database

数据库技能

为 Golden Armada AI Agent Fleet Platform 提供全面的数据库功能。

何时使用此技能

在处理以下工作时启用此技能：

• 数据库架构设计
• SQL 查询编写与优化
• 数据库迁移
• 索引优化
• 数据建模

PostgreSQL 快速参考

连接

# Connect
psql -h localhost -U postgres -d golden_armada

# Connection string
postgresql://user:password@host:5432/database

# Common psql commands
\l                  # List databases
\c database_name    # Connect to database
\dt                 # List tables
\d table_name       # Describe table
\di                 # List indexes
\q                  # Quit

架构设计

-- Create table with common patterns
CREATE TABLE agents (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    name VARCHAR(100) NOT NULL,
    type VARCHAR(50) NOT NULL CHECK (type IN ('claude', 'gpt', 'gemini')),
    status VARCHAR(20) DEFAULT 'idle',
    config JSONB DEFAULT '{}',
    created_at TIMESTAMPTZ DEFAULT NOW(),
    updated_at TIMESTAMPTZ DEFAULT NOW(),
    deleted_at TIMESTAMPTZ  -- Soft delete
);

-- Create index
CREATE INDEX idx_agents_type ON agents(type);
CREATE INDEX idx_agents_status ON agents(status) WHERE deleted_at IS NULL;
CREATE INDEX idx_agents_config ON agents USING GIN(config);

-- Add foreign key
CREATE TABLE tasks (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    agent_id UUID NOT NULL REFERENCES agents(id) ON DELETE CASCADE,
    message TEXT NOT NULL,
    result TEXT,
    status VARCHAR(20) DEFAULT 'pending',
    created_at TIMESTAMPTZ DEFAULT NOW()
);

常用查询

-- Basic CRUD
INSERT INTO agents (name, type) VALUES ('agent-1', 'claude') RETURNING *;
SELECT * FROM agents WHERE type = 'claude' AND deleted_at IS NULL;
UPDATE agents SET status = 'active' WHERE id = $1 RETURNING *;
DELETE FROM agents WHERE id = $1;

-- Joins
SELECT a.name, COUNT(t.id) as task_count
FROM agents a
LEFT JOIN tasks t ON a.id = t.agent_id
WHERE a.deleted_at IS NULL
GROUP BY a.id;

-- JSON operations
SELECT * FROM agents WHERE config->>'model' = 'claude-sonnet-5';
SELECT * FROM agents WHERE config @> '{"enabled": true}';
UPDATE agents SET config = config || '{"version": "2.0"}' WHERE id = $1;

-- Window functions
SELECT
    name,
    created_at,
    ROW_NUMBER() OVER (ORDER BY created_at) as row_num,
    LAG(created_at) OVER (ORDER BY created_at) as prev_created
FROM agents;

-- CTEs
WITH active_agents AS (
    SELECT * FROM agents WHERE status = 'active'
)
SELECT * FROM active_agents WHERE type = 'claude';

迁移

-- Migration: 001_create_agents.sql
BEGIN;

CREATE TABLE agents (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    name VARCHAR(100) NOT NULL,
    type VARCHAR(50) NOT NULL,
    created_at TIMESTAMPTZ DEFAULT NOW()
);

CREATE INDEX idx_agents_type ON agents(type);

COMMIT;

-- Rollback: 001_create_agents.sql
BEGIN;
DROP TABLE IF EXISTS agents;
COMMIT;

性能优化

-- Analyze query plan
EXPLAIN ANALYZE SELECT * FROM agents WHERE type = 'claude';

-- Check index usage
SELECT
    schemaname,
    tablename,
    indexname,
    idx_scan,
    idx_tup_read
FROM pg_stat_user_indexes;

-- Find slow queries
SELECT
    query,
    calls,
    mean_time,
    total_time
FROM pg_stat_statements
ORDER BY mean_time DESC
LIMIT 10;

-- Vacuum and analyze
VACUUM ANALYZE agents;

SQLAlchemy ORM

from sqlalchemy import Column, String, DateTime, ForeignKey, JSON
from sqlalchemy.dialects.postgresql import UUID
from sqlalchemy.orm import relationship
from sqlalchemy.sql import func
import uuid

class Agent(Base):
    __tablename__ = 'agents'

    id = Column(UUID(as_uuid=True), primary_key=True, default=uuid.uuid4)
    name = Column(String(100), nullable=False)
    type = Column(String(50), nullable=False)
    status = Column(String(20), default='idle')
    config = Column(JSON, default={})
    created_at = Column(DateTime(timezone=True), server_default=func.now())
    updated_at = Column(DateTime(timezone=True), onupdate=func.now())

    tasks = relationship("Task", back_populates="agent", cascade="all, delete-orphan")

class Task(Base):
    __tablename__ = 'tasks'

    id = Column(UUID(as_uuid=True), primary_key=True, default=uuid.uuid4)
    agent_id = Column(UUID(as_uuid=True), ForeignKey('agents.id'), nullable=False)
    message = Column(String, nullable=False)
    result = Column(String)

    agent = relationship("Agent", back_populates="tasks")

最佳实践

1. 在分布式系统中使用 UUID 作为主键
2. 为经常查询的列添加索引
3. 对重要数据使用软删除（deleted_at）
4. 使用 JSONB 存储灵活数据，并添加 GIN 索引
5. 使用外键约束确保数据完整性
6. 记录创建和更新时间戳以便审计
7. 使用连接池提升性能