Note

Go to the end to download the full example code.

Model

In this tutorial, we introduce the model APIs integrated in AgentScope, how to use them and how to integrate new model APIs. The supported model APIs and providers include:

API

Class

Compatible

Streaming

Tools

Vision

Reasoning

OpenAI

OpenAIChatModel

vLLM, DeepSeek

✅

✅

✅

✅

DashScope

DashScopeChatModel

✅

✅

✅

✅

Anthropic

AnthropicChatModel

✅

✅

✅

✅

Gemini

GeminiChatModel

✅

✅

✅

✅

Ollama

OllamaChatModel

✅

✅

✅

✅

Note

When using vLLM, you need to configure the appropriate tool calling parameters for different models during deployment, such as --enable-auto-tool-choice, --tool-call-parser, etc. For more details, refer to the official vLLM documentation.

Note

For OpenAI-compatible models (e.g. vLLM, Deepseek), developers can use the OpenAIChatModel class, and specify the API endpoint by the client_kwargs parameter: client_kwargs={"base_url": "http://your-api-endpoint"}. For example:

OpenAIChatModel(client_kwargs={"base_url": "http://localhost:8000/v1"})

Note

Model behavior parameters (such as temperature, maximum length, etc.) can be preset in the constructor function via the generate_kwargs parameter. For example:

OpenAIChatModel(generate_kwargs={"temperature": 0.3, "max_tokens": 1000})

To provide unified model interfaces, the above model classes has the following common methods:

  • The first three arguments of the __call__ method are messages , tools and tool_choice, representing the input messages, JSON schema of tool functions, and tool selection mode, respectively.

  • The return type are either a ChatResponse instance or an async generator of ChatResponse in streaming mode.

Note

Different model APIs differ in the input message format, refer to Prompt Formatter for more details.

The ChatResponse instance contains the generated thinking/text/tool use content, identity, created time and usage information.

import asyncio
import json
import os

from agentscope.message import TextBlock, ToolUseBlock, ThinkingBlock, Msg
from agentscope.model import ChatResponse, DashScopeChatModel

response = ChatResponse(
    content=[
        ThinkingBlock(
            type="thinking",
            thinking="I should search for AgentScope on Google.",
        ),
        TextBlock(type="text", text="I'll search for AgentScope on Google."),
        ToolUseBlock(
            type="tool_use",
            id="642n298gjna",
            name="google_search",
            input={"query": "AgentScope?"},
        ),
    ],
)

print(response)

ChatResponse(content=[{'type': 'thinking', 'thinking': 'I should search for AgentScope on Google.'}, {'type': 'text', 'text': "I'll search for AgentScope on Google."}, {'type': 'tool_use', 'id': '642n298gjna', 'name': 'google_search', 'input': {'query': 'AgentScope?'}}], id='2026-04-20 12:49:12.580_43219b', created_at='2026-04-20 12:49:12.580', type='chat', usage=None, metadata=None)

Taking DashScopeChatModel as an example, we can use it to create a chat model instance and call it with messages and tools:

async def example_model_call() -> None:
    """An example of using the DashScopeChatModel."""
    model = DashScopeChatModel(
        model_name="qwen-max",
        api_key=os.environ["DASHSCOPE_API_KEY"],
        stream=False,
    )

    res = await model(
        messages=[
            {"role": "user", "content": "Hi!"},
        ],
    )

    # You can directly create a ``Msg`` object with the response content
    msg_res = Msg("Friday", res.content, "assistant")

    print("The response:", res)
    print("The response as Msg:", msg_res)


asyncio.run(example_model_call())

The response: ChatResponse(content=[{'type': 'text', 'text': 'Hello! How can I assist you today?'}], id='8af67c32-cb99-9d97-8c31-e744e6d45ac1', created_at='2026-04-20 12:49:14.272', type='chat', usage=ChatUsage(input_tokens=10, output_tokens=9, time=1.690446, type='chat', metadata=GenerationUsage(input_tokens=10, output_tokens=9)), metadata=None)
The response as Msg: Msg(id='Pob9KGTUk7zxk7TT2uJv33', name='Friday', content=[{'type': 'text', 'text': 'Hello! How can I assist you today?'}], role='assistant', metadata={}, timestamp='2026-04-20 12:49:14.272', invocation_id='None')

Streaming

To enable streaming model, set the stream parameter in the model constructor to True. When streaming is enabled, the __call__ method will return an async generator that yields ChatResponse instances as they are generated by the model.

Note

The streaming mode in AgentScope is designed to be cumulative, meaning the content in each chunk contains all the previous content plus the newly generated content.

async def example_streaming() -> None:
    """An example of using the streaming model."""
    model = DashScopeChatModel(
        model_name="qwen-max",
        api_key=os.environ["DASHSCOPE_API_KEY"],
        stream=True,
    )

    generator = await model(
        messages=[
            {
                "role": "user",
                "content": "Count from 1 to 20, and just report the number without any other information.",
            },
        ],
    )
    print("The type of the response:", type(generator))

    i = 0
    async for chunk in generator:
        print(f"Chunk {i}")
        print(f"\ttype: {type(chunk.content)}")
        print(f"\t{chunk}\n")
        i += 1


asyncio.run(example_streaming())

The type of the response: <class 'async_generator'>
Chunk 0
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:15.553', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=1, time=1.280277, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=1)), metadata=None)

Chunk 1
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:15.616', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=4, time=1.342982, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=4)), metadata=None)

Chunk 2
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:15.680', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=7, time=1.407266, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=7)), metadata=None)

Chunk 3
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4\n5\n'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:15.744', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=10, time=1.47084, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=10)), metadata=None)

Chunk 4
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4\n5\n6\n7\n8\n'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:16.971', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=16, time=2.697405, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=16)), metadata=None)

Chunk 5
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n1'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:17.078', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=22, time=2.804427, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=22)), metadata=None)

Chunk 6
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12\n1'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:17.208', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=28, time=2.935222, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=28)), metadata=None)

Chunk 7
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12\n13\n14\n1'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:18.112', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=34, time=3.839073, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=34)), metadata=None)

Chunk 8
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12\n13\n14\n15\n16\n1'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:18.236', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=40, time=3.962386, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=40)), metadata=None)

Chunk 9
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12\n13\n14\n15\n16\n17\n18\n1'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:18.369', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=46, time=4.09541, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=46)), metadata=None)

Chunk 10
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12\n13\n14\n15\n16\n17\n18\n19\n20'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:18.552', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=50, time=4.278763, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=50)), metadata=None)

Chunk 11
        type: <class 'list'>
        ChatResponse(content=[{'type': 'text', 'text': '1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12\n13\n14\n15\n16\n17\n18\n19\n20'}], id='66c19af6-bf4b-95d9-be9b-2c3181f4e9e3', created_at='2026-04-20 12:49:18.584', type='chat', usage=ChatUsage(input_tokens=27, output_tokens=50, time=4.310658, type='chat', metadata=GenerationUsage(input_tokens=27, output_tokens=50)), metadata=None)

Reasoning

AgentScope supports reasoning models by providing the ThinkingBlock.

async def example_reasoning() -> None:
    """An example of using the reasoning model."""
    model = DashScopeChatModel(
        model_name="qwen-turbo",
        api_key=os.environ["DASHSCOPE_API_KEY"],
        enable_thinking=True,
    )

    res = await model(
        messages=[
            {"role": "user", "content": "Who am I?"},
        ],
    )

    last_chunk = None
    async for chunk in res:
        last_chunk = chunk
    print("The final response:")
    print(last_chunk)


asyncio.run(example_reasoning())

The final response:
ChatResponse(content=[{'type': 'thinking', 'thinking': 'Okay, the user asked, "Who am I?" I need to figure out how to respond. First, I should consider the context. Since the user hasn\'t provided any specific information about themselves, I can\'t know their identity. My role is to be helpful and guide them to provide more details.\n\nI should start by acknowledging that I don\'t have enough information to determine their identity. Then, I can ask them to share more about themselves, like their interests, background, or what they\'re looking for. This way, I can better assist them. I need to keep the tone friendly and encouraging. Let me make sure my response is clear and not too long. Also, avoid any assumptions. Just a straightforward, polite question to get them to elaborate.'}, {'type': 'text', 'text': 'You are a unique individual with your own thoughts, experiences, and identity. Since I don’t have access to personal information about you, I can’t define "who you are" in a specific way. However, I’m here to help you explore your thoughts, answer questions, or support you in any way you need! What would you like to discuss or learn about? 😊'}], id='9ce7773c-8596-913a-a240-76dcbf590137', created_at='2026-04-20 12:49:21.682', type='chat', usage=ChatUsage(input_tokens=12, output_tokens=234, time=3.094356, type='chat', metadata=GenerationUsage(input_tokens=12, output_tokens=234)), metadata=None)

Tools API

Different model providers differ in their tools APIs, e.g. the tools JSON schema, the tool call/response format. To provide a unified interface, AgentScope solves the problem by:

  • Providing unified tool call block ToolUseBlock and tool response block ToolResultBlock, respectively.

  • Providing a unified tools interface in the __call__ method of the model classes, that accepts a list of tools JSON schemas as follows:

json_schemas = [
    {
        "type": "function",
        "function": {
            "name": "google_search",
            "description": "Search for a query on Google.",
            "parameters": {
                "type": "object",
                "properties": {
                    "query": {
                        "type": "string",
                        "description": "The search query.",
                    },
                },
                "required": ["query"],
            },
        },
    },
]

Further Reading

Total running time of the script: (0 minutes 9.106 seconds)

Gallery generated by Sphinx-Gallery