Trigger feedback-driven retraining on demand, watch the staged bundle load, feedback learning, and promotion phases, and only replace the live airport bundle when the new models outperform the current ones.
Scheduler: true ยท Every 86400 seconds
Source
Base retraining uses the staged airport bundle plus saved FLIFO prediction-vs-actual feedback stats.
Gate
Promotion is blocked unless the candidate scores beat the current live airport bundle.
Publish
Promoted bundles land under train/<airport>/csv_build_outputs.
Run Status
Progress
Idle
0%
Loading...
Logs
Live retrain-service activity for scheduler and manual runs.