2026 03 09 Vitto Memory Job Mastery Design

Documentation for Plans

Vitto Memory Enhancements + Job Mastery Framework — Design Document

Session: 196 Date: 2026-03-09 Status: Approved

Overview

Two major systems built together:

User Dossier System — Per-user structured knowledge in vitto-users database, replacing 60-read brute-force with single-doc pre-compiled briefs
Job Mastery Framework — Two-layer (global + per-user) composable skill system where Vitto tracks, learns, and self-optimizes at specific job types

Architecture

Database: vitto-users (new, same cluster as vitto-mind)

Collections:

dossiers — One doc per user (identity, personality, expertise, workPatterns, spaces, relationships, timeline, jobMastery, brief)
job-playbooks — Global playbook definitions (composable skill trees, strategies, graduation criteria)
memories — Migrated from vitto-mind.vitto-memories (future phase)
knowledge-graphs — Migrated from vitto-mind.knowledge-edges/entities (future phase)

Job Mastery: Two-Layer Design

Layer 1 — Global Playbooks (job-playbooks collection):

Composable sub-skills (each with embedding for semantic retrieval)
Self-optimizing strategies with version history
5-tier graduation criteria (intern → junior → mid → senior → principal)
Origin tracking: seeded vs emergent
Progressive disclosure: brief (~50 tokens) → strategy (~500 tokens) → deep context (~2K tokens)

Layer 2 — Per-User Overlay (embedded in dossier.jobMastery):

Per sub-skill mastery scores
Personal strategy refinements
Autonomy level (derived from tier)
Optimization log (audit trail of self-improvements)
Level history with graduation timestamps

Self-Optimization Loop

After 3+ failures at same sub-skill:

Cheap LLM analyzes failure patterns against current strategy
Proposes revised strategy → stored as new version
A/B tested on next executions
Better? Promote. Worse? Rollback + flag for human review.

Progressive Disclosure (context injection)

Always: playbook.brief (~50 tokens per active job)
On task start: playbook.activeStrategy (~500 tokens)
Per phase: skill.detailedInstructions (~200 tokens, loaded per-phase)
On failure: skill.pitfalls + optimizationLog (~500 tokens)

Backend Implementation (backend-vitto-api)

New Files

File	Purpose
`app/users/schemas/user-dossier.schema.ts`	Dossier schema (identity, personality, expertise, workPatterns, spaces, relationships, timeline, jobMastery, brief)
`app/users/schemas/job-playbook.schema.ts`	Global playbook schema (skills, strategies, graduation criteria)
`app/users/user-dossier.service.ts`	Dossier CRUD + brief compilation + hydration from gograb-api
`app/users/job-mastery.service.ts`	Mastery tracking, tier promotion, self-optimization loop
`app/users/users.module.ts`	NestJS module for vitto-users DB
`app/users/users.controller.ts`	REST endpoints for dossier + mastery

Modified Files

File	Change
`app/app.module.ts`	Add vitto-users DB connection
`app/engine/context-assembly.service.ts`	Wire dossier + mastery into prompt assembly
`app/mind/mind.service.ts`	Auto-populate dossier after executions
`libs/libs-api/backend-core/.../database-names.config.ts`	Add VITTO_USERS constant

Frontend Implementation

New Library: `@libs-ai/vitto-intelligence`

Location: libs/libs-ai/vitto-intelligence/

Components:

JobMasteryDashboardComponent — Overview of all job types + user mastery levels
JobMasteryDetailComponent — Deep dive: skills, scores, history, optimization log
PlaybookEditorComponent — Edit playbook (strategies, skills, criteria). Role-gated: roleLevel >= 90
PlaybookCreatorComponent — Create new job type with Vitto-assisted pre-fill
UserDossierViewComponent — User's dossier (identity, expertise, timeline, mastery)
ExecutiveInstructionsComponent — Manage executive directives. Role-gated: roleLevel >= 90

Services:

VittoIntelligenceApiService — HTTP client to backend-vitto-api /users/* endpoints
MasteryProgressionService — Client-side mastery state + progression tracking

New Routes (under /vitto/)

Route	Component	Access
`/vitto/mastery`	JobMasteryDashboardComponent	All authenticated
`/vitto/mastery/:jobType`	JobMasteryDetailComponent	All authenticated
`/vitto/mastery/:jobType/edit`	PlaybookEditorComponent	roleLevel >= 90
`/vitto/mastery/new`	PlaybookCreatorComponent	roleLevel >= 90
`/vitto/dossier`	UserDossierViewComponent	All authenticated (own)
`/vitto/dossier/:userId`	UserDossierViewComponent	roleLevel >= 90
`/vitto/instructions`	ExecutiveInstructionsComponent	roleLevel >= 90

New Chat Plugin: `mastery.plugin.ts`

Injects mastery context into chat based on current task type
Shows skill progression badges in chat
Enables "teach Vitto" flow — user can refine strategies via chat
Pre-fills PlaybookCreator form when user describes a new job type in chat

Modified: VittoUnifiedChildRoutes

Add new routes to existing unified shell navigation.

Seeded Job Type: code-platform-generation

First playbook, seeded with sub-skills:

analyze-space-profile — Understand space context, modules, tech stack
select-template — Choose appropriate generation template
plan-architecture — Design module structure before generation
generate-backend — Backend code generation (NestJS, MongoDB)
generate-frontend — Frontend code generation (Angular, standalone)
verify-build — Build verification, lint, type check
validate-output — Quality validation against success criteria

Key Constraints

gograb-api remains source of truth for platform data
All reads from gograb-api via HTTP (BackendApiClientService)
@Optional() on all new service injections for backward compatibility
Build must pass: npx nx build backend-vitto-api
Follow existing NestJS patterns (mind/ directory as reference)
Angular: standalone, OnPush, signals, inject(), separate files
RxJS only — no Promises, no setTimeout