๐Ÿงช Brain Smoke Test
Synthetic end-to-end pipeline verification ยท proves the methodology works in seconds, not 24 hours
Test suite status
โ€” READY โ€”
Click "Run smoke test" to execute all 12 assertions against a synthetic feature vector. The test exercises every stage of the self-learning pipeline: ensemble prediction, training, regime tracking, reward shaping, feature attribution, active learning.
๐Ÿ“‹ Test log
Awaiting test execution...
๐Ÿ”ฌ What this proves
The 12 assertions cover every link in the chain:
1-3. Modules loaded and globals available (FeatureExtractor, ModelStore, MultiHorizon)
4. FeatureExtractor produces a 22-element feature vector for a synthetic finding
5. Regime detector returns a valid regime from current live data
6. MultiHorizon.predictEnsemble returns prob + per-horizon breakdown + weights
7. MultiHorizon.trainHorizon runs SGD with reward-shaped weight
8. MultiHorizon.recordOutcome updates the (regime, horizon) accuracy log
9. Feature alpha map gets updated after training
10. Main model trains and saves correctly
11. Continuous learner can run a manual cycle without errors
12. After simulating 5 wins + 5 losses, model weights have diverged from zero

If all 12 pass, the methodology works end-to-end. The brain WILL learn from real captures, no waiting required for proof.