Agente CRM descobre → passa contexto → Agente Financeiro analisa

Agente CRM: "Cliente: Acme Corp, faturamento R$ 50M/ano" ✅
Agente Financeiro: "Quem é Acme Corp? 🤷" ❌

┌──────────────┐     ┌──────────────┐     ┌──────────────┐
│  AGENTE CRM  │     │  AGENTE FIN  │     │  AGENTE EXEC │
│  (escrita)   │     │  (leitura)   │     │  (consolida) │
└──────┬───────┘     └──────┬───────┘     └──────┬───────┘
       │                    │                    │
       │    HTTP/MCP        │    HTTP/MCP        │    HTTP/MCP
       ▼                    ▼                    ▼
┌──────────────────────────────────────────────────────┐
│              MCP SHARED MEMORY SERVER                 │
│   tools: write / read / search / upsert / flush      │
│   + links + export                                    │
└───────────────────────┬──────────────────────────────┘
                        │
                        ▼
              ┌─────────────────┐
              │     SQLite       │
              │  shared_memory   │
              │  shared_memory_  │
              │  links           │
              └─────────────────┘

# 1. Clone
git clone https://github.com/CidLucas/agent-shared-memory.git
cd agent-shared-memory

# 2. Instale dependências
pip install .

# 3. Inicialize o banco
python -m src.main --init

# 4. Inicie o servidor (Terminal 1)
python -m src.main

# 5. Rode o exemplo básico (Terminal 2)
python examples/basic_usage.py

# === AGENTE CRM: descobre e salva ===
await memory.write(
    entity_type="client",
    entity_name="tech_solutions",
    key="company_profile",
    value={
        "nome": "Tech Solutions Ltda",
        "faturamento_mensal": 850000,
        "funcionarios": 45,
        "segmento": "B2B",
    },
    source="memory_agent",
    ttl_tier="memory_agent_hi",  # 14 dias
)

# === AGENTE FINANCEIRO: lê e adiciona análise ===
context = await memory.read(entity_name="tech_solutions")
# → {"tech_solutions": {"company_profile": {...}}}

await memory.write(
    entity_type="client",
    entity_name="tech_solutions",
    key="analise_financeira",
    value={
        "receita_anual": 10200000,
        "margem_liquida": 0.24,
        "score_credito": "AA",
    },
    source="memory_agent",
    category="decision",
)

# === AGENTE EXECUTIVO: consolida tudo ===
from src.agent_integration import load_shared_memory_context

contexto = await load_shared_memory_context(
    agent_type="executivo",
    entity_names=["tech_solutions", "carlos_silva"],
    memory_client=memory,
)
# contexto → {"tech_solutions": {"company_profile": {...}, "analise_financeira": {...}}}

from langgraph.graph import StateGraph
from your_memory_client import SharedMemoryClient

memory = SharedMemoryClient()

class AgentState(TypedDict):
    messages: list
    shared_context: dict

async def crm_node(state: AgentState, config):
    # Salva resultados na shared memory
    await memory.write("client", "acme", "profile", {...})
    return state

async def financeiro_node(state: AgentState, config):
    # Carrega o que o CRM salvou
    context = await load_shared_memory_context(
        "financeiro", ["acme"], memory
    )
    state["shared_context"] = context
    return state

# Graph: CRM → Shared Memory → Financeiro
graph = StateGraph(AgentState)
graph.add_node("crm", crm_node)
graph.add_node("financeiro", financeiro_node)
graph.add_edge("crm", "financeiro")