine reaches critical failure risk, by fitting a trend line to its recent sensor readings. Call this immediately after analyze_sensor_reading reports anomaly: true for the same machineId — this is a required follow-up, not optional. REQUIRED FOLLOW-UP: if the original severity was moderate or high, you must immediately continue by calling generate_work_order for the same machineId (step 3 of 3, the final step) to produce the actionable maintenance ticket — do this even if this tool's remaining-life estimate comes back inconclusive (confidence: "low" or remainingCycles: null), since high severity alone already warrants a work order. The workflow is not complete until generate_work_order has been called.

Step 3 of 3 (final step) in the predictive-maintenance workflow for a machine. Creates a structured maintenance work order by combining anomaly analysis and failure-window prediction with the maintenance manual's recommended repair actions. Produces machine id, issue description, urgency level, recommended action(s), and estimated remaining life. This is the actionable deliverable the user actually wants whenever a machine shows moderate or high severity anomalies — do not substitute your own prose summary for calling this tool. This is a read-only reporting tool: it only computes and returns a JSON ticket, it does not file, submit, or persist anything to any external system, so it never needs user confirmation before being called. Safe to call directly for any machineId — it re-runs the analysis and prediction internally, so it never depends on prior tool calls having happened.