如何调度自定义回调事件

先决条件

本指南假定您熟悉以下概念：

回调
自定义回调处理程序
Astream 事件 APIastream_events方法将显示自定义回调事件。

在某些情况下，您可能希望从 Runnable 中调度自定义回调事件，以便它可以显示出来在自定义回调处理程序中或通过 Astream Events API 进行。

例如，如果您有一个包含多个步骤的长时间运行的工具，则可以在步骤之间调度自定义事件，并使用这些自定义事件来监控进度。您还可以向应用程序的最终用户显示这些自定义事件，以向他们展示当前任务的进度。

要调度自定义事件，您需要为事件确定两个属性：name和data.

属性	类型	描述
name	str	A user defined name for the event.
data	Any	The data associated with the event. This can be anything, though we suggest making it JSON serializable.

重要

调度自定义回调事件需要langchain-core>=0.2.15.
自定义回调事件只能从现有的Runnable.
如果使用astream_events，您必须使用version='v2'以查看自定义事件。
尚不支持在 LangSmith 中发送或呈现自定义回调事件。

兼容性

如果您在 python<=3.10 中运行异步代码，则 LangChain 无法自动将配置（包括 astream_events（）所需的回调）传播到子可运行对象。这是您可能无法看到从自定义可运行对象或工具发出的事件的常见原因。

如果您运行的是 python<=3.10，则需要手动传播RunnableConfigobject 添加到异步环境中的子 runnable 中。有关如何手动传播配置的示例，请参阅barRunnableLambda 的 Lambda。

如果您运行的是 python>=3.11，则RunnableConfig将自动传播到异步环境中的子可运行对象。但是，将RunnableConfigmanually （如果您的代码可以在其他 Python 版本中运行）。

Astream 事件 API

使用自定义事件最有用的方法是通过 Astream Events API。

我们可以使用async adispatch_custom_event用于在异步设置中发出自定义事件的 API。

重要

要通过 astream events API 查看自定义事件，您需要使用较新的v2API 的astream_events.

from langchain_core.callbacks.manager import (
    adispatch_custom_event,
)
from langchain_core.runnables import RunnableLambda
from langchain_core.runnables.config import RunnableConfig


@RunnableLambda
async def foo(x: str) -> str:
    await adispatch_custom_event("event1", {"x": x})
    await adispatch_custom_event("event2", 5)
    return x


async for event in foo.astream_events("hello world", version="v2"):
    print(event)

API 参考：adispatch_custom_event | RunnableLambda | RunnableConfig

{'event': 'on_chain_start', 'data': {'input': 'hello world'}, 'name': 'foo', 'tags': [], 'run_id': 'f354ffe8-4c22-4881-890a-c1cad038a9a6', 'metadata': {}, 'parent_ids': []}
{'event': 'on_custom_event', 'run_id': 'f354ffe8-4c22-4881-890a-c1cad038a9a6', 'name': 'event1', 'tags': [], 'metadata': {}, 'data': {'x': 'hello world'}, 'parent_ids': []}
{'event': 'on_custom_event', 'run_id': 'f354ffe8-4c22-4881-890a-c1cad038a9a6', 'name': 'event2', 'tags': [], 'metadata': {}, 'data': 5, 'parent_ids': []}
{'event': 'on_chain_stream', 'run_id': 'f354ffe8-4c22-4881-890a-c1cad038a9a6', 'name': 'foo', 'tags': [], 'metadata': {}, 'data': {'chunk': 'hello world'}, 'parent_ids': []}
{'event': 'on_chain_end', 'data': {'output': 'hello world'}, 'run_id': 'f354ffe8-4c22-4881-890a-c1cad038a9a6', 'name': 'foo', 'tags': [], 'metadata': {}, 'parent_ids': []}

在 python <= 3.10 中，您必须手动传播配置！

from langchain_core.callbacks.manager import (
    adispatch_custom_event,
)
from langchain_core.runnables import RunnableLambda
from langchain_core.runnables.config import RunnableConfig


@RunnableLambda
async def bar(x: str, config: RunnableConfig) -> str:
    """An example that shows how to manually propagate config.

    You must do this if you're running python<=3.10.
    """
    await adispatch_custom_event("event1", {"x": x}, config=config)
    await adispatch_custom_event("event2", 5, config=config)
    return x


async for event in bar.astream_events("hello world", version="v2"):
    print(event)

API 参考：adispatch_custom_event | RunnableLambda | RunnableConfig

{'event': 'on_chain_start', 'data': {'input': 'hello world'}, 'name': 'bar', 'tags': [], 'run_id': 'c787b09d-698a-41b9-8290-92aaa656f3e7', 'metadata': {}, 'parent_ids': []}
{'event': 'on_custom_event', 'run_id': 'c787b09d-698a-41b9-8290-92aaa656f3e7', 'name': 'event1', 'tags': [], 'metadata': {}, 'data': {'x': 'hello world'}, 'parent_ids': []}
{'event': 'on_custom_event', 'run_id': 'c787b09d-698a-41b9-8290-92aaa656f3e7', 'name': 'event2', 'tags': [], 'metadata': {}, 'data': 5, 'parent_ids': []}
{'event': 'on_chain_stream', 'run_id': 'c787b09d-698a-41b9-8290-92aaa656f3e7', 'name': 'bar', 'tags': [], 'metadata': {}, 'data': {'chunk': 'hello world'}, 'parent_ids': []}
{'event': 'on_chain_end', 'data': {'output': 'hello world'}, 'run_id': 'c787b09d-698a-41b9-8290-92aaa656f3e7', 'name': 'bar', 'tags': [], 'metadata': {}, 'parent_ids': []}

异步回调处理程序

您还可以通过异步回调处理程序使用 dispatched 事件。

from typing import Any, Dict, List, Optional
from uuid import UUID

from langchain_core.callbacks import AsyncCallbackHandler
from langchain_core.callbacks.manager import (
    adispatch_custom_event,
)
from langchain_core.runnables import RunnableLambda
from langchain_core.runnables.config import RunnableConfig


class AsyncCustomCallbackHandler(AsyncCallbackHandler):
    async def on_custom_event(
        self,
        name: str,
        data: Any,
        *,
        run_id: UUID,
        tags: Optional[List[str]] = None,
        metadata: Optional[Dict[str, Any]] = None,
        **kwargs: Any,
    ) -> None:
        print(
            f"Received event {name} with data: {data}, with tags: {tags}, with metadata: {metadata} and run_id: {run_id}"
        )


@RunnableLambda
async def bar(x: str, config: RunnableConfig) -> str:
    """An example that shows how to manually propagate config.

    You must do this if you're running python<=3.10.
    """
    await adispatch_custom_event("event1", {"x": x}, config=config)
    await adispatch_custom_event("event2", 5, config=config)
    return x


async_handler = AsyncCustomCallbackHandler()
await foo.ainvoke(1, {"callbacks": [async_handler], "tags": ["foo", "bar"]})

API 参考：AsyncCallbackHandler | adispatch_custom_event | RunnableLambda | RunnableConfig

Received event event1 with data: {'x': 1}, with tags: ['foo', 'bar'], with metadata: {} and run_id: a62b84be-7afd-4829-9947-7165df1f37d9
Received event event2 with data: 5, with tags: ['foo', 'bar'], with metadata: {} and run_id: a62b84be-7afd-4829-9947-7165df1f37d9

同步回调处理程序

让我们看看如何使用dispatch_custom_event.

您必须调用dispatch_custom_event从现有的Runnable.

from typing import Any, Dict, List, Optional
from uuid import UUID

from langchain_core.callbacks import BaseCallbackHandler
from langchain_core.callbacks.manager import (
    dispatch_custom_event,
)
from langchain_core.runnables import RunnableLambda
from langchain_core.runnables.config import RunnableConfig


class CustomHandler(BaseCallbackHandler):
    def on_custom_event(
        self,
        name: str,
        data: Any,
        *,
        run_id: UUID,
        tags: Optional[List[str]] = None,
        metadata: Optional[Dict[str, Any]] = None,
        **kwargs: Any,
    ) -> None:
        print(
            f"Received event {name} with data: {data}, with tags: {tags}, with metadata: {metadata} and run_id: {run_id}"
        )


@RunnableLambda
def foo(x: int, config: RunnableConfig) -> int:
    dispatch_custom_event("event1", {"x": x})
    dispatch_custom_event("event2", {"x": x})
    return x


handler = CustomHandler()
foo.invoke(1, {"callbacks": [handler], "tags": ["foo", "bar"]})

API 参考：BaseCallbackHandler | dispatch_custom_event | RunnableLambda | RunnableConfig 的

Received event event1 with data: {'x': 1}, with tags: ['foo', 'bar'], with metadata: {} and run_id: 27b5ce33-dc26-4b34-92dd-08a89cb22268
Received event event2 with data: {'x': 1}, with tags: ['foo', 'bar'], with metadata: {} and run_id: 27b5ce33-dc26-4b34-92dd-08a89cb22268

后续步骤

您已经了解了如何发出自定义事件，您可以查看更深入的 astream 事件指南，这是利用自定义事件的最简单方法。