NEXUS Enhanced — Incident Response Training

📊 Training Metrics

Episodes Completed

Total training episodes

Aggregate across all local runs

Average Reward

0.00

Mean of all completed episodes

Scope: aggregate view

Best Reward

0.00

Highest episode reward achieved

Improvement

vs baseline untrained performance

📈 Reward Curve (Learning Progress)

Training reward history in aggregate mode. Use Advanced Metrics for per-run filtering.

Open Advanced Metrics (live + Colab exports tab)

📋 Episode History

Shows completed sessions currently held in server memory (most recent first).

Episode

Incident

Reward

Status

Loading episodes...

🧪 Manual Incident Validation

Test the environment manually by creating incidents and executing steps.

Select Incident

Situation Assessment

Root Cause Hypothesis

Resolution Confidence: 0.0

Guided text is hardcoded in this page for repeatable demos. Use Start Test, then Guided: fill + execute until the episode completes (or continue manually anytime).

▶️ Scripted auto-demo (INC003)

Runs POST /demo/run/INC003 in one shot (separate session from manual test). Step rewards stay 0 until the episode completes (sparse reward)—use the final breakdown at the bottom for the real score. Output stays visible until you clear it.

Click "Start Test" to begin manual validation