> ## Documentation Index
> Fetch the complete documentation index at: https://cndoc-langchain.site/llms.txt
> Use this file to discover all available pages before exploring further.

# LangGraph 运行时

[`Pregel`](https://reference.langchain.com/python/langgraph/pregel/main/Pregel) 实现了 LangGraph 的运行时，负责管理 LangGraph 应用程序的执行。

编译一个 [StateGraph](https://reference.langchain.com/python/langgraph/graph/state/StateGraph) 或创建一个 [`@entrypoint`](https://reference.langchain.com/python/langgraph/func/entrypoint) 会产生一个 [`Pregel`](https://reference.langchain.com/python/langgraph/pregel/main/Pregel) 实例，该实例可以使用输入进行调用。

本指南从高层次解释了运行时，并提供了直接使用 Pregel 实现应用程序的说明。

> **注意：** [`Pregel`](https://reference.langchain.com/python/langgraph/pregel/main/Pregel) 运行时的名称来源于 [Google 的 Pregel 算法](https://research.google/pubs/pub37252/)，该算法描述了一种使用图进行大规模并行计算的高效方法。

## 概述

在 LangGraph 中，Pregel 将 [**参与者**](https://en.wikipedia.org/wiki/Actor_model) 和 **通道** 结合成一个单一的应用程序。**参与者** 从通道读取数据并向通道写入数据。Pregel 将应用程序的执行组织成多个步骤，遵循 **Pregel 算法**/**批量同步并行** 模型。

每个步骤包含三个阶段：

* **计划**：确定此步骤要执行哪些 **参与者**。例如，在第一步中，选择订阅特殊 **输入** 通道的 **参与者**；在后续步骤中，选择订阅在上一步中更新的通道的 **参与者**。
* **执行**：并行执行所有选定的 **参与者**，直到全部完成、某个失败或达到超时。在此阶段，通道更新对参与者不可见，直到下一步。
* **更新**：使用此步骤中 **参与者** 写入的值更新通道。

重复此过程，直到没有选定的 **参与者** 需要执行，或达到最大步骤数。

## 参与者

一个 **参与者** 是一个 `PregelNode`。它订阅通道，从通道读取数据，并向通道写入数据。它可以被视为 Pregel 算法中的一个 **参与者**。`PregelNodes` 实现了 LangChain 的 Runnable 接口。

## 通道

通道用于在参与者（PregelNodes）之间进行通信。每个通道有一个值类型、一个更新类型和一个更新函数——该函数接收一系列更新并修改存储的值。通道可用于将数据从一个链发送到另一个链，或在未来的步骤中将数据从链发送到自身。

### LastValue

[`LastValue`](https://reference.langchain.com/python/langgraph/channels/last_value/LastValue) 是默认的通道类型。它存储最后写入的值，覆盖任何先前的值。用于输入和输出值，或在步骤之间传递数据。

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
from langgraph.channels import LastValue

channel: LastValue[int] = LastValue(int)
```

### Topic

[`Topic`](https://reference.langchain.com/python/langgraph/channels/topic/Topic) 是一个可配置的发布订阅通道，适用于在参与者之间发送多个值或跨步骤累积输出。可以配置为对值进行去重，或累积运行期间写入的所有值。

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
from langgraph.channels import Topic

# 累积跨步骤写入的所有值
channel: Topic[str] = Topic(str, accumulate=True)
```

### BinaryOperatorAggregate

[`BinaryOperatorAggregate`](https://reference.langchain.com/python/langgraph/channels/binop/BinaryOperatorAggregate) 存储一个持久值，该值通过将二元运算符应用于当前值和每个新更新来更新。用于计算跨步骤的运行聚合。

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
import operator
from langgraph.channels import BinaryOperatorAggregate

# 运行总计：每次写入都添加到当前值
total = BinaryOperatorAggregate(int, operator.add)
```

### DeltaChannel (beta)

<Warning>
  `DeltaChannel` 需要 `langgraph>=1.2`，目前处于 beta 版本。API 可能在未来版本中更改。
</Warning>

[`DeltaChannel`](https://reference.langchain.com/python/langgraph/channels/delta/DeltaChannel) 在每个步骤仅存储增量变化，而不是完整的累积值。这对于频繁写入且随时间累积大量值的通道最有用——例如，长时间运行线程中的对话消息列表。没有增量存储时，完整列表会在每个检查点重新序列化；使用 `DeltaChannel`，则仅存储每个步骤写入的新消息。

<Tip>
  当通道既频繁写入又随时间增长时，请考虑使用 `DeltaChannel`。一个好迹象是：如果你注意到某个通道的检查点大小随线程长度线性增长，那么 `DeltaChannel` 可能很适合。
</Tip>

在 `Annotated` 类型注解中使用 `DeltaChannel`，就像使用普通归约器一样：

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
from typing import Annotated, Sequence
from typing_extensions import TypedDict
from langgraph.channels import DeltaChannel


def my_reducer(state: list[str], writes: Sequence[list[str]]) -> list[str]:
    result = list(state)
    for write in writes:
        result.extend(write)
    return result


class State(TypedDict):
    messages: Annotated[list[str], DeltaChannel(my_reducer)]
```

#### 批量归约器要求

传递给 `DeltaChannel` 的 `reducer` 是一个 **批量归约器**：它在单次调用中接收当前状态和当前步骤所有写入的 *序列*——而不是像标准归约器那样成对处理。这与在 `StateGraph` 中与 `Annotated` 一起使用的按键归约器不同，后者每次更新调用一次。

<Warning>
  批量归约器 **必须是可结合的**（批处理不变）：

  ```
  reducer(reducer(state, [xs]), [ys]) == reducer(state, [xs, ys])
  ```

  如果你的归约器不是可结合的，重建的状态可能会因 LangGraph 如何跨步骤批处理写入而异，导致不一致的行为。
</Warning>

以下是两种最常见情况的批量归约器：

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
from typing import Any, Sequence


# 列表：按顺序追加所有写入
def list_reducer(state: list[Any], writes: Sequence[list[Any]]) -> list[Any]:
    result = list(state)
    for write in writes:
        result.extend(write)
    return result


# 字典：合并所有写入，键冲突时最后一次写入生效
def dict_reducer(
    state: dict[str, Any], writes: Sequence[dict[str, Any]]
) -> dict[str, Any]:
    result = dict(state)
    for write in writes:
        result.update(write)
    return result
```

两者都是可结合的：逐批应用与一起应用产生相同的结果。

#### 使用 snapshot\_frequency 控制有界读取延迟

没有快照时，读取 `DeltaChannel` 值需要重放完整的写入历史——对于有 N 个步骤的线程，复杂度为 O(N)。设置 `snapshot_frequency=K` 会每 K 个 pregel 步骤写入一个完整快照，将读取深度限制为最多 K 个步骤：

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
class State(TypedDict):
    messages: Annotated[
        list[str],
        DeltaChannel(my_reducer, snapshot_frequency=5),
    ]
```

较高的 `snapshot_frequency` 值会减少存储开销，但会增加读取延迟。较低的值以更大的检查点为代价更严格地限制延迟。`None`（默认值）完全跳过快照——适用于读取很少或线程较短的情况。

## 示例

虽然大多数用户将通过 [StateGraph](https://reference.langchain.com/python/langgraph/graph/state/StateGraph) API 或 [`@entrypoint`](https://reference.langchain.com/python/langgraph/func/entrypoint) 装饰器与 Pregel 交互，但也可以直接与 Pregel 交互。

以下是一些不同的示例，让你了解 Pregel API。

<Tabs>
  <Tab title="单节点">
    ```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    from langgraph.channels import EphemeralValue
    from langgraph.pregel import Pregel, NodeBuilder

    node1 = (
        NodeBuilder().subscribe_only("a")
        .do(lambda x: x + x)
        .write_to("b")
    )

    app = Pregel(
        nodes={"node1": node1},
        channels={
            "a": EphemeralValue(str),
            "b": EphemeralValue(str),
        },
        input_channels=["a"],
        output_channels=["b"],
    )

    app.invoke({"a": "foo"})
    ```

    ```con theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    {'b': 'foofoo'}
    ```
  </Tab>

  <Tab title="多节点">
    ```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    from langgraph.channels import LastValue, EphemeralValue
    from langgraph.pregel import Pregel, NodeBuilder

    node1 = (
        NodeBuilder().subscribe_only("a")
        .do(lambda x: x + x)
        .write_to("b")
    )

    node2 = (
        NodeBuilder().subscribe_only("b")
        .do(lambda x: x + x)
        .write_to("c")
    )


    app = Pregel(
        nodes={"node1": node1, "node2": node2},
        channels={
            "a": EphemeralValue(str),
            "b": LastValue(str),
            "c": EphemeralValue(str),
        },
        input_channels=["a"],
        output_channels=["b", "c"],
    )

    app.invoke({"a": "foo"})
    ```

    ```con theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    {'b': 'foofoo', 'c': 'foofoofoofoo'}
    ```
  </Tab>

  <Tab title="Topic">
    ```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    from langgraph.channels import EphemeralValue, Topic
    from langgraph.pregel import Pregel, NodeBuilder

    node1 = (
        NodeBuilder().subscribe_only("a")
        .do(lambda x: x + x)
        .write_to("b", "c")
    )

    node2 = (
        NodeBuilder().subscribe_to("b")
        .do(lambda x: x["b"] + x["b"])
        .write_to("c")
    )

    app = Pregel(
        nodes={"node1": node1, "node2": node2},
        channels={
            "a": EphemeralValue(str),
            "b": EphemeralValue(str),
            "c": Topic(str, accumulate=True),
        },
        input_channels=["a"],
        output_channels=["c"],
    )

    app.invoke({"a": "foo"})
    ```

    ```pycon theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    {'c': ['foofoo', 'foofoofoofoo']}
    ```
  </Tab>

  <Tab title="BinaryOperatorAggregate">
    此示例演示如何使用 [`BinaryOperatorAggregate`](https://reference.langchain.com/python/langgraph/channels/binop/BinaryOperatorAggregate) 通道实现归约器。

    ```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    from langgraph.channels import EphemeralValue, BinaryOperatorAggregate
    from langgraph.pregel import Pregel, NodeBuilder


    node1 = (
        NodeBuilder().subscribe_only("a")
        .do(lambda x: x + x)
        .write_to("b", "c")
    )

    node2 = (
        NodeBuilder().subscribe_only("b")
        .do(lambda x: x + x)
        .write_to("c")
    )

    def reducer(current, update):
        if current:
            return current + " | " + update
        else:
            return update

    app = Pregel(
        nodes={"node1": node1, "node2": node2},
        channels={
            "a": EphemeralValue(str),
            "b": EphemeralValue(str),
            "c": BinaryOperatorAggregate(str, operator=reducer),
        },
        input_channels=["a"],
        output_channels=["c"],
    )

    app.invoke({"a": "foo"})
    ```
  </Tab>

  <Tab title="循环">
    此示例演示如何在图中引入循环，方法是让一个链写入它订阅的通道。执行将继续，直到向通道写入 `None` 值。

    ```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    from langgraph.channels import EphemeralValue
    from langgraph.pregel import Pregel, NodeBuilder, ChannelWriteEntry

    example_node = (
        NodeBuilder().subscribe_only("value")
        .do(lambda x: x + x if len(x) < 10 else None)
        .write_to(ChannelWriteEntry("value", skip_none=True))
    )

    app = Pregel(
        nodes={"example_node": example_node},
        channels={
            "value": EphemeralValue(str),
        },
        input_channels=["value"],
        output_channels=["value"],
    )

    app.invoke({"value": "a"})
    ```

    ```pycon theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    {'value': 'aaaaaaaaaaaaaaaa'}
    ```
  </Tab>
</Tabs>

## 高级 API

LangGraph 提供了两个用于创建 Pregel 应用程序的高级 API：[StateGraph (Graph API)](/oss/python/langgraph/graph-api) 和 [Functional API](/oss/python/langgraph/functional-api)。

<Tabs>
  <Tab title="StateGraph (Graph API)">
    [StateGraph (Graph API)](https://reference.langchain.com/python/langgraph/graph/state/StateGraph) 是一个更高级的抽象，简化了 Pregel 应用程序的创建。它允许你定义节点和边的图。当你编译图时，StateGraph API 会自动为你创建 Pregel 应用程序。

    ```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    from typing import TypedDict

    from langgraph.constants import START
    from langgraph.graph import StateGraph

    class Essay(TypedDict):
        topic: str
        content: str | None
        score: float | None

    def write_essay(essay: Essay):
        return {
            "content": f"Essay about {essay['topic']}",
        }

    def score_essay(essay: Essay):
        return {
            "score": 10
        }

    builder = StateGraph(Essay)
    builder.add_node(write_essay)
    builder.add_node(score_essay)
    builder.add_edge(START, "write_essay")
    builder.add_edge("write_essay", "score_essay")

    # 编译图。
    # 这将返回一个 Pregel 实例。
    graph = builder.compile()
    ```

    编译后的 Pregel 实例将与一组节点和通道相关联。你可以通过打印它们来检查节点和通道。

    ```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    print(graph.nodes)
    ```

    你会看到类似这样的内容：

    ```pycon theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    {'__start__': <langgraph.pregel.read.PregelNode at 0x7d05e3ba1810>,
     'write_essay': <langgraph.pregel.read.PregelNode at 0x7d05e3ba14d0>,
     'score_essay': <langgraph.pregel.read.PregelNode at 0x7d05e3ba1710>}
    ```

    ```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    print(graph.channels)
    ```

    你应该会看到类似这样的内容

    ```pycon theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    {'topic': <langgraph.channels.last_value.LastValue at 0x7d05e3294d80>,
     'content': <langgraph.channels.last_value.LastValue at 0x7d05e3295040>,
     'score': <langgraph.channels.last_value.LastValue at 0x7d05e3295980>,
     '__start__': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e3297e00>,
     'write_essay': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e32960c0>,
     'score_essay': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e2d8ab80>,
     'branch:__start__:__self__:write_essay': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e32941c0>,
     'branch:__start__:__self__:score_essay': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e2d88800>,
     'branch:write_essay:__self__:write_essay': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e3295ec0>,
     'branch:write_essay:__self__:score_essay': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e2d8ac00>,
     'branch:score_essay:__self__:write_essay': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e2d89700>,
     'branch:score_essay:__self__:score_essay': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e2d8b400>,
     'start:write_essay': <langgraph.channels.ephemeral_value.EphemeralValue at 0x7d05e2d8b280>}
    ```
  </Tab>

  <Tab title="Functional API">
    在 [Functional API](/oss/python/langgraph/functional-api) 中，你可以使用 [`@entrypoint`](https://reference.langchain.com/python/langgraph/func/entrypoint) 创建 Pregel 应用程序。`entrypoint` 装饰器允许你定义一个接收输入并返回输出的函数。

    ```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    from typing import TypedDict

    from langgraph.checkpoint.memory import InMemorySaver
    from langgraph.func import entrypoint

    class Essay(TypedDict):
        topic: str
        content: str | None
        score: float | None


    checkpointer = InMemorySaver()

    @entrypoint(checkpointer=checkpointer)
    def write_essay(essay: Essay):
        return {
            "content": f"Essay about {essay['topic']}",
        }

    print("Nodes: ")
    print(write_essay.nodes)
    print("Channels: ")
    print(write_essay.channels)
    ```

    ```pycon theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
    Nodes:
    {'write_essay': <langgraph.pregel.read.PregelNode object at 0x7d05e2f9aad0>}
    Channels:
    {'__start__': <langgraph.channels.ephemeral_value.EphemeralValue object at 0x7d05e2c906c0>, '__end__': <langgraph.channels.last_value.LastValue object at 0x7d05e2c90c40>, '__previous__': <langgraph.channels.last_value.LastValue object at 0x7d05e1007280>}
    ```
  </Tab>
</Tabs>

***

<div className="source-links">
  <Callout icon="terminal-2">
    [将这些文档](/use-these-docs) 通过 MCP 连接到 Claude、VSCode 等，以获取实时答案。
  </Callout>

  <Callout icon="edit">
    [在 GitHub 上编辑此页面](https://github.com/langchain-ai/docs/edit/main/src/oss/langgraph/pregel.mdx) 或 [提交问题](https://github.com/langchain-ai/docs/issues/new/choose)。
  </Callout>
</div>