Continual Learning
Status: demo-backed / active
PodMan's continual-learning track owns team memory: what the system learns about files, collisions, interventions, outcomes, and future routing for a pod.
Files
| File | Purpose |
|---|---|
spec.md |
Data model and observe/store/predict/outcome/adapt loop |
policy.md |
What PodMan may and may not remember |
prompt.md |
Memory-agent prompt for outcome-backed learning |
plan.md |
Demo build order and acceptance criteria |
What Is Implemented Now
- MongoDB-backed
observations,collisions,interventions,outcomes,engineer_states, andteam_modelrecords. - Exact signature recall for prior accepted and dismissed outcomes.
- Outcome writes through
POST /api/outcome. - Team memory graph edges from accepted real outcomes.
- No raw screenshots or recordings are stored.
What Is Intentionally Cut
- Autonomous model training.
- Broad cross-pod generalization.
- Raw screen capture retention.
- Vector recall as a dependency for the demo proof.
Demo Proof Path
Observe screen/git state -> detect collision -> send intervention -> accept or dismiss outcome -> recall similar event -> show changed graph or changed behavior.