hashgraph-online/awesome-codex-plugins / plugins/Colin4k1024/tsp/skills/adk-framework-adapters
adk-framework-adapters
VerifiedADK 框架适配层。为 LangChain / EINO / AutoGen / AgentScope / CrewAI 提供框架特定的 代码模板、惯用模式、API 映射和项目结构,供 agent-dev-workshop Phase 5 代码生成使用。 每个框架 reference 文件标注 verified_date 用于版本锁定。
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.7 KB |
| references/agentscope.md | markdown | 5.7 KB |
| references/autogen.md | markdown | 5.0 KB |
| references/crewai.md | markdown | 6.3 KB |
| references/eino.md | markdown | 5.8 KB |
| references/langchain.md | markdown | 4.6 KB |
references/crewai.md
View on GitHub---
framework: crewai
version: "0.100.x"
language: python
verified_date: "2025-01-15"
---
# CrewAI 适配器
## 框架概述
- **语言**: Python
- **核心库**: `crewai`, `crewai-tools`
- **特点**: 角色扮演式多 Agent、高层抽象、任务委派、流程编排
- **适用**: 业务流程自动化、角色化团队协作、报告生成
## 项目模板
```
{project_name}/
├── pyproject.toml
├── .env.example
├── README.md
├── src/
│ ├── __init__.py
│ ├── crew.py # Crew 定义(核心)
│ ├── agents.py # Agent 角色定义
│ ├── tasks.py # 任务定义
│ ├── tools.py # 自定义工具
│ └── config.py # 配置
├── tests/
│ ├── __init__.py
│ ├── test_crew.py
│ └── conftest.py
└── scripts/
└── run.py
```
## 依赖配置
```toml
# pyproject.toml
[project]
name = "{project_name}"
version = "0.1.0"
requires-python = ">=3.10"
dependencies = [
"crewai>=0.100.0",
"crewai-tools>=0.17.0",
"python-dotenv>=1.0.0",
]
[project.optional-dependencies]
dev = ["pytest>=8.0"]
```
## Agent 创建模式
### 角色定义
```python
from crewai import Agent
researcher = Agent(
role="高级研究员",
goal="查找和整理关于{topic}的最新信息",
backstory="""你是一位经验丰富的研究员,擅长从海量信息中
提取关键洞察。你以严谨著称,总是引用可靠来源。""",
tools=[search_tool, web_scraper],
llm="openai/gpt-4o",
verbose=True,
allow_delegation=True, # 允许委派任务给其他 Agent
)
analyst = Agent(
role="数据分析师",
goal="分析研究数据并提取有价值的商业洞察",
backstory="""你是一位数据分析专家,擅长将复杂数据转化为
清晰的商业建议。你的分析总是数据驱动的。""",
tools=[analysis_tool],
llm="openai/gpt-4o",
verbose=True,
)
writer = Agent(
role="报告撰写人",
goal="将研究和分析结果撰写成专业报告",
backstory="""你是一位专业的商业写手,擅长将复杂信息
组织成清晰、可操作的报告。""",
llm="openai/gpt-4o",
verbose=True,
)
```
## 任务定义模式
```python
from crewai import Task
research_task = Task(
description="研究{topic}的最新趋势和关键数据点",
expected_output="包含关键发现、数据点和来源引用的研究报告",
agent=researcher,
)
analysis_task = Task(
description="基于研究报告,分析数据并提取商业洞察",
expected_output="包含数据分析、趋势预测和建议的分析报告",
agent=analyst,
context=[research_task], # 依赖上一个任务
)
report_task = Task(
description="将研究和分析整合成最终报告",
expected_output="格式规范的最终报告,包含摘要、分析和建议",
agent=writer,
context=[research_task, analysis_task],
output_file="output/report.md", # 输出到文件
)
```
## 工具定义模式
```python
from crewai.tools import BaseTool
from pydantic import BaseModel, Field
class SearchOrdersInput(BaseModel):
query: str = Field(description="搜索关键词:订单号或客户名")
status: str | None = Field(default=None, description="可选,按状态过滤")
class SearchOrdersTool(BaseTool):
name: str = "search_orders"
description: str = "按订单号或客户名搜索订单列表"
args_schema: type[BaseModel] = SearchOrdersInput
def _run(self, query: str, status: str = None) -> str:
results = do_search(query, status)
return str(results)
# 或使用 crewai-tools 内置工具
from crewai_tools import SerperDevTool, WebsiteSearchTool
search_tool = SerperDevTool()
web_tool = WebsiteSearchTool()
```
## Crew 编排
```python
from crewai import Crew, Process
# 顺序流程
crew = Crew(
agents=[researcher, analyst, writer],
tasks=[research_task, analysis_task, report_task],
process=Process.sequential,
verbose=True,
)
# 层级流程(带管理者)
crew = Crew(
agents=[researcher, analyst, writer],
tasks=[research_task, analysis_task, report_task],
process=Process.hierarchical,
manager_llm="openai/gpt-4o",
verbose=True,
)
```
### CrewAI Flow(高级编排)
```python
from crewai.flow.flow import Flow, listen, start
class ResearchFlow(Flow):
@start()
def collect_requirements(self):
return self.state.get("topic", "AI Agent")
@listen(collect_requirements)
def run_research(self, topic):
crew = Crew(
agents=[researcher],
tasks=[research_task],
process=Process.sequential,
)
return crew.kickoff(inputs={"topic": topic})
@listen(run_research)
def run_analysis(self, research_result):
crew = Crew(
agents=[analyst, writer],
tasks=[analysis_task, report_task],
process=Process.sequential,
)
return crew.kickoff(inputs={"research": research_result})
flow = ResearchFlow()
result = flow.kickoff(inputs={"topic": "AI Agent 开发趋势"})
```
## Memory 配置
```python
# CrewAI 内置 Memory 支持
crew = Crew(
agents=[researcher, analyst],
tasks=[research_task, analysis_task],
process=Process.sequential,
memory=True, # 启用记忆
# embedder 配置(用于语义记忆)
embedder={
"provider": "openai",
"config": {"model": "text-embedding-3-small"},
},
)
```
## 运行入口
```python
from dotenv import load_dotenv
from src.crew import build_crew
load_dotenv()
def main():
crew = build_crew()
topic = input("请输入研究主题: ")
result = crew.kickoff(inputs={"topic": topic})
print("\n" + "=" * 50)
print("最终结果:")
print("=" * 50)
print(result.raw)
if __name__ == "__main__":
main()
```
## 注意事项
- CrewAI 的核心抽象是 Agent(角色)+ Task(任务)+ Crew(团队)
- `backstory` 对 Agent 行为影响很大,是角色扮演的关键
- `allow_delegation=True` 允许 Agent 将子任务委派给其他 Agent
- `context` 参数建立 Task 之间的依赖关系
- `Process.hierarchical` 自动添加 Manager Agent 进行协调
- CrewAI Flow 提供更灵活的编排,支持条件分支和循环
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |