feat: Add adapter for CrewAI

Author: XiuShenAl

README.md

@@ -60,6 +60,7 @@
 > | AutoGen                                                      | 🚧             | ✅    | 🚧       |
 > | Microsoft Agent Framework                                    | 🚧             | 🚧    | 🚧       |
 > | [Agno](https://runtime.agentscope.io/en/agno_guidelines.html) | ✅             | ✅    | 🚧       |
+> | [CrewAI](https://runtime.agentscope.io/en/crewai_guidelines.html) | ✅             | ✅    | 🚧       |
 
 ---
 

README_zh.md

@@ -60,6 +60,7 @@
 > | AutoGen                                                      | 🚧           | ✅    | 🚧    |
 > | Microsoft Agent Framework                                    | 🚧           | 🚧    | 🚧    |
 > | [Agno](https://runtime.agentscope.io/zh/agno_guidelines.html) | 🧪           | ✅    | 🚧    |
+> | [CrewAI](https://runtime.agentscope.io/zh/crewai_guidelines.html) | ✅           | ✅    | 🚧    |
 
 ---
 

cookbook/_toc.yml

@@ -44,6 +44,7 @@ parts:
       - file: en/ut.md
       - file: en/langgraph_guidelines.md
       - file: en/agno_guidelines.md
+      - file: en/crewai_guidelines.md
   - file: en/contribute.md
     sections:
       - file: en/README.md
@@ -107,6 +108,7 @@ parts:
       - file: zh/ut.md
       - file: zh/langgraph_guidelines.md
       - file: zh/agno_guidelines.md
+      - file: zh/crewai_guidelines.md
   - file: zh/contribute.md
     sections:
     - file: zh/README.md

cookbook/en/crewai_guidelines.md

@@ -0,0 +1,205 @@
+---
+jupytext:
+  formats: md:myst
+  text_representation:
+    extension: .md
+    format_name: myst
+    format_version: 0.13
+    jupytext_version: 1.11.5
+kernelspec:
+  display_name: Python 3
+  language: python
+  name: python3
+---
+
+# CrewAI Integration Guide
+
+This document describes how to integrate and use the CrewAI framework within AgentScope Runtime to build collaborative autonomous agents that support multi-turn conversations, session memory, and streaming responses.
+
+## 📦 Example Overview
+
+The following example demonstrates how to use the CrewAI framework inside AgentScope Runtime:
+
+- Uses the Qwen-Plus model from DashScope.
+- Orchestrates a simple research task with one agent.
+- Supports multi-turn conversation and session memory.
+- Employs streaming output (SSE) to return responses in real-time.
+- Implements session history storage via an in-memory service (InMemorySessionHistoryService).
+- Can be accessed through an OpenAI-compatible API mode.
+
+Here’s the core code:
+
+```{code-cell}
+# crewai_agent.py
+# -*- coding: utf-8 -*-
+import os
+from agentscope_runtime.engine import AgentApp
+from agentscope_runtime.engine.schemas.agent_schemas import AgentRequest
+from agentscope_runtime.engine.services.session_history import InMemorySessionHistoryService
+from agentscope_runtime.adapters.crewai.memory import create_crewai_session_history_memory
+
+from crewai import Agent, LLM, Crew, Task
+
+PORT = 8090
+
+def run_app():
+    """Start AgentApp and enable streaming output."""
+    agent_app = AgentApp(
+        app_name="Friday",
+        app_description="A helpful assistant",
+    )
+
+    @agent_app.init
+    async def init_func(self):
+        # Initialize the session history service
+        self.session_history_service = InMemorySessionHistoryService()
+
+
+    @agent_app.query(framework="crewai")
+    async def query_func(
+        self,
+        msgs,
+        request: AgentRequest = None,
+        **kwargs,
+    ):
+        """Handle agent queries using CrewAI."""
+
+        # Extract user query from the input message
+        user_question = msgs[0]["content"][0]["text"]
+
+        # Initialize the LLM
+        llm = LLM(
+            model="qwen-plus",
+            api_key=os.environ["DASHSCOPE_API_KEY"],
+            base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
+            stream=True,
+        )
+
+        # Create session-specific memory for the crew
+        memory = await create_crewai_session_history_memory(
+            service_or_class=self.session_history_service,
+            user_id=request.user_id,
+            session_id=request.session_id,
+        )
+
+        # Define the Research Agent
+        research_analyst = Agent(
+            role="Expert Research Analyst",
+            goal="Analyze the user's question and provide a clear, concise, and accurate answer.",
+            backstory=(
+                "You are an expert analyst at a world-renowned research institute. "
+                "You are known for your ability to break down complex questions and "
+                "deliver well-structured, easy-to-understand answers."
+            ),
+            llm=llm,
+        )
+
+        # Define the Research Task
+        research_task = Task(
+            description=f"Investigate the following user query: '{user_question}'",
+            expected_output=(
+                "A comprehensive yet easy-to-read answer that directly addresses the user's query. "
+                "The answer should be well-formatted and factually correct."
+            ),
+            agent=research_analyst,
+        )
+
+        # Assemble the crew
+        crew = Crew(
+            agents=[research_analyst],
+            tasks=[research_task],
+            external_memory=memory,
+            stream=True,
+        )
+
+        # Kick off the crew and stream the results
+        async for chunk in await crew.akickoff():
+            yield chunk
+
+
+    agent_app.run(host="127.0.0.1", port=PORT)
+
+
+if __name__ == "__main__":
+    run_app()
+```
+
+## ⚙️ Prerequisites
+
+```{note}
+Before starting, make sure you have installed AgentScope Runtime and CrewAI, and configured the required API keys.
+```
+
+1. **Install dependencies**:
+
+   ```bash
+   pip install "agentscope-runtime[ext]"
+   ```
+
+2. **Set environment variables** (DashScope provides the API key for Qwen models):
+
+   ```bash
+   export DASHSCOPE_API_KEY="your-dashscope-api-key"
+   ```
+
+## ▶️ Run the Example
+
+Run the example:
+
+```bash
+python crewai_agent.py
+```
+
+## 🌐 API Interaction
+
+### 1. Ask the Agent (`/process`)
+
+You can send an HTTP POST request to interact with the agent, with SSE streaming enabled:
+
+```bash
+curl -N \
+  -X POST "http://localhost:8090/process" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "input": [
+      {
+        "role": "user",
+        "content": [
+          { "type": "text", "text": "What is the capital of France?" }
+        ]
+      }
+    ],
+    "session_id": "session_1"
+  }'
+```
+
+### 2. OpenAI-Compatible Mode
+
+This example also supports the **OpenAI Compatible API**:
+
+```python
+from openai import OpenAI
+
+client = OpenAI(base_url="http://127.0.0.1:8090/compatible-mode/v1")
+resp = client.responses.create(
+    model="any_model",
+    input="What is CrewAI?",
+)
+print(resp.response["output"][0]["content"][0]["text"])
+```
+
+## 🔧 Customization
+
+You can extend this example by:
+
+1. **Changing the model**: Replace `LLM(model="qwen-plus", ...)` with another model.
+2. **Adding system prompts**:
+   - Modify the agent's role, goal, and backstory to change its persona and expertise.
+   - Improve the task's description and expected_output for more specific results.
+   - Add more Agent and Task instances to the Crew to build more complex, multi-agent workflows for collaboration and delegation.
+3. **Use Different Tools**: Assign tools to your agents to allow them to interact with external services, such as searching the web or accessing databases.
+
+## 📚 References
+
+- [CrewAI Documentation](https://docs.crewai.com/)
+- [AgentScope Runtime Documentation](https://runtime.agentscope.io/)