Superagent: 一条 CLI 命令将 AI Agent 部署到生产环境 — 2026 最小化部署指南

npm install -g superagent-cli

# 验证安装
superagent --version
# 输出: superagent/0.4.2 linux-x64 node-v20.12.0

# 安装 Python SDK
pip install superagent-py

# 或从源码安装以获取最新功能
git clone https://github.com/superagent-ai/superagent.git
cd superagent/libs/superagent-py
pip install -e .

# 在项目根目录创建 .env 文件
cat > .env << 'EOF'
OPENAI_API_KEY=sk-your-openai-key-here
SUPERAGENT_API_URL=https://api.superagent.sh
SUPERAGENT_API_KEY=sa-your-superagent-key

# 可选：向量数据库凭证
PINECONE_API_KEY=your-pinecone-key
PINECONE_ENVIRONMENT=us-east-1

# 可选：本地 Ollama 开发
OLLAMA_BASE_URL=http://localhost:11434
EOF

# 登录 Superagent Cloud（或自建实例）
superagent login

# 创建新项目目录
mkdir my-first-agent && cd my-first-agent

# 使用模板初始化
superagent init --template qa-agent

# 部署到生产环境
superagent deploy

# 通过 curl 查询 Agent
curl -X POST https://api.superagent.sh/v1/agents/ag_01hwxyz123/invoke \
  -H "Authorization: Bearer $SUPERAGENT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input": "What are the key features of Superagent?",
    "enableStreaming": false
  }'

{
  "output": "Superagent provides: (1) One-command deployment, (2) Multi-LLM support including OpenAI and local models, (3) Built-in RAG with vector database integration, (4) REST API with streaming support, (5) Python and TypeScript SDKs, and (6) Workflow automation for chaining agents.",
  "intermediate_steps": [],
  "total_tokens": 142,
  "total_cost": 0.0021
}

from superagent.client import Superagent

client = Superagent()

# 使用 GPT-4o 创建 Agent
agent = client.agent.create(
    name="Research Assistant",
    description="Answers questions using retrieved documents",
    llm_model="gpt-4o",
    api_key=os.getenv("OPENAI_API_KEY")
)

# 切换到 Claude 3.5 Sonnet
agent_claude = client.agent.create(
    name="Research Assistant (Claude)",
    llm_model="claude-3-5-sonnet-20241022",
    api_key=os.getenv("ANTHROPIC_API_KEY")
)

from langchain.tools import DuckDuckGoSearchRun
from superagent.client import Superagent

search = DuckDuckGoSearchRun()

client = Superagent()
agent = client.agent.create(
    name="Web Search Agent",
    tools=[{
        "name": "web_search",
        "description": "Search the web for current information",
        "langchain_tool": search  # 直接传入 LangChain 工具
    }]
)

import os
from superagent.client import Superagent

client = Superagent()

# 连接 Pinecone 进行文档检索
datasource = client.datasource.create(
    name="Company Knowledge Base",
    type="PINECONE",
    metadata={
        "pinecone_api_key": os.getenv("PINECONE_API_KEY"),
        "pinecone_index_name": "company-docs",
        "pinecone_environment": "us-east-1"
    }
)

# 或使用 Weaviate
datasource_weaviate = client.datasource.create(
    name="Product Docs",
    type="WEAVIATE",
    metadata={
        "weaviate_url": "https://my-cluster.weaviate.network",
        "weaviate_api_key": os.getenv("WEAVIATE_API_KEY"),
        "class_name": "Document"
    }
)

# FastAPI 集成示例
from fastapi import FastAPI
from superagent.client import Superagent
import os

app = FastAPI()
client = Superagent(api_key=os.getenv("SUPERAGENT_API_KEY"))

@app.post("/api/ask")
async def ask_question(question: str):
    response = await client.agent.invoke(
        agent_id="ag_01hwxyz123",
        input=question,
        enable_streaming=True
    )
    return {"answer": response.output}

# 拉取官方镜像
docker pull superagentai/superagent:latest

# 使用环境变量运行
docker run -d \
  --name superagent \
  -p 3000:3000 \
  -e OPENAI_API_KEY=$OPENAI_API_KEY \
  -e DATABASE_URL=postgresql://user:pass@db:5432/superagent \
  -e NEXTAUTH_SECRET=$(openssl rand -hex 32) \
  superagentai/superagent:latest

# 验证容器运行中
docker ps | grep superagent

# docker-compose.yml 生产配置
version: "3.8"
services:
  superagent:
    image: superagentai/superagent:latest
    ports:
      - "3000:3000"
    environment:
      - OPENAI_API_KEY=${OPENAI_API_KEY}
      - DATABASE_URL=postgresql://postgres:postgres@db:5432/superagent
      - NEXTAUTH_SECRET=${NEXTAUTH_SECRET}
    depends_on:
      - db
      - redis

  db:
    image: postgres:16-alpine
    volumes:
      - pgdata:/var/lib/postgresql/data
    environment:
      - POSTGRES_PASSWORD=postgres
      - POSTGRES_DB=superagent

  redis:
    image: redis:7-alpine
    volumes:
      - redisdata:/data

volumes:
  pgdata:
  redisdata:

from superagent.client import Superagent
import requests

client = Superagent()

def get_stock_price(symbol: str) -> str:
    """从金融 API 获取实时股价。"""
    resp = requests.get(
        f"https://api.example.com/stocks/{symbol}",
        headers={"Authorization": f"Bearer {API_KEY}"}
    )
    data = resp.json()
    return f"{symbol}: ${data['price']} (change: {data['change']})"

# 注册自定义工具
client.tool.create(
    name="stock_price",
    description="Get the current stock price for a given ticker symbol",
    function=get_stock_price
)

from superagent.client import Superagent

client = Superagent()

# 方案 1：对话缓冲（默认，滑动窗口）
agent = client.agent.create(
    name="Chat Agent",
    memory={"type": "conversation_buffer", "k": 10}
)

# 方案 2：向量内存（语义检索历史对话）
agent = client.agent.create(
    name="Long Context Agent",
    memory={"type": "vector_memory", "vector_db": "pinecone"}
)

# 方案 3：Redis 会话内存（多用户应用）
agent = client.agent.create(
    name="Multi-User Agent",
    memory={"type": "redis", "ttl": 3600}  # 1 小时 TTL
)

from superagent.client import Superagent

client = Superagent()

# 定义内容生成工作流
workflow = client.workflow.create(
    name="Blog Post Pipeline",
    steps=[
        {
            "agent": "research-agent",
            "input": "Research the topic: {{topic}}",
            "output_key": "research_notes"
        },
        {
            "agent": "writer-agent",
            "input": "Write a blog post based on: {{research_notes}}",
            "output_key": "draft"
        },
        {
            "agent": "editor-agent",
            "input": "Review and improve: {{draft}}",
            "output_key": "final_post"
        }
    ]
)

# 执行工作流
result = client.workflow.invoke(
    workflow_id=workflow.id,
    inputs={"topic": "AI Agent Deployment Best Practices"}
)
print(result.steps[-1].output)  # 最终编辑后的文章

# 配置 API Key 认证
superagent config set auth.type=api_key
superagent config set auth.rate_limit=100/minute

# 启用请求日志用于审计
superagent config set logging.level=info
superagent config set logging.retention=30d

# 内置健康端点
curl https://your-superagent-instance.com/health

# 预期响应：
# {"status": "ok", "version": "0.4.2", "uptime": 86400}

# Prometheus 指标端点（启用后）
curl https://your-superagent-instance.com/metrics