oryon-framework 1.0.0

package/.knowledge/dossiers/DOSSIER-AGENTES-IA.md

@@ -0,0 +1,122 @@
+# DOSSIER-AGENTES-IA
+## Agentes de IA: Arquitetura, Criação e Aplicações Práticas
+
+> **Versão:** 1.0.0
+> **Criado:** 2026-03-26
+> **Fontes:** EP038
+> **Aplicação:** Criação de agentes de IA especializados sem programação
+
+---
+
+## VISÃO GERAL
+
+Este dossier consolida o conhecimento sobre **criação e uso de agentes de IA especializados**. Foco em abordagem no-code: como criar agentes que agem como especialistas em domínios específicos, coletam dados via formulários automatizados, e analisam resultados com contexto do negócio.
+
+---
+
+## 1. ARQUITETURA DE AGENTES
+
+### 1.1 Dual-Role Architecture
+
+**FW-EP038-001: Arquitetura Pesquisador + SuperAgente**
+```
+PAPEL 1: PESQUISADOR (humano)
+├── Faz perguntas e define escopo
+├── Valida outputs
+└── Direciona o agente para próximos passos
+
+PAPEL 2: SUPERAGENTE (ChatGPT configurado)
+├── Responde como especialista nos dados do negócio
+├── Acessa contexto injetado (documentos, dados)
+├── Gera estruturas automatizadas (formulários, relatórios)
+└── Analisa dados coletados com lente especialista
+
+SEPARAÇÃO CRÍTICA:
+└── Misturar papéis = degradação de qualidade
+└── Pesquisador pergunta, SuperAgente responde
+```
+
+### 1.2 Geração Automática de Google Forms
+
+**FW-EP038-002:**
+```
+INPUT:
+├── Objetivo da pesquisa
+├── Público-alvo
+└── Informações que precisa coletar
+
+OUTPUT DO SUPERAGENTE:
+├── Estrutura completa do Google Form
+├── Perguntas ordenadas por prioridade
+├── Tipos de resposta adequados
+└── Lógica de fluxo condicional
+
+CICLO COMPLETO:
+1. ChatGPT gera estrutura do formulário
+2. Humano implementa no Google Forms
+3. Respostas coletadas
+4. ChatGPT analisa respostas como IA especialista
+5. Insights acionáveis gerados
+```
+
+---
+
+## 2. PRINCÍPIOS DE CRIAÇÃO DE AGENTES
+
+### Context Quality = Agent Quality
+
+O agente é tão bom quanto o contexto que recebe:
+
+| Contexto Injetado | Qualidade das Respostas |
+|-----------------|------------------------|
+| Genérico ("empresa de marketing") | Respostas genéricas e previsíveis |
+| Específico (docs, dados, histórico, ICP) | Insights acionáveis e específicos |
+| Incompleto (informações parciais) | Respostas com lacunas e premissas erradas |
+
+### Especialização vs Generalização
+
+```
+AGENTE GENERALISTA:
+├── Responde sobre tudo
+├── Superficial em tudo
+└── Pouco valor prático
+
+AGENTE ESPECIALISTA:
+├── Responde profundamente sobre um domínio
+├── Contexto rico injetado
+└── Alto valor prático
+
+REGRA: 1 agente por domínio > 1 agente para tudo
+```
+
+---
+
+## 3. HEURÍSTICAS
+
+**HEUR-EP038-001:** Sem programação não significa sem estratégia — a qualidade do agente depende da qualidade do contexto injetado.
+
+**HEUR-EP038-002:** Google Forms gerado por IA tem melhor estrutura — ChatGPT considera viés, sequência lógica e tipos de resposta.
+
+**HEUR-EP038-003:** Dois papéis distintos evitam confusão — Pesquisador pergunta, SuperAgente responde. Misturar degrada qualidade.
+
+**HEUR-EP038-004:** Especialização por domínio é mais poderosa que generalização — um agente especialista em atendimento > um agente que "faz tudo".
+
+---
+
+## 4. CASOS DE USO
+
+| Domínio | Agente Especialista | Output |
+|--------|---------------------|--------|
+| Pesquisa de mercado | SuperAgente pesquisador | Google Form + análise de respostas |
+| Atendimento | Bot especialista em produto | Respostas sem necessidade humana |
+| RH | Agente de triagem | Análise de candidatos + scoring |
+| Marketing | Agente de ICP | Personas + copy baseado em dados |
+| Vendas | Agente de objeções | Scripts personalizados por perfil |
+
+---
+
+## CONEXÕES
+
+- **Agentes enriquecidos:** CMO, CRO, CLOSER
+- **Sources:** EP038
+- **Relacionado:** DOSSIER-WHATSAPP-CHATBOT.md, DOSSIER-CHATGPT-FERRAMENTAS.md

package/.knowledge/dossiers/DOSSIER-AI-OPERATING-SYSTEMS.md

@@ -0,0 +1,209 @@
+# DOSSIER: AI OPERATING SYSTEMS (AIOS)
+
+> **Version:** 1.0.0
+> **Created:** 2026-03-12
+> **Last Updated:** 2026-03-12
+> **Primary Source:** Liam Ottley (LO-001, LO-002, LO-003)
+> **Contributing Sources:** (Pending — cross-reference as more sources are processed)
+
+---
+
+## DEFINITION
+
+An AI Operating System (AIOS) is a comprehensive AI-powered layer that wraps around an
+existing business model to automate operations, augment decision-making, and free founder
+bandwidth. It is NOT a chatbot, NOT a single automation tool, and NOT a workflow builder.
+It is an integrated system that understands the business fully (context), connects to all
+data sources (data), synthesizes information (intelligence), takes action (automate), and
+creates new capabilities (build).
+
+The term was coined and popularized by **Liam Ottley** of Morningside AI, who describes
+AIOS as "the third major shift" in business history:
+
+```
+Industrial Revolution  →  changed how we MAKE
+Internet               →  changed how we REACH
+AIOS                   →  changes how we RUN
+```
+
+---
+
+## THE 5-LAYER ARCHITECTURE
+
+The AIOS is built in 5 concentric layers. Each layer wraps around the business model and
+compounds on the previous layer. The order is critical — skipping layers creates fragile systems.
+
+```
+                    ┌─────────────────────────────────────────┐
+                    │            LAYER 5: BUILD               │
+                    │     Create new capabilities              │
+                    │  ┌─────────────────────────────────┐    │
+                    │  │       LAYER 4: AUTOMATE          │    │
+                    │  │     Act on intelligence           │    │
+                    │  │  ┌─────────────────────────┐     │    │
+                    │  │  │   LAYER 3: INTELLIGENCE  │     │    │
+                    │  │  │   Make sense of data      │     │    │
+                    │  │  │  ┌───────────────────┐   │     │    │
+                    │  │  │  │  LAYER 2: DATA     │   │     │    │
+                    │  │  │  │  Connect sources   │   │     │    │
+                    │  │  │  │  ┌─────────────┐  │   │     │    │
+                    │  │  │  │  │ L1: CONTEXT  │  │   │     │    │
+                    │  │  │  │  │  Understand  │  │   │     │    │
+                    │  │  │  │  │  the biz     │  │   │     │    │
+                    │  │  │  │  └─────────────┘  │   │     │    │
+                    │  │  │  └───────────────────┘   │     │    │
+                    │  │  └─────────────────────────┘     │    │
+                    │  └─────────────────────────────────┘    │
+                    └─────────────────────────────────────────┘
+```
+
+### Layer Details
+
+| Layer | Name | Purpose | Implementation | Failure Mode |
+|-------|------|---------|---------------|-------------|
+| L1 | **Context** | System understands the business | CLAUDE.md, context files, SOPs, team docs | AI gives generic, wrong answers |
+| L2 | **Data** | System connects to reality | MCP servers, APIs, database connections | AI is blind to what is happening |
+| L3 | **Intelligence** | System synthesizes understanding | RAG pipeline, knowledge graph, analytics | AI has data but no insight |
+| L4 | **Automate** | System takes action | Skills, cron jobs, webhooks, agent actions | AI understands but cannot act |
+| L5 | **Build** | System creates new capabilities | Epic cycles, development pipeline, self-improvement | System is static, cannot evolve |
+
+---
+
+## THREE KPIs OF AIOS
+
+An AIOS is measured by exactly three KPIs. If all three are improving month over month,
+the system is working. If any is flat or declining, there is a specific layer failure.
+
+| KPI | What It Measures | Target | Diagnostic |
+|-----|-----------------|--------|------------|
+| **Away-From-Desk Autonomy** | Can the business run without the founder physically present? | Full day from phone | If NO: Layer 4 (Automate) or Layer 6 (Mobile Bridge) has gaps |
+| **Task Automation %** | What fraction of operational tasks are handled by AI? | 60-70% within 30 days | If LOW: Layer 1 (Context) incomplete or Task Audit not done |
+| **Revenue Per Employee** | How much output does each human produce with AI leverage? | Increasing month/month | If FLAT: AI workforce not deployed effectively |
+
+---
+
+## CLAUDE CODE AIOS STACK
+
+Liam's specific technology stack for AIOS implementation:
+
+```
+┌──────────────────────────────────────────────────────────────────────────────┐
+│                        CLAUDE CODE AIOS STACK                                │
+├──────────────────────────────────────────────────────────────────────────────┤
+│                                                                              │
+│  ENGINE:   Claude Code                                                       │
+│            AI reasoning + execution layer                                    │
+│            Complex multi-step tasks, code gen, file management               │
+│                                                                              │
+│  CONTEXT:  CLAUDE.md + structured context files                              │
+│            Business rules, SOPs, team info, product details                  │
+│            Loaded at session start, referenced throughout                    │
+│                                                                              │
+│  DATA:     MCP servers                                                       │
+│            Typed interfaces to external systems                              │
+│            CRM, email, calendar, analytics, databases                        │
+│                                                                              │
+│  SKILLS:   Claude Code skills                                                │
+│            Reusable capabilities triggered by keywords/commands              │
+│            /daily-brief, /process-inbox, /source-sync                        │
+│                                                                              │
+│  CRON:     Scheduled tasks                                                   │
+│            Automated recurring actions on schedule                           │
+│            Daily briefs, weekly reports, inbox processing                    │
+│                                                                              │
+│  MOBILE:   Telegram bot                                                      │
+│            Mobile command interface (away-from-desk bridge)                  │
+│            Send commands, receive updates, approve decisions                 │
+│                                                                              │
+└──────────────────────────────────────────────────────────────────────────────┘
+```
+
+---
+
+## MAPPING TO MEGA BRAIN
+
+Mega Brain has independently built an AIOS that maps almost 1:1 to Liam's architecture.
+This table documents the equivalences:
+
+| Liam's Layer | Liam's Implementation | Mega Brain Equivalent | Status |
+|-------------|----------------------|----------------------|--------|
+| **L1 Context** | CLAUDE.md + context files | `CLAUDE.md` + `.claude/rules/` (30 rules, 6 rule groups) + `system/` (JARVIS identity) | BUILT |
+| **L2 Data** | MCP servers | `n8n-mcp` (automation) + `read-ai` (meetings) + `mega-brain-knowledge` (RAG) + Fireflies integration | BUILT |
+| **L3 Intelligence** | RAG + knowledge graph | `core/intelligence/rag/` (BM25 + hybrid + graph) + `core/intelligence/pipeline/` (MCE) + knowledge graph (1,302 entities) | BUILT |
+| **L4 Automate** | Skills + cron | `.claude/skills/` (20+ skills) + `.claude/hooks/` (20+ hooks) + LaunchAgents (Fireflies cron) | BUILT |
+| **L5 Build** | Epic cycles | GSD workflow (`.planning/`) + development pipeline + epic cycles + this very session | BUILT |
+| **Mobile** | Telegram bot | NOT YET BUILT — identified gap | PENDING |
+
+### Identified Gaps from Liam's Framework
+
+| Gap | Liam's Approach | Mega Brain Status | Priority |
+|-----|----------------|-------------------|----------|
+| **Mobile Bridge** | Telegram bot for away-from-desk command | Not implemented | MEDIUM |
+| **Task Audit Formalization** | Structured task audit matrix as first step | Informal (not documented) | LOW |
+| **Daily Brief Automation** | Automated morning brief as first AIOS milestone | `/jarvis-briefing` exists but not automated via cron | MEDIUM |
+| **AI Workforce Packaging** | Package AIOS components for sale ($3K-$10K+) | Not applicable (internal system) | N/A |
+
+---
+
+## AI WORKFORCE CONCEPT
+
+Beyond individual AIOS, Liam describes **AI Workforces** — digital teams that replace
+entire departments. This is the second-order application of AIOS principles.
+
+| Department | AI Team Roles | Coordinator |
+|-----------|--------------|-------------|
+| **Marketing** | Content creator, distributor, analytics, SEO, social media | Marketing coordinator agent |
+| **Sales** | Lead qualifier, outreach, follow-up, proposal | Sales coordinator agent |
+| **Operations** | Task router, scheduler, documentation, QA | Ops coordinator agent |
+| **Creative** | Design, copywriting, video editing, brand consistency | Creative coordinator agent |
+
+In Mega Brain terms, this maps to the **agents/cargo/** architecture where functional role
+agents (CLOSER, CRO, CFO, CMO) serve as the equivalent of Liam's department-level AI teams.
+
+---
+
+## KEY PREDICTIONS
+
+| Prediction | Source | Timeframe |
+|-----------|--------|-----------|
+| 15% of day-to-day work decisions made by agentic AI | Gartner (cited by Liam) | By 2028 |
+| First one-person billion-dollar company | Sam Altman (cited by Liam) | Near-term |
+| "Bloodbath" for small businesses that don't adopt AI workforces | Liam Ottley | Already happening |
+
+---
+
+## CROSS-REFERENCES
+
+| Theme | Related Dossiers | Connection |
+|-------|-----------------|-----------|
+| Automation | (pending DOSSIER-AUTOMATION) | AIOS Layer 4 = automation layer |
+| Sales Operations | DOSSIER-SALES (if exists) | AI workforce for sales department |
+| Founder Productivity | (pending) | Bandwidth inversion (80/20 → 20/80) |
+| Revenue Efficiency | (pending) | Revenue per employee as primary metric |
+
+---
+
+## SOURCE MATERIAL
+
+| ID | Title | Type | Key Contribution |
+|----|-------|------|-----------------|
+| LO-001 | The AI Operating System — What It Is and Why You Need One | YouTube | 5-layer architecture, AIOS definition, implementation sequence |
+| LO-002 | I Built an AI Workforce — Here's What Happened | YouTube | AI workforce concept, department replacement, $3K-$10K pricing |
+| LO-003 | How I Made $1M in 7 Days Using AI | YouTube | 7-day sprint methodology, proof of AIOS capability |
+
+---
+
+## METADATA
+
+| Field | Value |
+|-------|-------|
+| Theme Slug | `ai-operating-systems` |
+| DNA Path | `knowledge/external/dna/persons/liam-ottley/` |
+| Dossier Path | `knowledge/external/dossiers/themes/DOSSIER-AI-OPERATING-SYSTEMS.md` |
+| Primary Expert | Liam Ottley |
+| Contributing Experts | (none yet — update as more sources contribute) |
+| Total DNA Elements | 27 (from Liam Ottley) |
+
+---
+
+*Dossier compiled by Pickle Rick (The Resourceful Builder) -- 2026-03-12*