Equipment maintenance · Knowledge & decisionsRecorded 30 September 2026
From a temperature reading to a reviewable recommendation.
These are synthetic software examples, not customer deployments or live equipment tests. The agent reads two published sample rules; it does not connect to a machine, send a notification or create a work order.
Actual product capture. Original prompts and answers are in Chinese; English descriptions below explain the same cases.
CASE A · COMPLETE INPUT
Recognize the rule. Show the source.
Input
DEMO-MOTOR-01 · 94°C · 45s
Knowledge consulted
At 90°C or above for at least 30 seconds, flag for human review. Include the original reading and the rule used.
Observed result
Tubun identifies the review condition and cites the source entries. It separates suggested checks from actions that have actually happened.
Create a knowledge space with equipment and procedure dimensions, add the two sample entries, assign the space to Plan, and use the prompts in the run records. Bring your own supported model connection. Results may vary with the model.
For HTTP, MQTT, RabbitMQ or Laya decision routing, we agree a real integration scope and demonstrate it with your team. These recorded cases validate knowledge-based reasoning only.
