# DEB-379: Training and Learning Lifecycle

> **Jira:** [DEB-379](https://divami.atlassian.net/browse/DEB-379)  
> **Source Report:** [20260315.md](../../reports/jira/20260315.md)

- **Status:** Start
- **Assignee:** Rajashekar Gummadidala
- **Priority:** Medium
- **Created:** 2026-03-13T16:12:41.823+0530
- **Updated:** 2026-03-13T16:24:32.784+0530

**Description:**

- RAG DB should be updated with the new training lessons
- On next similar query Check similarity against stored lessons 

If high match - Use learned pattern instead of generating new query



Receive training from RT agent
Validate trainer persona and permissions
Identify the target skill or pattern to update
Store lesson in RAG DB (vector + metadata)
Update QA Cache if applicable
Acknowledge back to RT agent with status
"On next similar query Check similarity against stored lessons 
If high match - Use learned pattern instead of generating new query"
Log usage of learned lesson

**Architecture Diagram:**

```mermaid
sequenceDiagram
    participant RT as RT Decision Agent
    participant DSA as Data Source Agent
    participant RBAC as RBAC Controller
    participant RAGDB as RAG DB (Vector)
    participant Cache as QA Cache

    RT->>DSA: Send Training Lesson
    DSA->>RBAC: Validate Trainer Persona & Permissions
    RBAC-->>DSA: Authorized
    DSA->>DSA: Identify Target Skill / Pattern
    DSA->>RAGDB: Store Lesson (vector + metadata)
    RAGDB-->>DSA: Stored
    DSA->>Cache: Update QA Cache (if applicable)
    Cache-->>DSA: Updated
    DSA-->>RT: Acknowledge (Success/Failure)
    note over DSA,RAGDB: On next similar query: check similarity
    DSA->>RAGDB: Similarity Check against stored lessons
    alt High similarity match
        RAGDB-->>DSA: Return learned pattern
        DSA->>DSA: Use learned pattern (skip new query gen)
    else No match
        DSA->>DSA: Generate new query
    end
    DSA->>DSA: Log usage of learned lesson
```