Behavioral Learning

Archived: describes an older release that is no longer supported.

Behavioral Learning

SuperLocalMemory learns from what happens after memories are recalled — tracking which memories lead to successful outcomes and surfacing them more often.

How It Works

1. You recall a memory — search finds a relevant memory for your task
2. You take action — use the information, write code, make decisions
3. Outcome is recorded — explicitly via report_outcome or implicitly from your behavior
4. Patterns emerge — over time, the system learns which memories are most useful

Reporting Outcomes

Explicit Reporting

report_outcome memory_id="abc123" outcome="success" context="Used this API pattern, it worked"

Outcome values: success, failure, partial

Implicit Inference

The system also infers outcomes from your behavior:

• Recalled a memory and continued working? Likely success
• Recalled a memory then immediately searched again? Likely failure
• Asked a follow-up question? Likely partial

Viewing Patterns

get_behavioral_patterns

Shows learned patterns like:

• "API documentation memories have 85% success rate in this project"
• "Architecture decisions from Project A transfer well to Project B"

Cross-Project Transfer

Behavioral patterns learned in one project can improve recommendations in similar projects. This happens automatically and preserves privacy — only pattern metadata transfers, never raw memory content.

Privacy Guarantees

• All learning happens 100% locally
• No data leaves your machine
• No LLM inference calls — pure statistical pattern recognition
• Learning data is stored in learning.db (separate from memories, GDPR-erasable)

---

See also: Memory-Lifecycle · Enterprise-Compliance · Learning-System