x

2026-08-18 21:55:46 +02:00 · 2025-06-25 10:25:06 -04:00 · 2025-06-24 14:47:20 -04:00 · 2025-06-24 12:35:31 -04:00 · 2025-06-24 11:17:21 -04:00 · 2025-06-24 09:56:19 -04:00
88 changed files with 6047 additions and 3245 deletions
@@ -61,6 +61,7 @@ REDIRECT_MAP = {
    "how-tos/subgraph-persistence.ipynb": "how-tos/persistence.ipynb#use-with-subgraphs",
    "how-tos/cross-thread-persistence.ipynb": "how-tos/persistence.ipynb#add-long-term-memory",
    "cloud/how-tos/copy_threads": "cloud/how-tos/use_threads",
+    "cloud/concepts/threads.md": "concepts/persistence.md#threads",
    # tool calling how-tos
    "how-tos/tool-calling-errors.ipynb": "how-tos/tool-calling.ipynb#handle-errors",
    "how-tos/pass-config-to-tools.ipynb": "how-tos/tool-calling.ipynb#access-config",
@@ -86,6 +87,8 @@ REDIRECT_MAP = {
    "cloud/how-tos/stream_events.md": "cloud/how-tos/streaming.md#stream-events",
    "cloud/how-tos/stream_debug.md": "cloud/how-tos/streaming.md#debug",
    "cloud/how-tos/stream_multiple.md": "cloud/how-tos/streaming.md#stream-multiple-modes",
+    "cloud/concepts/streaming.md": "concepts/streaming.md",
+    "agents/streaming.md": "how-tos/streaming.md",
    # prebuit redirects
    "how-tos/create-react-agent.ipynb": "agents/agents.md#basic-configuration",
    "how-tos/create-react-agent-memory.ipynb": "agents/memory.md",
@@ -107,8 +110,10 @@ REDIRECT_MAP = {
    # deployment redirects
    "how-tos/deploy-self-hosted.md": "cloud/deployment/self_hosted_data_plane.md",
    "concepts/self_hosted.md": "concepts/langgraph_self_hosted_data_plane.md",
+    "tutorials/deployment.md": "concepts/deployment_options.md",
    # assistant redirects
    "cloud/how-tos/assistant_versioning.md": "cloud/how-tos/configuration_cloud.md",
+    "cloud/concepts/runs.md": "concepts/assistants.md#execution",
 }


@@ -89,4 +89,4 @@ LangGraph Studio Web is a specialized UI that you can connect to LangGraph API s

 ## Deployment

-Once your LangGraph app is running locally, you can deploy it using LangGraph Platform. Refer to the [deployment options guide](../tutorials/deployment.md) for detailed instructions on all supported deployment models.
+Once your LangGraph app is running locally, you can deploy it using LangGraph Platform. Refer to the [deployment options guide](../concepts/deployment_options.md) for detailed instructions on all supported deployment models.
@@ -29,10 +29,10 @@ LangGraph includes several capabilities essential for building robust, productio

 - [**Memory integration**](./memory.md): Native support for *short-term* (session-based) and *long-term* (persistent across sessions) memory, enabling stateful behaviors in chatbots and assistants.
 - [**Human-in-the-loop control**](./human-in-the-loop.md): Execution can pause *indefinitely* to await human feedback—unlike websocket-based solutions limited to real-time interaction. This enables asynchronous approval, correction, or intervention at any point in the workflow.
- [**Streaming support**](./streaming.md): Real-time streaming of agent state, model tokens, tool outputs, or combined streams.
+- [**Streaming support**](../how-tos/streaming.md): Real-time streaming of agent state, model tokens, tool outputs, or combined streams.
 - [**Deployment tooling**](./deployment.md): Includes infrastructure-free deployment tools. [**LangGraph Platform**](https://langchain-ai.github.io/langgraph/concepts/langgraph_platform/) supports testing, debugging, and deployment.
    - **[Studio](https://langchain-ai.github.io/langgraph/concepts/langgraph_studio/)**: A visual IDE for inspecting and debugging workflows.
-    - Supports multiple [**deployment options**](https://langchain-ai.github.io/langgraph/tutorials/deployment/) for production.
+    - Supports multiple [**deployment options**](https://langchain-ai.github.io/langgraph/concepts/deployment_options.md) for production.

 ## High-level building blocks

@@ -109,7 +109,7 @@ Streaming is available in both sync and async modes:

 !!! tip

-    For full details, see the [streaming guide](./streaming.md).
+    For full details, see the [streaming guide](../how-tos/streaming.md).

 ## Max iterations

@@ -1,223 +0,0 @@
---
-search:
-  boost: 2
-tags:
-  - agent
-hide:
-  - tags
---
-
-# Streaming
-
-Streaming is key to building responsive applications. There are a few types of data you’ll want to stream:
-
-1. [**Agent progress**](#agent-progress) — get updates after each node in the agent graph is executed.
-2. [**LLM tokens**](#llm-tokens) — stream tokens as they are generated by the language model.
-3. [**Custom updates**](#tool-updates) — emit custom data from tools during execution (e.g., "Fetched 10/100 records")
-
-You can stream [more than one type of data](#stream-multiple-modes) at a time. 
-
-
-<figure markdown="1">
-![image](./assets/fast_parrot.png){: style="max-height:300px"}
-<figcaption>
-Waiting is for pigeons.
-</figcaption>
-</figure>
-
-## Agent progress
-
-To stream agent progress, use the [`stream()`][langgraph.graph.state.CompiledStateGraph.stream] or [`astream()`][langgraph.graph.state.CompiledStateGraph.astream] methods with [`stream_mode="updates"`](https://langchain-ai.github.io/langgraph/how-tos/streaming/#updates). This emits an event after every agent step.
-
-For example, if you have an agent that calls a tool once, you should see the following updates:
-
-* **LLM node**: AI message with tool call requests
-* **Tool node**: Tool message with execution result
-* **LLM node**: Final AI response
-
-=== "Sync"
-
-    ```python
-    agent = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=[get_weather],
-    )
-    # highlight-next-line
-    for chunk in agent.stream(
-        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-        # highlight-next-line
-        stream_mode="updates"
-    ):
-        print(chunk)
-        print("\n")
-    ```
-
-=== "Async"
-
-    ```python
-    agent = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=[get_weather],
-    )
-    # highlight-next-line
-    async for chunk in agent.astream(
-        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-        # highlight-next-line
-        stream_mode="updates"
-    ):
-        print(chunk)
-        print("\n")
-    ```
-
-## LLM tokens
-
-To stream tokens as they are produced by the LLM, use `stream_mode="messages"`:
-
-=== "Sync"
-
-    ```python
-    agent = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=[get_weather],
-    )
-    # highlight-next-line
-    for token, metadata in agent.stream(
-        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-        # highlight-next-line
-        stream_mode="messages"
-    ):
-        print("Token", token)
-        print("Metadata", metadata)
-        print("\n")
-    ```
-
-=== "Async"
-
-    ```python
-    agent = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=[get_weather],
-    )
-    # highlight-next-line
-    async for token, metadata in agent.astream(
-        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-        # highlight-next-line
-        stream_mode="messages"
-    ):
-        print("Token", token)
-        print("Metadata", metadata)
-        print("\n")
-    ```
-
-## Tool updates
-
-To stream updates from tools as they are executed, you can use [get_stream_writer][langgraph.config.get_stream_writer].
-
-=== "Sync"
-
-    ```python
-    # highlight-next-line
-    from langgraph.config import get_stream_writer
-
-    def get_weather(city: str) -> str:
-        """Get weather for a given city."""
-        # highlight-next-line
-        writer = get_stream_writer()
-        # stream any arbitrary data
-        # highlight-next-line
-        writer(f"Looking up data for city: {city}")
-        return f"It's always sunny in {city}!"
-
-    agent = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=[get_weather],
-    )
-
-    for chunk in agent.stream(
-        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-        # highlight-next-line
-        stream_mode="custom"
-    ):
-        print(chunk)
-        print("\n")
-    ```
-
-=== "Async"
-
-    ```python
-    # highlight-next-line
-    from langgraph.config import get_stream_writer
-
-    def get_weather(city: str) -> str:
-        """Get weather for a given city."""
-        # highlight-next-line
-        writer = get_stream_writer()
-        # stream any arbitrary data
-        # highlight-next-line
-        writer(f"Looking up data for city: {city}")
-        return f"It's always sunny in {city}!"
-
-    agent = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=[get_weather],
-    )
-
-    async for chunk in agent.astream(
-        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-        # highlight-next-line
-        stream_mode="custom"
-    ):
-        print(chunk)
-        print("\n")
-    ```
-
-!!! Note
-    If you add `get_stream_writer` inside your tool, you won't be able to invoke the tool outside of a LangGraph execution context. 
-
-## Stream multiple modes
-
-You can specify multiple streaming modes by passing stream mode as a list: `stream_mode=["updates", "messages", "custom"]`:
-
-=== "Sync"
-
-    ```python
-    agent = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=[get_weather],
-    )
-
-    for stream_mode, chunk in agent.stream(
-        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-        # highlight-next-line
-        stream_mode=["updates", "messages", "custom"]
-    ):
-        print(chunk)
-        print("\n")
-    ```
-
-=== "Async"
-
-    ```python
-    agent = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=[get_weather],
-    )
-
-    async for stream_mode, chunk in agent.astream(
-        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-        # highlight-next-line
-        stream_mode=["updates", "messages", "custom"]
-    ):
-        print(chunk)
-        print("\n")
-    ```
-
-## Disable streaming
-
-In some applications you might need to disable streaming of individual tokens for a given model. This is useful in [multi-agent](./multi-agent.md) systems to control which agents stream their output.
-
-See the [Models](./models.md#disable-streaming) guide to learn how to disable streaming.
-
-## Additional resources
-
-* [Streaming in LangGraph](https://langchain-ai.github.io/langgraph/how-tos/streaming)
@@ -1,310 +1 @@
---
-search:
-  boost: 2
-tags:
-  - agent
-hide:
-  - tags
---
-
-# Tools
-
-[Tools](https://python.langchain.com/docs/concepts/tools/) are a way to encapsulate a function and its input schema in a way that can be passed to a chat model that supports tool calling. This allows the model to request the execution of this function with specific inputs.
-
-You can either [define your own tools](#define-simple-tools) or use [prebuilt integrations](#prebuilt-tools) that LangChain provides.
-
-## Define simple tools
-
-You can pass a vanilla function to `create_react_agent` to use as a tool:
-
-```python
-from langgraph.prebuilt import create_react_agent
-
-def multiply(a: int, b: int) -> int:
-    """Multiply two numbers."""
-    return a * b
-
-create_react_agent(
-    model="anthropic:claude-3-7-sonnet",
-    tools=[multiply]
-)
-```
-
-`create_react_agent` automatically converts vanilla functions to [LangChain tools](https://python.langchain.com/docs/concepts/tools/#tool-interface).
-
-## Customize tools
-
-For more control over tool behavior, use the `@tool` decorator:
-
-```python
-# highlight-next-line
-from langchain_core.tools import tool
-
-# highlight-next-line
-@tool("multiply_tool", parse_docstring=True)
-def multiply(a: int, b: int) -> int:
-    """Multiply two numbers.
-
-    Args:
-        a: First operand
-        b: Second operand
-    """
-    return a * b
-```
-
-You can also define a custom input schema using Pydantic:
-
-```python
-from pydantic import BaseModel, Field
-
-class MultiplyInputSchema(BaseModel):
-    """Multiply two numbers"""
-    a: int = Field(description="First operand")
-    b: int = Field(description="Second operand")
-
-# highlight-next-line
-@tool("multiply_tool", args_schema=MultiplyInputSchema)
-def multiply(a: int, b: int) -> int:
-    return a * b
-```
-
-For additional customization, refer to the [custom tools guide](https://python.langchain.com/docs/how_to/custom_tools/).
-
-## Hide arguments from the model
-
-Some tools require runtime-only arguments (e.g., user ID or session context) that should not be controllable by the model.
-
-You can put these arguments in the `state` or `config` of the agent, and access
-this information inside the tool:
-
-```python
-from langgraph.prebuilt import InjectedState
-from langgraph.prebuilt.chat_agent_executor import AgentState
-from langchain_core.runnables import RunnableConfig
-
-def my_tool(
-    # This will be populated by an LLM
-    tool_arg: str,
-    # access information that's dynamically updated inside the agent
-    # highlight-next-line
-    state: Annotated[AgentState, InjectedState],
-    # access static data that is passed at agent invocation
-    # highlight-next-line
-    config: RunnableConfig,
-) -> str:
-    """My tool."""
-    do_something_with_state(state["messages"])
-    do_something_with_config(config)
-    ...
-```
-
-## Disable parallel tool calling
-
-Some model providers support executing multiple tools in parallel, but
-allow users to disable this feature.
-
-For supported providers, you can disable parallel tool calling by setting `parallel_tool_calls=False` via the `model.bind_tools()` method:
-
-```python
-from langchain.chat_models import init_chat_model
-
-def add(a: int, b: int) -> int:
-    """Add two numbers"""
-    return a + b
-
-def multiply(a: int, b: int) -> int:
-    """Multiply two numbers."""
-    return a * b
-
-model = init_chat_model("anthropic:claude-3-5-sonnet-latest", temperature=0)
-tools = [add, multiply]
-agent = create_react_agent(
-    # disable parallel tool calls
-    # highlight-next-line
-    model=model.bind_tools(tools, parallel_tool_calls=False),
-    tools=tools
-)
-
-agent.invoke(
-    {"messages": [{"role": "user", "content": "what's 3 + 5 and 4 * 7?"}]}
-)
-```
-
-## Return tool results directly
-
-Use `return_direct=True` to return tool results immediately and stop the agent loop:
-
-```python
-from langchain_core.tools import tool
-
-# highlight-next-line
-@tool(return_direct=True)
-def add(a: int, b: int) -> int:
-    """Add two numbers"""
-    return a + b
-
-agent = create_react_agent(
-    model="anthropic:claude-3-7-sonnet-latest",
-    tools=[add]
-)
-
-agent.invoke(
-    {"messages": [{"role": "user", "content": "what's 3 + 5?"}]}
-)
-```
-
-## Force tool use
-
-To force the agent to use specific tools, you can set the `tool_choice` option in `model.bind_tools()`:
-
-```python
-from langchain_core.tools import tool
-
-# highlight-next-line
-@tool(return_direct=True)
-def greet(user_name: str) -> int:
-    """Greet user."""
-    return f"Hello {user_name}!"
-
-tools = [greet]
-
-agent = create_react_agent(
-    # highlight-next-line
-    model=model.bind_tools(tools, tool_choice={"type": "tool", "name": "greet"}),
-    tools=tools
-)
-
-agent.invoke(
-    {"messages": [{"role": "user", "content": "Hi, I am Bob"}]}
-)
-```
-
-!!! Warning "Avoid infinite loops"
-
-    Forcing tool usage without stopping conditions can create infinite loops. Use one of the following safeguards:
-
-    - Mark the tool with [`return_direct=True`](#return-tool-results-directly) to end the loop after execution.
-    - Set [`recursion_limit`](../concepts/low_level.md#recursion-limit) to restrict the number of execution steps.
-
-## Handle tool errors
-
-By default, the agent will catch all exceptions raised during tool calls and will pass those as tool messages to the LLM. To control how the errors are handled, you can use the prebuilt [`ToolNode`][langgraph.prebuilt.tool_node.ToolNode] — the node that executes tools inside `create_react_agent` — via its `handle_tool_errors` parameter:
-
-=== "Enable error handling (default)"
-
-    ```python
-    from langgraph.prebuilt import create_react_agent
-
-    def multiply(a: int, b: int) -> int:
-        """Multiply two numbers."""
-        if a == 42:
-            raise ValueError("The ultimate error")
-        return a * b
-
-    # Run with error handling (default)
-    agent = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=[multiply]
-    )
-    agent.invoke(
-        {"messages": [{"role": "user", "content": "what's 42 x 7?"}]}
-    )
-    ```
-
-=== "Disable error handling"
-
-    ```python
-    from langgraph.prebuilt import create_react_agent, ToolNode
-
-    def multiply(a: int, b: int) -> int:
-        """Multiply two numbers."""
-        if a == 42:
-            raise ValueError("The ultimate error")
-        return a * b
-
-    # highlight-next-line
-    tool_node = ToolNode(
-        [multiply],
-        # highlight-next-line
-        handle_tool_errors=False  # (1)!
-    )
-    agent_no_error_handling = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=tool_node
-    )
-    agent_no_error_handling.invoke(
-        {"messages": [{"role": "user", "content": "what's 42 x 7?"}]}
-    )
-    ```
-
-    1. This disables error handling (enabled by default). See all available strategies in the [API reference][langgraph.prebuilt.tool_node.ToolNode].
-
-=== "Custom error handling"
-
-    ```python
-    from langgraph.prebuilt import create_react_agent, ToolNode
-
-    def multiply(a: int, b: int) -> int:
-        """Multiply two numbers."""
-        if a == 42:
-            raise ValueError("The ultimate error")
-        return a * b
-
-    # highlight-next-line
-    tool_node = ToolNode(
-        [multiply],
-        # highlight-next-line
-        handle_tool_errors=(
-            "Can't use 42 as a first operand, you must switch operands!"  # (1)!
-        )
-    )
-    agent_custom_error_handling = create_react_agent(
-        model="anthropic:claude-3-7-sonnet-latest",
-        tools=tool_node
-    )
-    agent_custom_error_handling.invoke(
-        {"messages": [{"role": "user", "content": "what's 42 x 7?"}]}
-    )
-    ```
-
-    1. This provides a custom message to send to the LLM in case of an exception. See all available strategies in the [API reference][langgraph.prebuilt.tool_node.ToolNode].
-
-See [API reference][langgraph.prebuilt.tool_node.ToolNode] for more information on different tool error handling options.
-
-## Working with memory
-
-LangGraph allows access to short-term and long-term memory from tools. See [Memory](./memory.md) guide for more information on:
-
-* how to [read](./memory.md#read-short-term) from and [write](./memory.md#write-short-term) to **short-term** memory
-* how to [read](./memory.md#read-long-term) from and [write](./memory.md#write-long-term) to **long-term** memory
-
-## Prebuilt tools
-
-You can use prebuilt tools from model providers by passing a dictionary with tool specs to the `tools` parameter of `create_react_agent`. For example, to use the `web_search_preview` tool from OpenAI:
-
-```python
-from langgraph.prebuilt import create_react_agent
-
-agent = create_react_agent(
-    model="openai:gpt-4o-mini", 
-    tools=[{"type": "web_search_preview"}]
-)
-response = agent.invoke(
-    {"messages": ["What was a positive news story from today?"]}
-)
-```
-
-Additionally, LangChain supports a wide range of prebuilt tool integrations for interacting with APIs, databases, file systems, web data, and more. These tools extend the functionality of agents and enable rapid development.
-
-You can browse the full list of available integrations in the [LangChain integrations directory](https://python.langchain.com/docs/integrations/tools/).
-
-Some commonly used tool categories include:
-
- **Search**: Bing, SerpAPI, Tavily
- **Code interpreters**: Python REPL, Node.js REPL
- **Databases**: SQL, MongoDB, Redis
- **Web data**: Web scraping and browsing
- **APIs**: OpenWeatherMap, NewsAPI, and others
-
-These integrations can be configured and added to your agents using the same `tools` parameter shown in the examples above.
-
+delete me
@@ -1,5 +0,0 @@
-# Runs
-
-A run is an invocation of an [assistant](../../concepts/assistants.md). Each run may have its own input, configuration, and metadata, which may affect execution and output of the underlying graph. A run can optionally be executed on a [thread](./threads.md).
-
-The LangGraph Platform API provides several endpoints for creating and managing runs. See the [API reference](../../cloud/reference/api/api_ref.html#tag/thread-runs/) for more details.
@@ -1,138 +0,0 @@
-# Streaming
-
-Streaming is critical for making LLM applications feel responsive to end users.  
-When creating a streaming run, the **streaming mode** determines what kinds of data are streamed back to the API client.
-
-## Supported streaming modes
-
-LangGraph Platform supports the following streaming modes:
-
-| Mode                 | Description                                                                                                                                                    | LangGraph Library Method                                                |
-|----------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------|
-| **`values`**         | Stream the full graph state after each [super-step](https://langchain-ai.github.io/langgraph/concepts/low_level/#graphs). [Guide](../how-tos/streaming.md#stream-graph-state) | `.stream()` / `.astream()` with `stream_mode="values"`                  |
-| **`updates`**        | Stream only the updates to the graph state after each node. [Guide](../how-tos/streaming.md#stream-graph-state)                                                              | `.stream()` / `.astream()` with `stream_mode="updates"`                 |
-| **`messages-tuple`** | Stream LLM tokens for any messages generated inside the graph (useful for chat apps). [Guide](../how-tos/streaming.md#messages)                                   | `.stream()` / `.astream()` with `stream_mode="messages"`                |
-| **`debug`**          | Stream debug information throughout graph execution. [Guide](../how-tos/streaming.md#debug)                                                                       | `.stream()` / `.astream()` with `stream_mode="debug"`                   |
-| **`custom`**         | Stream custom data. [Guide](../../how-tos/streaming.md#stream-custom-data)                                                                                                                                             | `.stream()` / `.astream()` with `stream_mode="custom"`                  |
-| **`events`**         | Stream all events (including the state of the graph); mainly useful when migrating large LCEL apps. [Guide](../how-tos/streaming.md#stream-events)                       | `.astream_events()`                                                     |
-
-✅ You can also **combine multiple modes** at the same time.  See the [how-to guide](../how-tos/streaming.md#stream-multiple-modes) for configuration details.
-
-## Stateless runs
-
-If you don't want to **persist the outputs** of a streaming run in the [checkpointer](../../concepts/persistence.md) DB, you can create a stateless run without creating a thread:
-
-=== "Python"
-
-    ```python
-    from langgraph_sdk import get_client
-    client = get_client(url=<DEPLOYMENT_URL>, api_key=<API_KEY>)
-
-    async for chunk in client.runs.stream(
-        # highlight-next-line
-        None,  # (1)!
-        assistant_id,
-        input=inputs,
-        stream_mode="updates"
-    ):
-        print(chunk.data)
-    ```
-
-    1. We are passing `None` instead of a `thread_id` UUID.
-
-=== "JavaScript"
-
-    ```js
-    import { Client } from "@langchain/langgraph-sdk";
-    const client = new Client({ apiUrl: <DEPLOYMENT_URL>, apiKey: <API_KEY> });
-
-    // create a streaming run
-    // highlight-next-line
-    const streamResponse = client.runs.stream(
-      // highlight-next-line
-      null,  // (1)!
-      assistantID,
-      {
-        input,
-        streamMode: "updates"
-      }
-    );
-    for await (const chunk of streamResponse) {
-      console.log(chunk.data);
-    }
-    ```
-
-    1. We are passing `None` instead of a `thread_id` UUID.
-
-=== "cURL"
-
-    ```bash
-    curl --request POST \
-    --url <DEPLOYMENT_URL>/runs/stream \
-    --header 'Content-Type: application/json' \
-    --header 'x-api-key: <API_KEY>'
-    --data "{
-      \"assistant_id\": \"agent\",
-      \"input\": <inputs>,
-      \"stream_mode\": \"updates\"
-    }"
-    ```
-
-## Join and stream
-
-LangGraph Platform allows you to join an active [background run](../how-tos/background_run.md) and stream outputs from it. To do so, you can use [LangGraph SDK's](https://langchain-ai.github.io/langgraph/cloud/reference/sdk/python_sdk_ref/) `client.runs.join_stream` method:
-
-=== "Python"
-
-    ```python
-    from langgraph_sdk import get_client
-    client = get_client(url=<DEPLOYMENT_URL>, api_key=<API_KEY>)
-
-    # highlight-next-line
-    async for chunk in client.runs.join_stream(
-        thread_id,
-        # highlight-next-line
-        run_id,  # (1)!
-    ):
-        print(chunk)
-    ```
-
-    1. This is the `run_id` of an existing run you want to join.
-
-
-=== "JavaScript"
-
-    ```js
-    import { Client } from "@langchain/langgraph-sdk";
-    const client = new Client({ apiUrl: <DEPLOYMENT_URL>, apiKey: <API_KEY> });
-
-    // highlight-next-line
-    const streamResponse = client.runs.joinStream(
-      threadID,
-      // highlight-next-line
-      runId  // (1)!
-    );
-    for await (const chunk of streamResponse) {
-      console.log(chunk);
-    }
-    ```
-
-    1. This is the `run_id` of an existing run you want to join.
-
-=== "cURL"
-
-    ```bash
-    curl --request GET \
-    --url <DEPLOYMENT_URL>/threads/<THREAD_ID>/runs/<RUN_ID>/stream \
-    --header 'Content-Type: application/json' \
-    --header 'x-api-key: <API_KEY>'
-    ```
-
-!!! warning "Outputs not buffered"
-
-    When you use `.join_stream`, output is not buffered, so any output produced before joining will not be received.
-
-## API Reference
-
-For API usage and implementation, refer to the [API reference](../reference/api/api_ref.html#tag/thread-runs/POST/threads/{thread_id}/runs/stream). 
-
@@ -62,6 +62,15 @@ Starting from the `LangGraph Platform` view...
 1. In the panel, select the `Server` tab to view server logs for the revision. Server logs are only available after a revision has been deployed.
 1. Within the `Server` tab, adjust the date/time range picker as needed. By default, the date/time range picker is set to the `Last 7 days`.

+## View Deployment Metrics
+
+Starting from the <a href="https://smith.langchain.com/" target="_blank">LangSmith UI</a>...
+
+1. In the left-hand navigation panel, select `LangGraph Platform`. The `LangGraph Platform` view contains a list of existing LangGraph Platform deployments.
+1. Select an existing deployment to monitor.
+1. Select the `Monitoring` tab to view the deployment metrics. See a list of [all available metrics](../../concepts/langgraph_control_plane.md#monitoring).
+1. Within the `Monitoring` tab, use the date/time range picker as needed. By default, the date/time range picker is set to the `Last 15 minutes`.
+
 ## Interrupt Revision

 Interrupting a revision will stop deployment of the revision.
@@ -20,7 +20,7 @@ my-app/
 |-- openai_agent.py     # code for your graph
 ```

-where the graph is defined in `openai_agent.py`.
+where the graph is defined in `openai_agent.py`. 

 ### No rebuild

@@ -28,11 +28,11 @@ In the standard LangGraph API configuration, the server uses the compiled graph

 ```python
 from langchain_openai import ChatOpenAI
-from langgraph.graph import END, START, StateGraph, MessagesState
+from langgraph.graph import END, START, MessageGraph

 model = ChatOpenAI(temperature=0)

-graph_workflow = StateGraph(MessagesState)
+graph_workflow = MessageGraph()

 graph_workflow.add_node("agent", model)
 graph_workflow.add_edge("agent", END)
@@ -61,7 +61,7 @@ To make your graph rebuild on each new run with custom configuration, you need t
 from typing import Annotated
 from typing_extensions import TypedDict
 from langchain_openai import ChatOpenAI
-from langgraph.graph import END, START
+from langgraph.graph import END, START, MessageGraph
 from langgraph.graph.state import StateGraph
 from langgraph.graph.message import add_messages
 from langgraph.prebuilt import ToolNode
@@ -144,4 +144,4 @@ Finally, you need to specify the path to your graph-making function (`make_graph
 }
 ```

-See more info on LangGraph API configuration file [here](../reference/cli.md#configuration-file)
+See more info on LangGraph API configuration file [here](../reference/cli.md#configuration-file)
@@ -33,7 +33,7 @@ For more information on breakpoints see [here](../../concepts/breakpoints.md).

 ### Submit run

-To submit the run with the specified input and run settings, click the "Submit" button. This will add a [run](../concepts/runs.md) to the existing selected [thread](../concepts/threads.md). If no thread is currently selected, a new one will be created.
+To submit the run with the specified input and run settings, click the "Submit" button. This will add a [run](../concepts/runs.md) to the existing selected [thread](../../concepts/persistence.md#threads). If no thread is currently selected, a new one will be created.

 To cancel the ongoing run, click the "Cancel" button.

@@ -1,8 +1,12 @@
-# Stream outputs
+# Streaming API

-## Streaming API
+[LangGraph SDK](https://langchain-ai.github.io/langgraph/cloud/reference/sdk/python_sdk_ref/) allows you to [stream outputs](../../concepts/streaming.md) from the LangGraph API server.

-[LangGraph SDK](https://langchain-ai.github.io/langgraph/cloud/reference/sdk/python_sdk_ref/) allows you to stream outputs from the LangGraph API server.
+!!! note
+
+    LangGraph SDK and LangGraph Server are a part of [LangGraph Platform](../../concepts/langgraph_platform.md).
+
+## Basic usage

 Basic usage example:

@@ -833,3 +837,121 @@ To stream all events, including the state of the graph:
      \"stream_mode\": \"events\"
    }"
    ```
+
+## Stateless runs
+
+If you don't want to **persist the outputs** of a streaming run in the [checkpointer](../../concepts/persistence.md) DB, you can create a stateless run without creating a thread:
+
+=== "Python"
+
+    ```python
+    from langgraph_sdk import get_client
+    client = get_client(url=<DEPLOYMENT_URL>, api_key=<API_KEY>)
+
+    async for chunk in client.runs.stream(
+        # highlight-next-line
+        None,  # (1)!
+        assistant_id,
+        input=inputs,
+        stream_mode="updates"
+    ):
+        print(chunk.data)
+    ```
+
+    1. We are passing `None` instead of a `thread_id` UUID.
+
+=== "JavaScript"
+
+    ```js
+    import { Client } from "@langchain/langgraph-sdk";
+    const client = new Client({ apiUrl: <DEPLOYMENT_URL>, apiKey: <API_KEY> });
+
+    // create a streaming run
+    // highlight-next-line
+    const streamResponse = client.runs.stream(
+      // highlight-next-line
+      null,  // (1)!
+      assistantID,
+      {
+        input,
+        streamMode: "updates"
+      }
+    );
+    for await (const chunk of streamResponse) {
+      console.log(chunk.data);
+    }
+    ```
+
+    1. We are passing `None` instead of a `thread_id` UUID.
+
+=== "cURL"
+
+    ```bash
+    curl --request POST \
+    --url <DEPLOYMENT_URL>/runs/stream \
+    --header 'Content-Type: application/json' \
+    --header 'x-api-key: <API_KEY>'
+    --data "{
+      \"assistant_id\": \"agent\",
+      \"input\": <inputs>,
+      \"stream_mode\": \"updates\"
+    }"
+    ```
+
+## Join and stream
+
+LangGraph Platform allows you to join an active [background run](../how-tos/background_run.md) and stream outputs from it. To do so, you can use [LangGraph SDK's](https://langchain-ai.github.io/langgraph/cloud/reference/sdk/python_sdk_ref/) `client.runs.join_stream` method:
+
+=== "Python"
+
+    ```python
+    from langgraph_sdk import get_client
+    client = get_client(url=<DEPLOYMENT_URL>, api_key=<API_KEY>)
+
+    # highlight-next-line
+    async for chunk in client.runs.join_stream(
+        thread_id,
+        # highlight-next-line
+        run_id,  # (1)!
+    ):
+        print(chunk)
+    ```
+
+    1. This is the `run_id` of an existing run you want to join.
+
+
+=== "JavaScript"
+
+    ```js
+    import { Client } from "@langchain/langgraph-sdk";
+    const client = new Client({ apiUrl: <DEPLOYMENT_URL>, apiKey: <API_KEY> });
+
+    // highlight-next-line
+    const streamResponse = client.runs.joinStream(
+      threadID,
+      // highlight-next-line
+      runId  // (1)!
+    );
+    for await (const chunk of streamResponse) {
+      console.log(chunk);
+    }
+    ```
+
+    1. This is the `run_id` of an existing run you want to join.
+
+=== "cURL"
+
+    ```bash
+    curl --request GET \
+    --url <DEPLOYMENT_URL>/threads/<THREAD_ID>/runs/<RUN_ID>/stream \
+    --header 'Content-Type: application/json' \
+    --header 'x-api-key: <API_KEY>'
+    ```
+
+!!! warning "Outputs not buffered"
+
+    When you use `.join_stream`, output is not buffered, so any output produced before joining will not be received.
+
+## API Reference
+
+For API usage and implementation, refer to the [API reference](../reference/api/api_ref.html#tag/thread-runs/POST/threads/{thread_id}/runs/stream). 
@@ -13,7 +13,7 @@ LangGraph Studio is accessed from the LangSmith UI, within the LangGraph Platfor

 For applications that are [deployed](../../quick_start.md) on LangGraph Platform, you can access Studio as part of that deployment. To do so, navigate to the deployment in LangGraph Platform within the LangSmith UI and click the "LangGraph Studio" button.

-This will load the Studio UI connected to your live deployment, allowing you to create, read, and update the [threads](../../concepts/threads.md), [assistants](../../../concepts/assistants.md), and [memory](../../../concepts//memory.md) in that deployment.
+This will load the Studio UI connected to your live deployment, allowing you to create, read, and update the [threads](../../../concepts/persistence.md#threads), [assistants](../../../concepts/assistants.md), and [memory](../../../concepts//memory.md) in that deployment.

 ## Local development server

@@ -0,0 +1,57 @@
+# Run experiments over a dataset
+
+LangGraph Studio supports evaluations by allowing you to run your assistant over a pre-defined LangSmith dataset. This enables you to understand how your application performs over a variety of inputs, compare the results to reference outputs, and score the results using [evaluators](../../../agents/evals.md).
+
+This guide shows you how to run an experiment end-to-end from Studio.
+
+---
+
+## Prerequisites
+
+Before running an experiment, ensure you have the following:
+
+1.  **A LangSmith dataset**: Your dataset should contain the inputs you want to test and optionally, reference outputs for comparison.
+
+    - The schema for the inputs must match the required input schema for the assistant. For more information on schemas, see [here](../../../concepts/low_level.md#schema).
+    - For more on creating datasets, see [How to Manage Datasets](https://docs.smith.langchain.com/evaluation/how_to_guides/manage_datasets_in_application#set-up-your-dataset).
+
+2.  **(Optional) Evaluators**: You can attach evaluators (e.g., LLM-as-a-Judge, heuristics, or custom functions) to your dataset in LangSmith. These will run automatically after the graph has processed all inputs.
+
+    - To learn more, read about [Evaluation Concepts](https://docs.smith.langchain.com/evaluation/concepts#evaluators).
+
+3.  **A running application**: The experiment can be run against:
+    - An application deployed on [LangGraph Platform](../../quick_start.md).
+    - A locally running application started via the [langgraph-cli](../../../tutorials/langgraph-platform/local-server.md).
+
+---
+
+## Step-by-step guide
+
+### 1. Launch the experiment
+
+Click the **Run experiment** button in the top right corner of the Studio page.
+
+### 2. Select your dataset
+
+In the modal that appears, select the dataset (or a specific dataset split) to use for the experiment and click **Start**.
+
+### 3. Monitor the progress
+
+All of the inputs in the dataset will now be run against the active assistant. Monitor the experiment's progress via the badge in the top right corner.
+
+You can continue to work in Studio while the experiment runs in the background. Click the arrow icon button at any time to navigate to LangSmith and view the detailed experiment results.
+
+---
+
+## Troubleshooting
+
+### "Run experiment" button is disabled
+
+If the "Run experiment" button is disabled, check the following:
+
+- **Deployed application**: If your application is deployed on LangGraph Platform, you may need to create a new revision to enable this feature.
+- **Local development server**: If you are running your application locally, make sure you have upgraded to the latest version of the `langgraph-cli` (`pip install -U langgraph-cli`). Additionally, ensure you have tracing enabled by setting the `LANGSMITH_API_KEY` in your project's `.env` file.
+
+### Evaluator results are missing
+
+When you run an experiment, any attached evaluators are scheduled for execution in a queue. If you don't see results immediately, it likely means they are still pending.
@@ -1,10 +1,6 @@
 # Manage threads

-!!! info "Prerequisites"
-
-    - [Threads Overview](../concepts/threads.md)
-
-Studio allows you to view threads from the server and edit their state.
+Studio allows you to view [threads](../../concepts/persistence.md#threads) from the server and edit their state.

 ## View threads

@@ -1,10 +1,6 @@
 # Use threads

-!!! info "Prerequisites"
-
-    - [Threads Overview](../concepts/threads.md)
-
-In this guide, we will show how to create, view, and inspect threads.
+In this guide, we will show how to create, view, and inspect [threads](../../concepts/persistence.md#threads).

 ## Create a thread

@@ -1,29 +1,31 @@
 # Assistants

-!!! info "Prerequisites"
+**Assistants** allow you to manage configurations (like prompts, LLM selection, tools) separately from your graph's core logic, enabling rapid changes that don't alter the graph architecture. It is a way to create multiple specialized versions of the same graph architecture, each optimized for different use cases through configuration variations rather than structural changes.

-    - [LangGraph Server](./langgraph_server.md)
-    - [Configuration](./low_level.md#configuration)
-
-When building agents, it is common to make rapid changes that _do not_ alter the graph logic. For example, simply changing prompts or the LLM selection can have significant impacts on the behavior of the agent but does not require updating your graph's architecture. Assistants offer a straightforward way to manage these configurations separately from your graph's core logic.
-
-Imagine a general-purpose writing agent built on a common graph architecture. While the structure remains the same, different writing styles—such as blog posts and tweets—require tailored configurations to optimize performance. To support these variations, you can create multiple assistants (e.g., one for blogs and another for tweets) that share the underlying graph but differ in model selection and system prompt.
+For example, imagine a general-purpose writing agent built on a common graph architecture. While the structure remains the same, different writing styles—such as blog posts and tweets—require tailored configurations to optimize performance. To support these variations, you can create multiple assistants (e.g., one for blogs and another for tweets) that share the underlying graph but differ in model selection and system prompt.

 ![assistant versions](img/assistants.png)

-## Configuring assistants
+The LangGraph Cloud API provides several endpoints for creating and managing assistants and their versions. See the [API reference](../cloud/reference/api/api_ref.html#tag/assistants) for more details.
+
+!!! info
+
+    Assistants are a [LangGraph Platform](langgraph_platform.md) concept. They are not available in the open source LangGraph library.
+
+## Configuration

 Assistants build on the LangGraph open source concept of [configuration](low_level.md#configuration).
-While configuration is available in the open source LangGraph library, assistants are only present in [LangGraph Platform](langgraph_platform.md).
-This is due to the fact that assistants are tightly coupled to your deployed graph. Upon deployment, LangGraph Server will automatically create a default assistant for each graph using the graph's default configuration settings.
+While configuration is available in the open source LangGraph library, assistants are only present in [LangGraph Platform](langgraph_platform.md). This is due to the fact that assistants are tightly coupled to your deployed graph. Upon deployment, LangGraph Server will automatically create a default assistant for each graph using the graph's default configuration settings.

 In practice, an assistant is just an _instance_ of a graph with a specific configuration. Therefore, multiple assistants can reference the same graph but can contain different configurations (e.g. prompts, models, tools). The LangGraph Server API provides several endpoints for creating and managing assistants. See the [API reference](../cloud/reference/api/api_ref.html) and [this how-to](../cloud/how-tos/configuration_cloud.md) for more details on how to create assistants.

-## Versioning assistants
+## Versioning

 Assistants support versioning to track changes over time.
 Once you've created an assistant, subsequent edits to that assistant will create new versions. See [this how-to](../cloud/how-tos/configuration_cloud.md#create-a-new-version-for-your-assistant) for more details on how to manage assistant versions.

-## Learn more
+## Execution

-* The LangGraph Cloud API provides several endpoints for creating and managing assistants and their versions. See the [API reference](../cloud/reference/api/api_ref.html#tag/assistants) for more details.
+A **run** is an invocation of an assistant. Each run may have its own input, configuration, and metadata, which may affect execution and output of the underlying graph. A run can optionally be executed on a [thread](../../concepts/persistence.md#threads).
+
+The LangGraph Platform API provides several endpoints for creating and managing runs. See the [API reference](../../cloud/reference/api/api_ref.html#tag/thread-runs/) for more details.
@@ -5,7 +5,16 @@ search:

 # Deployment Options

-There are 4 main options for deploying with the LangGraph Platform:
+## Free deployment
+
+There are two free options for deploying LangGraph applications via the LangGraph Server:
+
+1. [Local](../tutorials/langgraph-platform/local-server.md): Deploy for local testing and development. 
+1. [Standalone Container (Lite)](../concepts/langgraph_standalone_container.md): A limited version of Standalone Container for deployments unlikely to see more that 1 million node executions per year and that do not need crons and other enterprise features. Standalone Container (Lite) deployment option is free with a LangSmith API key.
+
+## Production deployment
+
+There are 4 main options for deploying with the [LangGraph Platform](langgraph_platform.md):

 1. [Cloud SaaS](#cloud-saas)

@@ -22,7 +31,7 @@ A quick comparison:
 |----------------------|----------------|----------------------------|-------------------------------|--------------------------|
 | **[Control plane UI/API](../concepts/langgraph_control_plane.md)** | Yes | Yes | Yes | No |
 | **CI/CD** | Managed internally by platform | Managed externally by you | Managed externally by you | Managed externally by you |
-| **Data/compute residency** | LangChain’s cloud | Your cloud | Your cloud | Your cloud |
+| **Data/compute residency** | LangChain's cloud | Your cloud | Your cloud | Your cloud |
 | **LangSmith compatibility** | Trace to LangSmith SaaS | Trace to LangSmith SaaS | Trace to Self-Hosted LangSmith | Optional tracing |
 | **[Server version compatibility](../concepts/langgraph_server.md#server-versions)** | Enterprise | Enterprise | Enterprise | Lite, Enterprise |
 | **[Pricing](https://www.langchain.com/pricing-langgraph-platform)** | Plus | Enterprise | Enterprise | Developer |
@@ -19,6 +19,7 @@ From the control plane UI, you can:
 - Update a deployment.
 - Update environment variables for a deployment.
 - View build and server logs of a deployment.
+- View deployment metrics like CPU and memory usage.
 - Delete a deployment.

 The Control Plane UI is embedded in [LangSmith](https://docs.smith.langchain.com/langgraph_cloud).
@@ -88,6 +89,15 @@ Infrastructure for deployments and revisions are provisioned and deployed asynch

 The control plane and [LangGraph Data Plane](./langgraph_data_plane.md) "listener" application coordinate to achieve asynchronous deployments.

+### Monitoring
+
+After a deployment is ready, the control plane monitors the deployment and records various metrics, such as:
+
+- CPU and memory usage of the deployment.
+- Number of container restarts.
+
+These metrics are displayed as charts in the Control Plane UI.
+
 ### LangSmith Integration

 A [LangSmith](https://docs.smith.langchain.com/) tracing project is automatically created for each deployment. The tracing project has the same name as the deployment. When creating a deployment, the `LANGCHAIN_TRACING` and `LANGSMITH_API_KEY`/`LANGCHAIN_API_KEY` environment variables do not need to be specified; they are set automatically by the control plane.
@@ -17,7 +17,7 @@ Develop, deploy, scale, and manage agents with **LangGraph Platform** — the pu

 LangGraph Platform makes it easy to get your agent running in production —  whether it’s built with LangGraph or another framework — so you can focus on your app logic, not infrastructure. Deploy with one click to get a live endpoint, and use our robust APIs and built-in task queues to handle production scale. 

- **[Streaming Support](../cloud/concepts/streaming.md)**: As agents grow more sophisticated, they often benefit from streaming both token outputs and intermediate states back to the user. Without this, users are left waiting for potentially long operations with no feedback. LangGraph Server provides multiple streaming modes optimized for various application needs.
+- **[Streaming Support](../cloud/how-tos/streaming.md)**: As agents grow more sophisticated, they often benefit from streaming both token outputs and intermediate states back to the user. Without this, users are left waiting for potentially long operations with no feedback. LangGraph Server provides multiple streaming modes optimized for various application needs.

 - **[Background Runs](../cloud/how-tos/background_run.md)**: For agents that take longer to process (e.g., hours), maintaining an open connection can be impractical. The LangGraph Server supports launching agent runs in the background and provides both polling endpoints and webhooks to monitor run status effectively.
 
@@ -3,7 +3,7 @@
 There are two versions of the self-hosted deployment: [Self-Hosted Data Plane](./deployment_options.md#self-hosted-data-plane) and [Self-Hosted Control Plane](./deployment_options.md#self-hosted-control-plane).

 !!! info "Important"
-    The Self-Hosted Control Plane deployment option is currently in beta stage and requires an [Enterprise](../../concepts/plans.md) plan.
+    The Self-Hosted Control Plane deployment option is currently in beta stage and requires an [Enterprise](plans.md) plan.

 ## Requirements

@@ -8,7 +8,7 @@ search:
 There are two versions of the self-hosted deployment: [Self-Hosted Data Plane](./deployment_options.md#self-hosted-data-plane) and [Self-Hosted Control Plane](./deployment_options.md#self-hosted-control-plane).

 !!! info "Important"
-    The Self-Hosted Data Plane deployment option is currently in beta stage and requires an [Enterprise](../../concepts/plans.md) plan.
+    The Self-Hosted Data Plane deployment option is currently in beta stage and requires an [Enterprise](plans.md) plan.

 ## Requirements

@@ -7,7 +7,7 @@ search:

 **LangGraph Server** offers an API for creating and managing agent-based applications. It is built on the concept of [assistants](assistants.md), which are agents configured for specific tasks, and includes built-in [persistence](persistence.md#memory-store) and a **task queue**. This versatile API supports a wide range of agentic application use cases, from background processing to real-time interactions.

-Use LangGraph Server to create and manage [assistants](assistants.md), [threads](../cloud/concepts/threads.md), [runs](../cloud/concepts/runs.md), [cron jobs](../cloud/concepts/cron_jobs.md), [webhooks](../cloud/concepts/webhooks.md), and more.
+Use LangGraph Server to create and manage [assistants](assistants.md), [threads](./persistence.md#threads), [runs](../cloud/concepts/runs.md), [cron jobs](../cloud/concepts/cron_jobs.md), [webhooks](../cloud/concepts/webhooks.md), and more.

 !!! tip "API reference"
  
@@ -24,6 +24,7 @@ Key features of LangGraph Studio:
 - [Manage assistants](../cloud/how-tos/studio/manage_assistants.md)
 - [Manage threads](../cloud/how-tos/threads_studio.md)
 - [Iterate on prompts](../cloud/how-tos/iterate_graph_studio.md)
+- [Run experiments over a dataset](../cloud/how-tos/studio/run_evals.md)
 - Manage [long term memory](memory.md)
 - Debug agent state via [time travel](time-travel.md)

@@ -41,4 +42,4 @@ Chat mode is a simpler UI for iterating on and testing chat-specific agents. It

 ## Learn more

- See this guide on how to [get started](../cloud/how-tos/studio/quick_start.md) with LangGraph Studio.
+- See this guide on how to [get started](../cloud/how-tos/studio/quick_start.md) with LangGraph Studio.
@@ -15,15 +15,19 @@ LangGraph has a built-in persistence layer, implemented through checkpointers. W

 ## Threads

-A thread is a unique ID or [thread identifier](#threads) assigned to each checkpoint saved by a checkpointer. When invoking graph with a checkpointer, you **must** specify a `thread_id` as part of the `configurable` portion of the config:
+A thread is a unique ID or thread identifier assigned to each checkpoint saved by a checkpointer. It contains the accumulated state of a sequence of [runs](../cloud/concepts/runs.md). When a run is executed, the [state](../concepts/low_level.md#state) of the underlying graph of the assistant will be persisted to the thread.
+
+When invoking graph with a checkpointer, you **must** specify a `thread_id` as part of the `configurable` portion of the config:

 ```python
 {"configurable": {"thread_id": "1"}}
 ```

+A thread's current and historical state can be retrieved. To persist state, a thread must be created prior to executing a run. The LangGraph Platform API provides several endpoints for creating and managing threads and thread state. See the [API reference](../cloud/reference/api/api_ref.html#tag/threads) for more details.
+
 ## Checkpoints

-Checkpoint is a snapshot of the graph state saved at each super-step and is represented by `StateSnapshot` object with the following key properties:
+The state of a thread at a particular point in time is called a checkpoint. Checkpoint is a snapshot of the graph state saved at each super-step and is represented by `StateSnapshot` object with the following key properties:

 - `config`: Config associated with this checkpoint. 
 - `metadata`: Metadata associated with this checkpoint.
@@ -31,6 +35,8 @@ Checkpoint is a snapshot of the graph state saved at each super-step and is repr
 - `next` A tuple of the node names to execute next in the graph.
 - `tasks`: A tuple of `PregelTask` objects that contain information about next tasks to be executed. If the step was previously attempted, it will include error information. If a graph was interrupted [dynamically](../how-tos/human_in_the_loop/breakpoints.ipynb#dynamic-breakpoints) from within a node, tasks will contain additional data associated with interrupts.

+Checkpoints are persisted and can be used to restore the state of a thread at a later time.
+
 Let's see what checkpoints are saved when a simple graph is invoked as follows:

 ```python
@@ -18,6 +18,6 @@ There are three main categories of data you can stream:

 - [**Stream LLM tokens**](../how-tos/streaming.md#messages) — capture token streams from anywhere: inside nodes, subgraphs, or tools.
 - [**Emit progress notifications from tools**](../how-tos/streaming.md#stream-custom-data) — send custom updates or progress signals directly from tool functions.
- [**Stream from subgraphs**](../how-tos/streaming.md#subgraphs) — include outputs from both the parent graph and any nested subgraphs.
+- [**Stream from subgraphs**](../how-tos/streaming.md#stream-subgraph-outputs) — include outputs from both the parent graph and any nested subgraphs.
 - [**Use any LLM**](../how-tos/streaming.md#use-with-any-llm) — stream tokens from any LLM, even if it's not a LangChain model using the `custom` streaming mode.
 - [**Use multiple streaming modes**](../how-tos/streaming.md#stream-multiple-modes) — choose from `values` (full state), `updates` (state deltas), `messages` (LLM tokens + metadata), `custom` (arbitrary user data), or `debug` (detailed traces).
@@ -1,62 +1,64 @@
 # Tools

-Many AI applications interact directly with humans. In these cases, it is appropriate for models to respond in natural language.
-But what about cases where we want a model to also interact *directly* with systems, such as databases or an API?
-These systems often have a particular input schema; for example, APIs frequently have a required payload structure. You can use [tool calling](https://platform.openai.com/docs/guides/function-calling/example-use-cases) to request model responses that match a particular schema.
+Many AI applications interact with users via natural language. However, some use cases require models to interface directly with external systems—such as APIs, databases, or file systems—using structured input. In these scenarios, **tool calling** enables models to generate requests that conform to a specified input schema.

-[Tools](https://python.langchain.com/docs/concepts/tools/) are a way to encapsulate a function and its input schema in a way that can be passed to a chat model that supports tool calling. This allows the model to request the execution of this function with specific inputs.
-
-**Tools** can be passed to [chat models](https://python.langchain.com/docs/concepts/chat_models) that support [tool calling](https://python.langchain.com/docs/concepts/tool_calling) allowing the model to request the execution of a specific function with specific inputs.
-
-You can [create custom tools](https://python.langchain.com/docs/how_to/custom_tools/) or use [prebuilt](#prebuilt-tools) tools.
+[Tools](https://python.langchain.com/docs/concepts/tools/) encapsulate a callable function and its input schema. These can be passed to compatible [chat models](https://python.langchain.com/docs/concepts/chat_models), allowing the model to decide whether to invoke a tool and with what arguments.

 ## Tool calling

 ![Diagram of a tool call by a model](./img/tool_call.png)

-A key principle of tool calling is that the model decides when to use a tool based on the input's relevance. The model doesn't always need to call a tool.
-For example, given an input that is *irrelevant to the tool*, the model would not call the tool:
+Tool calling is typically **conditional**. Based on the user input and available tools, the model may choose to issue a tool call request. This request is returned in an `AIMessage` object, which includes a `tool_calls` field that specifies the tool name and input arguments:

 ```python
-result = llm_with_tools.invoke("Hello world!")
+llm_with_tools.invoke("What is 2 multiplied by 3?")
+# -> AIMessage(tool_calls=[{'name': 'multiply', 'args': {'a': 2, 'b': 3}, ...}])
 ```

-The result would be an `AIMessage` containing the model's response in natural language (e.g., "Hello!").
-However, if we pass an input *relevant to the tool*, the model should choose to call it:
+If the input is unrelated to any tool, the model returns only a natural language message:

 ```python
-result = llm_with_tools.invoke("What is 2 multiplied by 3?")
+llm_with_tools.invoke("Hello world!")  # -> AIMessage(content="Hello!")
 ```

-As before, the output `result` will be an `AIMessage`. 
-But, if the tool was called, `result` will have a `tool_calls` attribute.
-This attribute includes everything needed to execute the tool, including the tool name and input arguments:
+Importantly, the model does not execute the tool—it only generates a request. A separate executor (such as a runtime or agent) is responsible for handling the tool call and returning the result.

-```
-result.tool_calls
-{'name': 'multiply', 'args': {'a': 2, 'b': 3}, 'id': 'xxx', 'type': 'tool_call'}
-```
-
-For more details on usage, see the [how-to guide](../how-tos/tool-calling.ipynb).
-
-## Execute tools
-
-LangGraph offers pre-built components — [`ToolNode`][langgraph.prebuilt.tool_node.ToolNode] and [`create_react_agent`][langgraph.prebuilt.chat_agent_executor.create_react_agent] — that invoke the tools on behalf of the user.
-
-See this [how-to guide](../how-tos/tool-calling.ipynb#use-prebuilt-toolnode) on tool calling.
+See the [tool calling guide](../how-tos/tool-calling.md) for more details.

 ## Prebuilt tools

-LangChain supports a wide range of prebuilt tool integrations for interacting with APIs, databases, file systems, web data, and more. These tools extend the functionality of agents and enable rapid development.
+LangChain provides prebuilt tool integrations for common external systems including APIs, databases, file systems, and web data.

-You can browse the full list of available integrations in the [LangChain integrations directory](https://python.langchain.com/docs/integrations/tools/).
+Browse the [integrations directory](https://python.langchain.com/docs/integrations/tools/) for available tools.

-Some commonly used tool categories include:
+Common categories:

- **Search**: Bing, SerpAPI, Tavily
- **Code interpreters**: Python REPL, Node.js REPL
- **Databases**: SQL, MongoDB, Redis
- **Web data**: Web scraping and browsing
- **APIs**: OpenWeatherMap, NewsAPI, and others
+* **Search**: Bing, SerpAPI, Tavily
+* **Code execution**: Python REPL, Node.js REPL
+* **Databases**: SQL, MongoDB, Redis
+* **Web data**: Scraping and browsing
+* **APIs**: OpenWeatherMap, NewsAPI, etc.

-These integrations can be configured and added to your agents using the same `tools` parameter shown in the examples above.
+## Custom tools
+
+You can define custom tools using the `@tool` decorator or plain Python functions. For example:
+
+```python
+from langchain_core.tools import tool
+
+@tool
+def multiply(a: int, b: int) -> int:
+    """Multiply two numbers."""
+    return a * b
+```
+
+See the [tool calling guide](../how-tos/tool-calling.md) for more details.
+
+## Tool execution
+
+While the model determines *when* to call a tool, **execution** of the tool call must be handled by a runtime component.
+
+LangGraph provides prebuilt components for this:
+
+* [`ToolNode`][oolNode]: Executes tools based on AI tool calls.
+* [`create_react_agent`][create_react_agent]: Constructs a full agent that manages tool calling automatically.
@@ -1,11 +1,220 @@
 # Stream outputs

-## Streaming API
+You can [stream outputs](../concepts/streaming.md) from a LangGraph agent or workflow.
+
+## Supported stream modes
+
+Pass one or more of the following stream modes as a list to the [`stream()`][langgraph.graph.state.CompiledStateGraph.stream] or [`astream()`][langgraph.graph.state.CompiledStateGraph.astream] methods:
+
+| Mode | Description |
+|------|-------------|
+| `values` | Streams the full value of the state after each step of the graph. |
+| `updates` | Streams the updates to the state after each step of the graph. If multiple updates are made in the same step (e.g., multiple nodes are run), those updates are streamed separately. |
+| `custom` | Streams custom data from inside your graph nodes. |
+| `messages` | Streams 2-tuples (LLM token, metadata) from any graph nodes where an LLM is invoked. |
+| `debug` | Streams as much information as possible throughout the execution of the graph.   
+
+## Stream from an agent
+
+### Agent progress
+
+To stream agent progress, use the [`stream()`][langgraph.graph.state.CompiledStateGraph.stream] or [`astream()`][langgraph.graph.state.CompiledStateGraph.astream] methods with `stream_mode="updates"`. This emits an event after every agent step.
+
+For example, if you have an agent that calls a tool once, you should see the following updates:
+
+* **LLM node**: AI message with tool call requests
+* **Tool node**: Tool message with execution result
+* **LLM node**: Final AI response
+
+=== "Sync"
+
+    ```python
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_weather],
+    )
+    # highlight-next-line
+    for chunk in agent.stream(
+        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
+        # highlight-next-line
+        stream_mode="updates"
+    ):
+        print(chunk)
+        print("\n")
+    ```
+
+=== "Async"
+
+    ```python
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_weather],
+    )
+    # highlight-next-line
+    async for chunk in agent.astream(
+        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
+        # highlight-next-line
+        stream_mode="updates"
+    ):
+        print(chunk)
+        print("\n")
+    ```
+
+### LLM tokens
+
+To stream tokens as they are produced by the LLM, use `stream_mode="messages"`:
+
+=== "Sync"
+
+    ```python
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_weather],
+    )
+    # highlight-next-line
+    for token, metadata in agent.stream(
+        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
+        # highlight-next-line
+        stream_mode="messages"
+    ):
+        print("Token", token)
+        print("Metadata", metadata)
+        print("\n")
+    ```
+
+=== "Async"
+
+    ```python
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_weather],
+    )
+    # highlight-next-line
+    async for token, metadata in agent.astream(
+        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
+        # highlight-next-line
+        stream_mode="messages"
+    ):
+        print("Token", token)
+        print("Metadata", metadata)
+        print("\n")
+    ```
+
+### Tool updates
+
+To stream updates from tools as they are executed, you can use [get_stream_writer][langgraph.config.get_stream_writer].
+
+=== "Sync"
+
+    ```python
+    # highlight-next-line
+    from langgraph.config import get_stream_writer
+
+    def get_weather(city: str) -> str:
+        """Get weather for a given city."""
+        # highlight-next-line
+        writer = get_stream_writer()
+        # stream any arbitrary data
+        # highlight-next-line
+        writer(f"Looking up data for city: {city}")
+        return f"It's always sunny in {city}!"
+
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_weather],
+    )
+
+    for chunk in agent.stream(
+        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
+        # highlight-next-line
+        stream_mode="custom"
+    ):
+        print(chunk)
+        print("\n")
+    ```
+
+=== "Async"
+
+    ```python
+    # highlight-next-line
+    from langgraph.config import get_stream_writer
+
+    def get_weather(city: str) -> str:
+        """Get weather for a given city."""
+        # highlight-next-line
+        writer = get_stream_writer()
+        # stream any arbitrary data
+        # highlight-next-line
+        writer(f"Looking up data for city: {city}")
+        return f"It's always sunny in {city}!"
+
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_weather],
+    )
+
+    async for chunk in agent.astream(
+        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
+        # highlight-next-line
+        stream_mode="custom"
+    ):
+        print(chunk)
+        print("\n")
+    ```
+
+!!! Note
+    If you add `get_stream_writer` inside your tool, you won't be able to invoke the tool outside of a LangGraph execution context. 
+
+### Stream multiple modes
+
+You can specify multiple streaming modes by passing stream mode as a list: `stream_mode=["updates", "messages", "custom"]`:
+
+=== "Sync"
+
+    ```python
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_weather],
+    )
+
+    for stream_mode, chunk in agent.stream(
+        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
+        # highlight-next-line
+        stream_mode=["updates", "messages", "custom"]
+    ):
+        print(chunk)
+        print("\n")
+    ```
+
+=== "Async"
+
+    ```python
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_weather],
+    )
+
+    async for stream_mode, chunk in agent.astream(
+        {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
+        # highlight-next-line
+        stream_mode=["updates", "messages", "custom"]
+    ):
+        print(chunk)
+        print("\n")
+    ```
+
+### Disable streaming
+
+In some applications you might need to disable streaming of individual tokens for a given model. This is useful in [multi-agent](../agents/multi-agent.md) systems to control which agents stream their output.
+
+See the [Models](../agents/models.md#disable-streaming) guide to learn how to disable streaming.
+
+## Stream from a workflow 
+
+### Basic usage example

 LangGraph graphs expose the [`.stream()`][langgraph.pregel.Pregel.stream] (sync) and [`.astream()`][langgraph.pregel.Pregel.astream] (async) methods to yield streamed outputs as iterators.

-Basic usage example:
-
 === "Sync"

    ```python
@@ -61,18 +270,7 @@ Basic usage example:
      ```output
      {'refine_topic': {'topic': 'ice cream and cats'}}
      {'generate_joke': {'joke': 'This is a joke about ice cream and cats'}}
-      ```
-
-
-### Supported stream modes
-
-| Mode                             | Description                                                                                                                                                                         |
-|----------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-| [`values`](#stream-graph-state)  | Streams the full value of the state after each step of the graph.                                                                                                                   |
-| [`updates`](#stream-graph-state) | Streams the updates to the state after each step of the graph. If multiple updates are made in the same step (e.g., multiple nodes are run), those updates are streamed separately. |
-| [`custom`](#stream-custom-data)  | Streams custom data from inside your graph nodes.                                                                                                                                   |
-| [`messages`](#messages)          | Streams 2-tuples (LLM token, metadata) from any graph nodes where an LLM is invoked.                                                                                                        |
-| [`debug`](#debug)                | Streams as much information as possible throughout the execution of the graph.                                                                                                      |
+      ```                                                                                                   |

 ### Stream multiple modes

@@ -94,7 +292,7 @@ The streamed outputs will be tuples of `(mode, chunk)` where `mode` is the name
        print(chunk)
    ```

-## Stream graph state
+### Stream graph state

 Use the stream modes `updates` and `values` to stream the state of the graph as it executes.

@@ -157,7 +355,7 @@ graph = (
    ```


-## Subgraphs
+### Stream subgraph outputs

 To include outputs from [subgraphs](../concepts/subgraphs.md) in the streamed outputs, you can set `subgraphs=True` in the `.stream()` method of the parent graph. This will stream outputs from both the parent graph and any subgraphs.

@@ -233,7 +431,7 @@ for chunk in graph.stream(

      **Note** that we are receiving not just the node updates, but we also the namespaces which tell us what graph (or subgraph) we are streaming from.

-## Debugging {#debug}
+### Debugging {#debug}

 Use the `debug` streaming mode to stream as much information as possible throughout the execution of the graph. The streamed outputs include the name of the node as well as the full state.

@@ -247,7 +445,7 @@ for chunk in graph.stream(
 ```


-## LLM tokens {#messages}
+### LLM tokens {#messages}

 Use the `messages` streaming mode to stream Large Language Model (LLM) outputs **token by token** from any part of your graph, including nodes, tools, subgraphs, or tasks.

@@ -307,7 +505,7 @@ for message_chunk, metadata in graph.stream( # (2)!
 2. The "messages" stream mode returns an iterator of tuples `(message_chunk, metadata)` where `message_chunk` is the token streamed by the LLM and `metadata` is a dictionary with information about the graph node where the LLM was called and other information.


-### Filter by LLM invocation
+#### Filter by LLM invocation

 You can associate `tags` with LLM invocations to filter the streamed tokens by LLM invocation.

@@ -391,7 +589,7 @@ async for msg, metadata in graph.astream(  # (3)!
      4. The `stream_mode` is set to "messages" to stream LLM tokens. The `metadata` contains information about the LLM invocation, including the tags.


-### Filter by node
+#### Filter by node

 To stream tokens only from specific nodes, use `stream_mode="messages"` and filter the outputs by the `langgraph_node` field in the streamed metadata:

@@ -464,7 +662,7 @@ for msg, metadata in graph.stream( # (1)!
      1. The "messages" stream mode returns a tuple of `(message_chunk, metadata)` where `message_chunk` is the token streamed by the LLM and `metadata` is a dictionary with information about the graph node where the LLM was called and other information.
      2. Filter the streamed tokens by the `langgraph_node` field in the metadata to only include the tokens from the `write_poem` node.

-## Stream custom data
+### Stream custom data

 To send **custom user-defined data** from inside a LangGraph node or tool, follow these steps:

@@ -541,7 +739,7 @@ To send **custom user-defined data** from inside a LangGraph node or tool, follo
      3. Emit another custom key-value pair.
      4. Set `stream_mode="custom"` to receive the custom data in the stream.

-## Use with any LLM
+### Use with any LLM

 You can use `stream_mode="custom"` to stream data from **any LLM API**  — even if that API does **not** implement the LangChain chat model interface.

@@ -701,7 +899,7 @@ for chunk in graph.stream(
      ```


-## Disable streaming for specific chat models
+### Disable streaming for specific chat models

 If your application mixes models that support streaming with those that do not, you may need to explicitly disable streaming for 
 models that do not support it.
@@ -733,7 +931,7 @@ Set `disable_streaming=True` when initializing the model.
      1. Set `disable_streaming=True` to disable streaming for the chat model.


-## Async with Python < 3.11 { #async }
+### Async with Python < 3.11 { #async }

 In Python versions < 3.11, [asyncio tasks](https://docs.python.org/3/library/asyncio-task.html#asyncio.create_task) do not support the `context` parameter.  
 This limits LangGraph ability to automatically propagate context, and affects LangGraph’s streaming mechanisms in two key ways:
@@ -1,22 +0,0 @@
---
-search:
-  boost: 2
---
-
-# Deployment 🚀
-
-There are two free options for deploying LangGraph applications via the LangGraph Server:
-
- [Local](./langgraph-platform/local-server.md): Deploy for local testing and development. 
- [Standalone Container (Lite)](../concepts/langgraph_standalone_container.md): A limited version of Standalone Container for deployments unlikely to see more that 1 million node executions per year and that do not need crons and other enterprise features. Standalone Container (Lite) deployment option is free with a LangSmith API key.
-
-## Other deployment options
-
-Additionally, you can deploy to production with [LangGraph Platform](../concepts/langgraph_platform.md):
-
- [Cloud SaaS](../concepts/langgraph_cloud.md): Connect your GitHub repositories and deploy LangGraph Servers within LangChain's cloud. *We manage everything.*
- [Self-Hosted Data Plane<sup>(Beta)</sup>](../concepts/langgraph_self_hosted_data_plane.md): Create deployments from the [Control Plane UI](../concepts/langgraph_control_plane.md#control-plane-ui) and deploy LangGraph Servers to **your** cloud. *We manage the [control plane](../concepts/langgraph_control_plane.md). You manage the deployments.*
- [Self-Hosted Control Plane<sup>(Beta)</sup>](../concepts/langgraph_self_hosted_control_plane.md): Create deployments from a self-hosted [Control Plane UI](../concepts/langgraph_control_plane.md#control-plane-ui) and deploy LangGraph Servers to **your** cloud. *You manage everything.*
- [Standalone Container](../concepts/langgraph_standalone_container.md): Deploy LangGraph Server Docker images however you like.
-
-For more information, see [Deployment options](../concepts/deployment_options.md).
@@ -89,7 +89,7 @@
  },
  {
   "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 2,
   "id": "baf669a0-04ee-492d-80d8-8fcb658ed128",
   "metadata": {},
   "outputs": [],
@@ -313,8 +313,8 @@
    "\n",
    "    builder.add_edge(\"finalizer\", END)\n",
    "\n",
-    "    # These functions let the step be used in a\n",
-    "    # StateGraph with 'messages' as the key.\n",
+    "    # These functions let the step be used in a MessageGraph\n",
+    "    # or a StateGraph with 'messages' as the key.\n",
    "    def encode(x: Union[Sequence[AnyMessage], PromptValue]) -> dict:\n",
    "        \"\"\"Ensure the input is the correct format.\"\"\"\n",
    "        if isinstance(x, PromptValue):\n",
@@ -89,162 +89,145 @@ plugins:
                - "!^_"

 nav:
-  - Guides:
+  - Get started:
    - index.md
-    - Get started:
-      - Quickstart: agents/agents.md
-      - LangGraph basics:
-        -  concepts/why-langgraph.md
-        -  Build a basic chatbot: tutorials/get-started/1-build-basic-chatbot.md
-        -  tutorials/get-started/2-add-tools.md
-        -  tutorials/get-started/3-add-memory.md
-        -  Add human-in-the-loop: tutorials/get-started/4-human-in-the-loop.md
-        -  tutorials/get-started/5-customize-state.md
-        -  tutorials/get-started/6-time-travel.md
-      - Deployment: tutorials/deployment.md
-    - Prebuilt agents:
-      - Overview: agents/overview.md
-      - agents/run_agents.md
-      - agents/streaming.md
-      - agents/models.md
-      - agents/tools.md
-      - agents/mcp.md
-      - agents/context.md
-      - agents/memory.md
-      - agents/human-in-the-loop.md
-      - agents/multi-agent.md
-      - agents/evals.md
-      - agents/deployment.md
-      - agents/ui.md
-    - LangGraph framework:
-      - Agent architectures:
-        - Overview: concepts/agentic_concepts.md
+    - Quickstarts:
+      - Agent: agents/agents.md
+      - Local server: tutorials/langgraph-platform/local-server.md
+      - Deployment: cloud/quick_start.md
+    - General concepts:
+      - Common patterns:
+        - Agent architectures: concepts/agentic_concepts.md
        - Workflows & agents: tutorials/workflows.md
-      - Graphs:
-        - Overview: concepts/low_level.md
-        - Runtime overview: concepts/pregel.md
-        - Use the Graph API: how-tos/graph-api.ipynb
-      - Streaming:
-          - Overview: concepts/streaming.md
-          - "Stream outputs": how-tos/streaming.md
-      - Persistence:
-        - Overview: concepts/persistence.md
-        - concepts/durable_execution.md
-        - how-tos/persistence.ipynb
-      - Memory:
-        - Overview: concepts/memory.md
-        - Manage memory: how-tos/memory.ipynb
-      - Human-in-the-loop:
-        - Overview: concepts/human_in_the_loop.md
-        - how-tos/human_in_the_loop/add-human-in-the-loop.md
-      - Breakpoints:
-        - Overview: concepts/breakpoints.md
-        - how-tos/human_in_the_loop/breakpoints.ipynb
-      - Time travel:
-        - Overview: concepts/time-travel.md
-        - how-tos/human_in_the_loop/time-travel.ipynb
-      - Tools:
-        - Overview: concepts/tools.md
-        - how-tos/tool-calling.ipynb
-      - Subgraphs:
-        - Overview: concepts/subgraphs.md
-        - how-tos/subgraph.ipynb
-      - Multi-agent:
-        - Overview: concepts/multi_agent.md
-        - how-tos/multi_agent.ipynb
-      - Functional API:
-        - Overview: concepts/functional_api.md
-        - how-tos/use-functional-api.md
-
-    - LangGraph Platform:
-      - Overview: concepts/langgraph_platform.md
-      - Get started:
-        - Quickstart: tutorials/langgraph-platform/local-server.md
-        - Deployment quickstart: cloud/quick_start.md
-      - Components:
-        - Overview: concepts/langgraph_components.md
-        - LangGraph Server:
-          - Overview: concepts/langgraph_server.md
-          - Application structure:
-            - Overview: concepts/application_structure.md
-            - cloud/deployment/setup.md
-            - cloud/deployment/setup_pyproject.md
-            - cloud/deployment/setup_javascript.md
-            - cloud/deployment/custom_docker.md
-        - LangGraph CLI: concepts/langgraph_cli.md
-        - LangGraph Studio:
-          - Overview: concepts/langgraph_studio.md
-          - Quickstart: cloud/how-tos/studio/quick_start.md
-          - cloud/how-tos/invoke_studio.md
-          - cloud/how-tos/studio/manage_assistants.md
-          - cloud/how-tos/threads_studio.md
-          - cloud/how-tos/iterate_graph_studio.md
-          - cloud/how-tos/clone_traces_studio.md
-          - cloud/how-tos/datasets_studio.md
-        - LangGraph SDK: concepts/sdk.md
-      - Data management:
-        - Add semantic search: cloud/deployment/semantic_search.md
-        - Add TTLs: how-tos/ttl/configure_ttl.md
+      - Agent development: agents/overview.md
+      - Workflow orchestration:
+        - Graphs: concepts/low_level.md
+        - Subgraphs: concepts/subgraphs.md
+        - Runtime: concepts/pregel.md
+        - Functional API: concepts/functional_api.md
+    - Core capabilities:
+      - Streaming: concepts/streaming.md
+      - Persistence: concepts/persistence.md
+      - Durable execution: concepts/durable_execution.md
+      - Memory: concepts/memory.md
+      - Tools: concepts/tools.md
+      - Human-in-the-loop: concepts/human_in_the_loop.md
+      - Breakpoints: concepts/breakpoints.md
+      - Time travel: concepts/time-travel.md
+      - Multi-agent: concepts/multi_agent.md
+    - Platform capabilities:
+      - LangGraph Platform:
+          - Overview: concepts/langgraph_platform.md
+          - Components:
+            - Overview: concepts/langgraph_components.md
+            - LangGraph Server:
+              - Overview: concepts/langgraph_server.md
+              - Data plane: concepts/langgraph_data_plane.md
+              - Control plane: concepts/langgraph_control_plane.md
+            - LangGraph CLI: concepts/langgraph_cli.md
+            - LangGraph Studio: concepts/langgraph_studio.md
+            - LangGraph SDK: concepts/sdk.md
+          - Plans & pricing: concepts/plans.md
+          - Application structure: concepts/application_structure.md
+          - Scalability & resilience: concepts/scalability_and_resilience.md
+      - Authentication & access control: concepts/auth.md
+      - Assistants: concepts/assistants.md
+      - Double-texting: concepts/double_texting.md
+      - Webhooks: cloud/concepts/webhooks.md
+      - Cron jobs: cloud/concepts/cron_jobs.md
+      - Deployment:
+        - Overview: concepts/deployment_options.md
+        - Deployment options:
+          - Cloud SaaS: concepts/langgraph_cloud.md
+          - Self-Hosted Data Plane: concepts/langgraph_self_hosted_data_plane.md
+          - Self-Hosted Control Plane: concepts/langgraph_self_hosted_control_plane.md
+          - Standalone Container: concepts/langgraph_standalone_container.md
+    
+  - Guides:
+    - LangGraph APIs:
+      - Use the Graph API: how-tos/graph-api.ipynb
+      - Use the Functional API: how-tos/use-functional-api.md
+    - Models:
+      - Configure model: agents/models.md
+    - Streaming:
+      - Stream outputs: how-tos/streaming.md
+      - Use Server API: cloud/how-tos/streaming.md
+    - Context:
+      - Use in agent: agents/context.md
+    - Memory:
+      - Basic implementation: agents/memory.md
+      - Persistence: how-tos/persistence.ipynb # MERGE
+      - Custom implementation: how-tos/memory.ipynb
+    - Human-in-the-loop:
+      - Add to agent: agents/human-in-the-loop.md
+      - Add to workflow: how-tos/human_in_the_loop/add-human-in-the-loop.md
+      - Use Server API: cloud/how-tos/add-human-in-the-loop.md
+    - Time travel:
+      - Use Server API: cloud/how-tos/human_in_the_loop_time_travel.md
+    - Breakpoints:
+      - Set breakpoints: how-tos/human_in_the_loop/breakpoints.ipynb
+      - Use Server API: cloud/how-tos/human_in_the_loop_breakpoint.md
+    - Tools:
+      - Call tools: how-tos/tool-calling.md
+    - Subgraphs:
+      - Use subgraphs: how-tos/subgraph.ipynb
+    - Multi-agent:
+      - Prebuilt implementation: agents/multi-agent.md
+      - Custom implementation: how-tos/multi_agent.ipynb
+    - MCP:
+      - Use MCP tools: agents/mcp.md
+      - Server deployment via MCP: concepts/server-mcp.md
+    - Deployment:
+      - Basic deployment: agents/deployment.md
+      - Set up your application:
+        - Use requirements.txt: cloud/deployment/setup.md
+        - Use pyproject.toml: cloud/deployment/setup_pyproject.md
+        - Use JavaScript: cloud/deployment/setup_javascript.md
+        - Use custom Docker: cloud/deployment/custom_docker.md
+      - Deploy to production:
+         - Cloud SaaS: cloud/deployment/cloud.md
+         - Self-Hosted Data Plane: cloud/deployment/self_hosted_data_plane.md
+         - Self-Hosted Control Plane: cloud/deployment/self_hosted_control_plane.md
+         - Standalone Container: cloud/deployment/standalone_container.md
+    - Evaluation:
+      - Basic implementation: agents/evals.md
+    - Platform capabilities:
+      - LangGraph Studio:
+        - Quickstart: cloud/how-tos/studio/quick_start.md
+        - cloud/how-tos/invoke_studio.md
+        - cloud/how-tos/studio/manage_assistants.md
+        - cloud/how-tos/threads_studio.md
+        - cloud/how-tos/iterate_graph_studio.md
+        - cloud/how-tos/studio/run_evals.md
+        - cloud/how-tos/clone_traces_studio.md
+        - cloud/how-tos/datasets_studio.md
      - Authentication & access control:
-        - Overview: concepts/auth.md
        - how-tos/auth/custom_auth.md
        - how-tos/auth/openapi_security.md
      - Assistants:
-        - Overview: concepts/assistants.md
        - cloud/how-tos/configuration_cloud.md
-      - Threads:
-        - Overview: cloud/concepts/threads.md
-        - cloud/how-tos/use_threads.md
-      - Runs:
-        - Overview: cloud/concepts/runs.md
-        - cloud/how-tos/background_run.md
-        - cloud/how-tos/same-thread.md
-        - cloud/how-tos/cron_jobs.md
-        - cloud/how-tos/stateless_runs.md
-        - cloud/how-tos/configurable_headers.md
-      - Streaming:
-        - Overview: cloud/concepts/streaming.md
-        - cloud/how-tos/streaming.md
-      - Human-in-the-loop: cloud/how-tos/add-human-in-the-loop.md
-      - Breakpoints: cloud/how-tos/human_in_the_loop_breakpoint.md
-      - Time travel: cloud/how-tos/human_in_the_loop_time_travel.md
-      - MCP: concepts/server-mcp.md
+        - Threads: cloud/how-tos/use_threads.md
+        - Runs:
+          - cloud/how-tos/background_run.md
+          - cloud/how-tos/same-thread.md
+          - cloud/how-tos/cron_jobs.md
+          - cloud/how-tos/stateless_runs.md
+          - cloud/how-tos/configurable_headers.md
      - Double-texting:
-        - Overview: concepts/double_texting.md
        - cloud/how-tos/interrupt_concurrent.md
        - cloud/how-tos/rollback_concurrent.md
        - cloud/how-tos/reject_concurrent.md
        - cloud/how-tos/enqueue_concurrent.md
-      - Webhooks:
-        - Overview: cloud/concepts/webhooks.md
-        - cloud/how-tos/webhooks.md
-      - Cron jobs:
-        - Overview: cloud/concepts/cron_jobs.md
-        - cloud/how-tos/cron_jobs.md
+      - Webhooks: cloud/how-tos/webhooks.md
+      - Cron jobs: cloud/how-tos/cron_jobs.md
      - Server customization:
        - how-tos/http/custom_lifespan.md
        - how-tos/http/custom_middleware.md
        - how-tos/http/custom_routes.md
-      - Deployment:
-        - Overview: concepts/deployment_options.md
-        - Data plane: concepts/langgraph_data_plane.md
-        - Control plane: concepts/langgraph_control_plane.md
-        - Deployment options:
-          - Cloud SaaS:
-            - Overview: concepts/langgraph_cloud.md
-            - Deploy Cloud SaaS: cloud/deployment/cloud.md
-          - Self-Hosted Data Plane:
-            - Overview: concepts/langgraph_self_hosted_data_plane.md
-            - Deploy Self-Hosted Data Plane: cloud/deployment/self_hosted_data_plane.md
-          - Self-Hosted Control Plane:
-            - Overview: concepts/langgraph_self_hosted_control_plane.md
-            - Deploy Self-Hosted Control Plane: cloud/deployment/self_hosted_control_plane.md
-          - Standalone Container:
-            - Overview: concepts/langgraph_standalone_container.md
-            - Deploy Standalone Container: cloud/deployment/standalone_container.md
-        - Scalability & resilience: concepts/scalability_and_resilience.md
-        - Plans & pricing: concepts/plans.md
-      
+      - Data management:
+        - Add semantic search: cloud/deployment/semantic_search.md
+        - Add TTLs: how-tos/ttl/configure_ttl.md
+
  - Reference:
    - reference/index.md
    - LangGraph:
@@ -273,9 +256,20 @@ nav:
      - Environment variables: cloud/reference/env_var.md

  - Examples:
+    - agents/run_agents.md
+    - LangGraph basics:
+      -  concepts/why-langgraph.md
+      -  Build a basic chatbot: tutorials/get-started/1-build-basic-chatbot.md
+      -  tutorials/get-started/2-add-tools.md
+      -  tutorials/get-started/3-add-memory.md
+      -  Add human-in-the-loop: tutorials/get-started/4-human-in-the-loop.md
+      -  tutorials/get-started/5-customize-state.md
+      -  tutorials/get-started/6-time-travel.md
+    - Template applications: concepts/template_applications.md  # TODO: make tutorial
    - Agentic RAG: tutorials/rag/langgraph_agentic_rag.ipynb
    - Agent Supervisor: tutorials/multi_agent/agent_supervisor.ipynb
    - SQL agent: tutorials/sql-agent.ipynb
+    - Prebuilt chat UI: agents/ui.md
    - Graph runs in LangSmith: how-tos/run-id-langsmith.ipynb
    - LangGraph Platform:
      - Authentication:
@@ -290,11 +284,12 @@ nav:
        - Integrate LangGraph into a React app: cloud/how-tos/use_stream_react.md
        - Implement generative UI with LangGraph: cloud/how-tos/generative_ui_react.md

-  - Resources:
-    - concepts/faq.md
-    - Template applications: concepts/template_applications.md  # TODO: make tutorial
-    - llms.txt: llms-txt-overview.md
+  - Additional resources:
    - agents/prebuilt.md # NOTE: prebuilt.md is auto-generated by `make build-prebuilt`
+    - LangGraph Academy course: https://academy.langchain.com/courses/intro-to-langgraph
+    - Case studies: adopters.md
+    - concepts/faq.md
+    - llms.txt: llms-txt-overview.md
    - Troubleshooting:
        - Errors:
          - troubleshooting/errors/index.md
@@ -305,9 +300,7 @@ nav:
          - troubleshooting/errors/INVALID_CHAT_HISTORY.md
          - troubleshooting/errors/INVALID_LICENSE.md
        - LangGraph Studio: troubleshooting/studio.md
-    - Learn:
-      - LangGraph Academy course: https://academy.langchain.com/courses/intro-to-langgraph
-      - Case studies: adopters.md
+    

 markdown_extensions:
  - abbr
@@ -364,16 +357,6 @@ markdown_extensions:
 hooks:
  - _scripts/notebook_hooks.py
 extra:
-  consent:
-    title: Cookie consent
-    actions:
-      - accept
-      - reject
-    description: >- 
-      We use cookies to recognize your repeated visits and preferences, as well
-      as to measure the effectiveness of our documentation and whether users
-      find what they're searching for. <strong>Clicking "Accept" makes our
-      documentation better. Thank you!</strong> ❤️
  social:
    - icon: fontawesome/brands/js
      link: https://langchain-ai.github.io/langgraphjs/
@@ -2590,7 +2590,7 @@ wheels = [

 [[package]]
 name = "langgraph"
-version = "0.4.7"
+version = "0.5.0rc1"
 source = { editable = "../libs/langgraph" }
 dependencies = [
    { name = "langchain-core" },
@@ -2641,7 +2641,7 @@ dev = [

 [[package]]
 name = "langgraph-checkpoint"
-version = "2.0.26"
+version = "2.1.0"
 source = { editable = "../libs/checkpoint" }
 dependencies = [
    { name = "langchain-core" },
@@ -2660,6 +2660,8 @@ dev = [
    { name = "dataclasses-json" },
    { name = "mypy" },
    { name = "numpy" },
+    { name = "pandas" },
+    { name = "pandas-stubs", specifier = ">=2.2.2.240807" },
    { name = "pytest" },
    { name = "pytest-asyncio" },
    { name = "pytest-mock" },
@@ -2892,7 +2894,7 @@ test = [

 [[package]]
 name = "langgraph-prebuilt"
-version = "0.2.2"
+version = "0.5.0rc0"
 source = { editable = "../libs/prebuilt" }
 dependencies = [
    { name = "langchain-core" },
@@ -23,6 +23,7 @@ from langgraph.checkpoint.base import (
 )
 from langgraph.checkpoint.postgres import _internal
 from langgraph.checkpoint.postgres.base import BasePostgresSaver
+from langgraph.checkpoint.postgres.shallow import ShallowPostgresSaver
 from langgraph.checkpoint.serde.base import SerializerProtocol

 Conn = _internal.Conn  # For backward compatibility
@@ -456,4 +457,4 @@ class PostgresSaver(BasePostgresSaver):
        )


-__all__ = ["PostgresSaver", "BasePostgresSaver", "Conn"]
+__all__ = ["PostgresSaver", "BasePostgresSaver", "ShallowPostgresSaver", "Conn"]
@@ -23,6 +23,7 @@ from langgraph.checkpoint.base import (
 )
 from langgraph.checkpoint.postgres import _ainternal
 from langgraph.checkpoint.postgres.base import BasePostgresSaver
+from langgraph.checkpoint.postgres.shallow import AsyncShallowPostgresSaver
 from langgraph.checkpoint.serde.base import SerializerProtocol

 Conn = _ainternal.Conn  # For backward compatibility
@@ -559,4 +560,4 @@ class AsyncPostgresSaver(BasePostgresSaver):
        ).result()


-__all__ = ["AsyncPostgresSaver", "Conn"]
+__all__ = ["AsyncPostgresSaver", "AsyncShallowPostgresSaver", "Conn"]
@@ -168,7 +168,7 @@ class BasePostgresSaver(BaseCheckpointSaver[str]):
        checkpoint["channel_versions"][TASKS] = (
            max(checkpoint["channel_versions"].values())
            if checkpoint["channel_versions"]
-            else self.get_next_version(None)
+            else self.get_next_version(None, None)
        )

    def _load_blobs(
@@ -246,7 +246,7 @@ class BasePostgresSaver(BaseCheckpointSaver[str]):
            for idx, (channel, value) in enumerate(writes)
        ]

-    def get_next_version(self, current: str | None) -> str:
+    def get_next_version(self, current: str | None, channel: None) -> str:
        if current is None:
            current_v = 0
        elif isinstance(current, int):
@@ -0,0 +1,959 @@
+import asyncio
+import threading
+import warnings
+from collections.abc import AsyncIterator, Iterator, Sequence
+from contextlib import asynccontextmanager, contextmanager
+from typing import Any, Optional
+
+from langchain_core.runnables import RunnableConfig
+from psycopg import (
+    AsyncConnection,
+    AsyncCursor,
+    AsyncPipeline,
+    Capabilities,
+    Connection,
+    Cursor,
+    Pipeline,
+)
+from psycopg.rows import DictRow, dict_row
+from psycopg.types.json import Jsonb
+from psycopg_pool import AsyncConnectionPool, ConnectionPool
+
+from langgraph.checkpoint.base import (
+    WRITES_IDX_MAP,
+    ChannelVersions,
+    Checkpoint,
+    CheckpointMetadata,
+    CheckpointTuple,
+    get_checkpoint_metadata,
+)
+from langgraph.checkpoint.postgres import _ainternal, _internal
+from langgraph.checkpoint.postgres.base import BasePostgresSaver
+from langgraph.checkpoint.serde.base import SerializerProtocol
+from langgraph.checkpoint.serde.types import TASKS
+
+"""
+To add a new migration, add a new string to the MIGRATIONS list.
+The position of the migration in the list is the version number.
+"""
+MIGRATIONS = [
+    """CREATE TABLE IF NOT EXISTS checkpoint_migrations (
+    v INTEGER PRIMARY KEY
+);""",
+    """CREATE TABLE IF NOT EXISTS checkpoints (
+    thread_id TEXT NOT NULL,
+    checkpoint_ns TEXT NOT NULL DEFAULT '',
+    type TEXT,
+    checkpoint JSONB NOT NULL,
+    metadata JSONB NOT NULL DEFAULT '{}',
+    PRIMARY KEY (thread_id, checkpoint_ns)
+);""",
+    """CREATE TABLE IF NOT EXISTS checkpoint_blobs (
+    thread_id TEXT NOT NULL,
+    checkpoint_ns TEXT NOT NULL DEFAULT '',
+    channel TEXT NOT NULL,
+    type TEXT NOT NULL,
+    blob BYTEA,
+    PRIMARY KEY (thread_id, checkpoint_ns, channel)
+);""",
+    """CREATE TABLE IF NOT EXISTS checkpoint_writes (
+    thread_id TEXT NOT NULL,
+    checkpoint_ns TEXT NOT NULL DEFAULT '',
+    checkpoint_id TEXT NOT NULL,
+    task_id TEXT NOT NULL,
+    idx INTEGER NOT NULL,
+    channel TEXT NOT NULL,
+    type TEXT,
+    blob BYTEA NOT NULL,
+    PRIMARY KEY (thread_id, checkpoint_ns, checkpoint_id, task_id, idx)
+);""",
+    """
+    CREATE INDEX CONCURRENTLY IF NOT EXISTS checkpoints_thread_id_idx ON checkpoints(thread_id);
+    """,
+    """
+    CREATE INDEX CONCURRENTLY IF NOT EXISTS checkpoint_blobs_thread_id_idx ON checkpoint_blobs(thread_id);
+    """,
+    """
+    CREATE INDEX CONCURRENTLY IF NOT EXISTS checkpoint_writes_thread_id_idx ON checkpoint_writes(thread_id);
+    """,
+    """
+    ALTER TABLE checkpoint_writes ADD COLUMN task_path TEXT NOT NULL DEFAULT '';
+    """,
+]
+
+SELECT_SQL = f"""
+select
+    thread_id,
+    checkpoint,
+    checkpoint_ns,
+    metadata,
+    (
+        select array_agg(array[bl.channel::bytea, bl.type::bytea, bl.blob])
+        from jsonb_each_text(checkpoint -> 'channel_versions')
+        inner join checkpoint_blobs bl
+            on bl.thread_id = checkpoints.thread_id
+            and bl.checkpoint_ns = checkpoints.checkpoint_ns
+            and bl.channel = jsonb_each_text.key
+    ) as channel_values,
+    (
+        select
+        array_agg(array[cw.task_id::text::bytea, cw.channel::bytea, cw.type::bytea, cw.blob] order by cw.task_id, cw.idx)
+        from checkpoint_writes cw
+        where cw.thread_id = checkpoints.thread_id
+            and cw.checkpoint_ns = checkpoints.checkpoint_ns
+            and cw.checkpoint_id = (checkpoint->>'id')
+    ) as pending_writes,
+    (
+        select array_agg(array[cw.type::bytea, cw.blob] order by cw.task_path, cw.task_id, cw.idx)
+        from checkpoint_writes cw
+        where cw.thread_id = checkpoints.thread_id
+            and cw.checkpoint_ns = checkpoints.checkpoint_ns
+            and cw.channel = '{TASKS}'
+    ) as pending_sends
+from checkpoints """
+
+UPSERT_CHECKPOINT_BLOBS_SQL = """
+    INSERT INTO checkpoint_blobs (thread_id, checkpoint_ns, channel, type, blob)
+    VALUES (%s, %s, %s, %s, %s)
+    ON CONFLICT (thread_id, checkpoint_ns, channel) DO UPDATE SET
+        type = EXCLUDED.type,
+        blob = EXCLUDED.blob;
+"""
+
+UPSERT_CHECKPOINTS_SQL = """
+    INSERT INTO checkpoints (thread_id, checkpoint_ns, checkpoint, metadata)
+    VALUES (%s, %s, %s, %s)
+    ON CONFLICT (thread_id, checkpoint_ns)
+    DO UPDATE SET
+        checkpoint = EXCLUDED.checkpoint,
+        metadata = EXCLUDED.metadata;
+"""
+
+UPSERT_CHECKPOINT_WRITES_SQL = """
+    INSERT INTO checkpoint_writes (thread_id, checkpoint_ns, checkpoint_id, task_id, task_path, idx, channel, type, blob)
+    VALUES (%s, %s, %s, %s, %s, %s, %s, %s, %s)
+    ON CONFLICT (thread_id, checkpoint_ns, checkpoint_id, task_id, idx) DO UPDATE SET
+        channel = EXCLUDED.channel,
+        type = EXCLUDED.type,
+        blob = EXCLUDED.blob;
+"""
+
+INSERT_CHECKPOINT_WRITES_SQL = """
+    INSERT INTO checkpoint_writes (thread_id, checkpoint_ns, checkpoint_id, task_id, task_path, idx, channel, type, blob)
+    VALUES (%s, %s, %s, %s, %s, %s, %s, %s, %s)
+    ON CONFLICT (thread_id, checkpoint_ns, checkpoint_id, task_id, idx) DO NOTHING
+"""
+
+
+def _dump_blobs(
+    serde: SerializerProtocol,
+    thread_id: str,
+    checkpoint_ns: str,
+    values: dict[str, Any],
+    versions: ChannelVersions,
+) -> list[tuple[str, str, str, str, Optional[bytes]]]:
+    if not versions:
+        return []
+
+    return [
+        (
+            thread_id,
+            checkpoint_ns,
+            k,
+            *(serde.dumps_typed(values[k]) if k in values else ("empty", None)),
+        )
+        for k in versions
+    ]
+
+
+class ShallowPostgresSaver(BasePostgresSaver):
+    """A checkpoint saver that uses Postgres to store checkpoints.
+
+    This checkpointer ONLY stores the most recent checkpoint and does NOT retain any history.
+    It is meant to be a light-weight drop-in replacement for the PostgresSaver that
+    supports most of the LangGraph persistence functionality with the exception of time travel.
+    """
+
+    SELECT_SQL = SELECT_SQL
+    MIGRATIONS = MIGRATIONS
+    UPSERT_CHECKPOINT_BLOBS_SQL = UPSERT_CHECKPOINT_BLOBS_SQL
+    UPSERT_CHECKPOINTS_SQL = UPSERT_CHECKPOINTS_SQL
+    UPSERT_CHECKPOINT_WRITES_SQL = UPSERT_CHECKPOINT_WRITES_SQL
+    INSERT_CHECKPOINT_WRITES_SQL = INSERT_CHECKPOINT_WRITES_SQL
+
+    lock: threading.Lock
+
+    def __init__(
+        self,
+        conn: _internal.Conn,
+        pipe: Optional[Pipeline] = None,
+        serde: Optional[SerializerProtocol] = None,
+    ) -> None:
+        warnings.warn(
+            "ShallowPostgresSaver is deprecated as of version 2.0.20 and will be removed in 3.0.0. "
+            "Use PostgresSaver instead, and invoke the graph with `graph.invoke(..., checkpoint_during=False)`.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
+        super().__init__(serde=serde)
+        if isinstance(conn, ConnectionPool) and pipe is not None:
+            raise ValueError(
+                "Pipeline should be used only with a single Connection, not ConnectionPool."
+            )
+
+        self.conn = conn
+        self.pipe = pipe
+        self.lock = threading.Lock()
+        self.supports_pipeline = Capabilities().has_pipeline()
+
+    @classmethod
+    @contextmanager
+    def from_conn_string(
+        cls, conn_string: str, *, pipeline: bool = False
+    ) -> Iterator["ShallowPostgresSaver"]:
+        """Create a new ShallowPostgresSaver instance from a connection string.
+
+        Args:
+            conn_string: The Postgres connection info string.
+            pipeline: whether to use Pipeline
+
+        Returns:
+            ShallowPostgresSaver: A new ShallowPostgresSaver instance.
+        """
+        with Connection.connect(
+            conn_string, autocommit=True, prepare_threshold=0, row_factory=dict_row
+        ) as conn:
+            if pipeline:
+                with conn.pipeline() as pipe:
+                    yield cls(conn, pipe)
+            else:
+                yield cls(conn)
+
+    def setup(self) -> None:
+        """Set up the checkpoint database asynchronously.
+
+        This method creates the necessary tables in the Postgres database if they don't
+        already exist and runs database migrations. It MUST be called directly by the user
+        the first time checkpointer is used.
+        """
+        with self._cursor() as cur:
+            cur.execute(self.MIGRATIONS[0])
+            results = cur.execute(
+                "SELECT v FROM checkpoint_migrations ORDER BY v DESC LIMIT 1"
+            )
+            row = results.fetchone()
+            if row is None:
+                version = -1
+            else:
+                version = row["v"]
+            for v, migration in zip(
+                range(version + 1, len(self.MIGRATIONS)),
+                self.MIGRATIONS[version + 1 :],
+            ):
+                cur.execute(migration)
+                cur.execute(f"INSERT INTO checkpoint_migrations (v) VALUES ({v})")
+        if self.pipe:
+            self.pipe.sync()
+
+    def list(
+        self,
+        config: Optional[RunnableConfig],
+        *,
+        filter: Optional[dict[str, Any]] = None,
+        before: Optional[RunnableConfig] = None,
+        limit: Optional[int] = None,
+    ) -> Iterator[CheckpointTuple]:
+        """List checkpoints from the database.
+
+        This method retrieves a list of checkpoint tuples from the Postgres database based
+        on the provided config. For ShallowPostgresSaver, this method returns a list with
+        ONLY the most recent checkpoint.
+        """
+        where, args = self._search_where(config, filter, before)
+        query = self.SELECT_SQL + where
+        if limit:
+            query += f" LIMIT {limit}"
+        with self._cursor() as cur:
+            cur.execute(self.SELECT_SQL + where, args, binary=True)
+            for value in cur:
+                checkpoint: Checkpoint = {
+                    **value["checkpoint"],
+                    "channel_values": self._load_blobs(value["channel_values"]),
+                    "pending_sends": [
+                        self.serde.loads_typed((t.decode(), v))
+                        for t, v in value["pending_sends"]
+                    ]
+                    if value["pending_sends"]
+                    else [],
+                }
+                yield CheckpointTuple(
+                    config={
+                        "configurable": {
+                            "thread_id": value["thread_id"],
+                            "checkpoint_ns": value["checkpoint_ns"],
+                            "checkpoint_id": checkpoint["id"],
+                        }
+                    },
+                    checkpoint=checkpoint,
+                    metadata=value["metadata"],
+                    pending_writes=self._load_writes(value["pending_writes"]),
+                )
+
+    def get_tuple(self, config: RunnableConfig) -> Optional[CheckpointTuple]:
+        """Get a checkpoint tuple from the database.
+
+        This method retrieves a checkpoint tuple from the Postgres database based on the
+        provided config (matching the thread ID in the config).
+
+        Args:
+            config: The config to use for retrieving the checkpoint.
+
+        Returns:
+            Optional[CheckpointTuple]: The retrieved checkpoint tuple, or None if no matching checkpoint was found.
+
+        Examples:
+
+            Basic:
+            >>> config = {"configurable": {"thread_id": "1"}}
+            >>> checkpoint_tuple = memory.get_tuple(config)
+            >>> print(checkpoint_tuple)
+            CheckpointTuple(...)
+
+            With timestamp:
+
+            >>> config = {
+            ...    "configurable": {
+            ...        "thread_id": "1",
+            ...        "checkpoint_ns": "",
+            ...        "checkpoint_id": "1ef4f797-8335-6428-8001-8a1503f9b875",
+            ...    }
+            ... }
+            >>> checkpoint_tuple = memory.get_tuple(config)
+            >>> print(checkpoint_tuple)
+            CheckpointTuple(...)
+        """  # noqa
+        thread_id = config["configurable"]["thread_id"]
+        checkpoint_ns = config["configurable"].get("checkpoint_ns", "")
+        args = (thread_id, checkpoint_ns)
+        where = "WHERE thread_id = %s AND checkpoint_ns = %s"
+
+        with self._cursor() as cur:
+            cur.execute(
+                self.SELECT_SQL + where,
+                args,
+                binary=True,
+            )
+
+            for value in cur:
+                checkpoint: Checkpoint = {
+                    **value["checkpoint"],
+                    "channel_values": self._load_blobs(value["channel_values"]),
+                    "pending_sends": [
+                        self.serde.loads_typed((t.decode(), v))
+                        for t, v in value["pending_sends"]
+                    ]
+                    if value["pending_sends"]
+                    else [],
+                }
+                return CheckpointTuple(
+                    config={
+                        "configurable": {
+                            "thread_id": thread_id,
+                            "checkpoint_ns": checkpoint_ns,
+                            "checkpoint_id": checkpoint["id"],
+                        }
+                    },
+                    checkpoint=checkpoint,
+                    metadata=value["metadata"],
+                    pending_writes=self._load_writes(value["pending_writes"]),
+                )
+
+    def put(
+        self,
+        config: RunnableConfig,
+        checkpoint: Checkpoint,
+        metadata: CheckpointMetadata,
+        new_versions: ChannelVersions,
+    ) -> RunnableConfig:
+        """Save a checkpoint to the database.
+
+        This method saves a checkpoint to the Postgres database. The checkpoint is associated
+        with the provided config. For ShallowPostgresSaver, this method saves ONLY the most recent
+        checkpoint and overwrites a previous checkpoint, if it exists.
+
+        Args:
+            config: The config to associate with the checkpoint.
+            checkpoint: The checkpoint to save.
+            metadata: Additional metadata to save with the checkpoint.
+            new_versions: New channel versions as of this write.
+
+        Returns:
+            RunnableConfig: Updated configuration after storing the checkpoint.
+
+        Examples:
+
+            >>> from langgraph.checkpoint.postgres import ShallowPostgresSaver
+            >>> DB_URI = "postgres://postgres:postgres@localhost:5432/postgres?sslmode=disable"
+            >>> with ShallowPostgresSaver.from_conn_string(DB_URI) as memory:
+            >>>     config = {"configurable": {"thread_id": "1", "checkpoint_ns": ""}}
+            >>>     checkpoint = {"ts": "2024-05-04T06:32:42.235444+00:00", "id": "1ef4f797-8335-6428-8001-8a1503f9b875", "channel_values": {"key": "value"}}
+            >>>     saved_config = memory.put(config, checkpoint, {"source": "input", "step": 1, "writes": {"key": "value"}}, {})
+            >>> print(saved_config)
+            {'configurable': {'thread_id': '1', 'checkpoint_ns': '', 'checkpoint_id': '1ef4f797-8335-6428-8001-8a1503f9b875'}}
+        """
+        configurable = config["configurable"].copy()
+        thread_id = configurable.pop("thread_id")
+        checkpoint_ns = configurable.pop("checkpoint_ns")
+
+        copy = checkpoint.copy()
+        next_config = {
+            "configurable": {
+                "thread_id": thread_id,
+                "checkpoint_ns": checkpoint_ns,
+                "checkpoint_id": checkpoint["id"],
+            }
+        }
+
+        with self._cursor(pipeline=True) as cur:
+            cur.execute(
+                """DELETE FROM checkpoint_writes
+                WHERE thread_id = %s AND checkpoint_ns = %s AND checkpoint_id NOT IN (%s, %s)""",
+                (
+                    thread_id,
+                    checkpoint_ns,
+                    checkpoint["id"],
+                    configurable.get("checkpoint_id", ""),
+                ),
+            )
+            cur.executemany(
+                self.UPSERT_CHECKPOINT_BLOBS_SQL,
+                _dump_blobs(
+                    self.serde,
+                    thread_id,
+                    checkpoint_ns,
+                    copy.pop("channel_values"),  # type: ignore[misc]
+                    new_versions,
+                ),
+            )
+            cur.execute(
+                self.UPSERT_CHECKPOINTS_SQL,
+                (
+                    thread_id,
+                    checkpoint_ns,
+                    Jsonb(copy),
+                    Jsonb(get_checkpoint_metadata(config, metadata)),
+                ),
+            )
+        return next_config
+
+    def put_writes(
+        self,
+        config: RunnableConfig,
+        writes: Sequence[tuple[str, Any]],
+        task_id: str,
+        task_path: str = "",
+    ) -> None:
+        """Store intermediate writes linked to a checkpoint.
+
+        This method saves intermediate writes associated with a checkpoint to the Postgres database.
+
+        Args:
+            config: Configuration of the related checkpoint.
+            writes: List of writes to store.
+            task_id: Identifier for the task creating the writes.
+        """
+        query = (
+            self.UPSERT_CHECKPOINT_WRITES_SQL
+            if all(w[0] in WRITES_IDX_MAP for w in writes)
+            else self.INSERT_CHECKPOINT_WRITES_SQL
+        )
+        with self._cursor(pipeline=True) as cur:
+            cur.executemany(
+                query,
+                self._dump_writes(
+                    config["configurable"]["thread_id"],
+                    config["configurable"]["checkpoint_ns"],
+                    config["configurable"]["checkpoint_id"],
+                    task_id,
+                    task_path,
+                    writes,
+                ),
+            )
+
+    @contextmanager
+    def _cursor(self, *, pipeline: bool = False) -> Iterator[Cursor[DictRow]]:
+        """Create a database cursor as a context manager.
+
+        Args:
+            pipeline: whether to use pipeline for the DB operations inside the context manager.
+                Will be applied regardless of whether the ShallowPostgresSaver instance was initialized with a pipeline.
+                If pipeline mode is not supported, will fall back to using transaction context manager.
+        """
+        with _internal.get_connection(self.conn) as conn:
+            if self.pipe:
+                # a connection in pipeline mode can be used concurrently
+                # in multiple threads/coroutines, but only one cursor can be
+                # used at a time
+                try:
+                    with conn.cursor(binary=True, row_factory=dict_row) as cur:
+                        yield cur
+                finally:
+                    if pipeline:
+                        self.pipe.sync()
+            elif pipeline:
+                # a connection not in pipeline mode can only be used by one
+                # thread/coroutine at a time, so we acquire a lock
+                if self.supports_pipeline:
+                    with (
+                        self.lock,
+                        conn.pipeline(),
+                        conn.cursor(binary=True, row_factory=dict_row) as cur,
+                    ):
+                        yield cur
+                else:
+                    # Use connection's transaction context manager when pipeline mode not supported
+                    with (
+                        self.lock,
+                        conn.transaction(),
+                        conn.cursor(binary=True, row_factory=dict_row) as cur,
+                    ):
+                        yield cur
+            else:
+                with self.lock, conn.cursor(binary=True, row_factory=dict_row) as cur:
+                    yield cur
+
+
+class AsyncShallowPostgresSaver(BasePostgresSaver):
+    """A checkpoint saver that uses Postgres to store checkpoints asynchronously.
+
+    This checkpointer ONLY stores the most recent checkpoint and does NOT retain any history.
+    It is meant to be a light-weight drop-in replacement for the AsyncPostgresSaver that
+    supports most of the LangGraph persistence functionality with the exception of time travel.
+    """
+
+    SELECT_SQL = SELECT_SQL
+    MIGRATIONS = MIGRATIONS
+    UPSERT_CHECKPOINT_BLOBS_SQL = UPSERT_CHECKPOINT_BLOBS_SQL
+    UPSERT_CHECKPOINTS_SQL = UPSERT_CHECKPOINTS_SQL
+    UPSERT_CHECKPOINT_WRITES_SQL = UPSERT_CHECKPOINT_WRITES_SQL
+    INSERT_CHECKPOINT_WRITES_SQL = INSERT_CHECKPOINT_WRITES_SQL
+    lock: asyncio.Lock
+
+    def __init__(
+        self,
+        conn: _ainternal.Conn,
+        pipe: Optional[AsyncPipeline] = None,
+        serde: Optional[SerializerProtocol] = None,
+    ) -> None:
+        warnings.warn(
+            "AsyncShallowPostgresSaver is deprecated as of version 2.0.20 and will be removed in 3.0.0. "
+            "Use AsyncPostgresSaver instead, and invoke the graph with `await graph.ainvoke(..., checkpoint_during=False)`.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
+        super().__init__(serde=serde)
+        if isinstance(conn, AsyncConnectionPool) and pipe is not None:
+            raise ValueError(
+                "Pipeline should be used only with a single AsyncConnection, not AsyncConnectionPool."
+            )
+
+        self.conn = conn
+        self.pipe = pipe
+        self.lock = asyncio.Lock()
+        self.loop = asyncio.get_running_loop()
+        self.supports_pipeline = Capabilities().has_pipeline()
+
+    @classmethod
+    @asynccontextmanager
+    async def from_conn_string(
+        cls,
+        conn_string: str,
+        *,
+        pipeline: bool = False,
+        serde: Optional[SerializerProtocol] = None,
+    ) -> AsyncIterator["AsyncShallowPostgresSaver"]:
+        """Create a new AsyncShallowPostgresSaver instance from a connection string.
+
+        Args:
+            conn_string: The Postgres connection info string.
+            pipeline: whether to use AsyncPipeline
+
+        Returns:
+            AsyncShallowPostgresSaver: A new AsyncShallowPostgresSaver instance.
+        """
+        async with await AsyncConnection.connect(
+            conn_string, autocommit=True, prepare_threshold=0, row_factory=dict_row
+        ) as conn:
+            if pipeline:
+                async with conn.pipeline() as pipe:
+                    yield cls(conn=conn, pipe=pipe, serde=serde)
+            else:
+                yield cls(conn=conn, serde=serde)
+
+    async def setup(self) -> None:
+        """Set up the checkpoint database asynchronously.
+
+        This method creates the necessary tables in the Postgres database if they don't
+        already exist and runs database migrations. It MUST be called directly by the user
+        the first time checkpointer is used.
+        """
+        async with self._cursor() as cur:
+            await cur.execute(self.MIGRATIONS[0])
+            results = await cur.execute(
+                "SELECT v FROM checkpoint_migrations ORDER BY v DESC LIMIT 1"
+            )
+            row = await results.fetchone()
+            if row is None:
+                version = -1
+            else:
+                version = row["v"]
+            for v, migration in zip(
+                range(version + 1, len(self.MIGRATIONS)),
+                self.MIGRATIONS[version + 1 :],
+            ):
+                await cur.execute(migration)
+                await cur.execute(f"INSERT INTO checkpoint_migrations (v) VALUES ({v})")
+        if self.pipe:
+            await self.pipe.sync()
+
+    async def alist(
+        self,
+        config: Optional[RunnableConfig],
+        *,
+        filter: Optional[dict[str, Any]] = None,
+        before: Optional[RunnableConfig] = None,
+        limit: Optional[int] = None,
+    ) -> AsyncIterator[CheckpointTuple]:
+        """List checkpoints from the database asynchronously.
+
+        This method retrieves a list of checkpoint tuples from the Postgres database based
+        on the provided config. For ShallowPostgresSaver, this method returns a list with
+        ONLY the most recent checkpoint.
+        """
+        where, args = self._search_where(config, filter, before)
+        query = self.SELECT_SQL + where
+        if limit:
+            query += f" LIMIT {limit}"
+        async with self._cursor() as cur:
+            await cur.execute(self.SELECT_SQL + where, args, binary=True)
+            async for value in cur:
+                checkpoint: Checkpoint = {
+                    **value["checkpoint"],
+                    "channel_values": self._load_blobs(value["channel_values"]),
+                    "pending_sends": [
+                        self.serde.loads_typed((t.decode(), v))
+                        for t, v in value["pending_sends"]
+                    ]
+                    if value["pending_sends"]
+                    else [],
+                }
+                yield CheckpointTuple(
+                    config={
+                        "configurable": {
+                            "thread_id": value["thread_id"],
+                            "checkpoint_ns": value["checkpoint_ns"],
+                            "checkpoint_id": checkpoint["id"],
+                        }
+                    },
+                    checkpoint=checkpoint,
+                    metadata=value["metadata"],
+                    pending_writes=await asyncio.to_thread(
+                        self._load_writes, value["pending_writes"]
+                    ),
+                )
+
+    async def aget_tuple(self, config: RunnableConfig) -> Optional[CheckpointTuple]:
+        """Get a checkpoint tuple from the database asynchronously.
+
+        This method retrieves a checkpoint tuple from the Postgres database based on the
+        provided config (matching the thread ID in the config).
+
+        Args:
+            config: The config to use for retrieving the checkpoint.
+
+        Returns:
+            Optional[CheckpointTuple]: The retrieved checkpoint tuple, or None if no matching checkpoint was found.
+        """
+        thread_id = config["configurable"]["thread_id"]
+        checkpoint_ns = config["configurable"].get("checkpoint_ns", "")
+        args = (thread_id, checkpoint_ns)
+        where = "WHERE thread_id = %s AND checkpoint_ns = %s"
+
+        async with self._cursor() as cur:
+            await cur.execute(
+                self.SELECT_SQL + where,
+                args,
+                binary=True,
+            )
+
+            async for value in cur:
+                checkpoint: Checkpoint = {
+                    **value["checkpoint"],
+                    "channel_values": self._load_blobs(value["channel_values"]),
+                    "pending_sends": [
+                        self.serde.loads_typed((t.decode(), v))
+                        for t, v in value["pending_sends"]
+                    ]
+                    if value["pending_sends"]
+                    else [],
+                }
+                return CheckpointTuple(
+                    config={
+                        "configurable": {
+                            "thread_id": thread_id,
+                            "checkpoint_ns": checkpoint_ns,
+                            "checkpoint_id": checkpoint["id"],
+                        }
+                    },
+                    checkpoint=checkpoint,
+                    metadata=value["metadata"],
+                    pending_writes=await asyncio.to_thread(
+                        self._load_writes, value["pending_writes"]
+                    ),
+                )
+
+    async def aput(
+        self,
+        config: RunnableConfig,
+        checkpoint: Checkpoint,
+        metadata: CheckpointMetadata,
+        new_versions: ChannelVersions,
+    ) -> RunnableConfig:
+        """Save a checkpoint to the database asynchronously.
+
+        This method saves a checkpoint to the Postgres database. The checkpoint is associated
+        with the provided config. For AsyncShallowPostgresSaver, this method saves ONLY the most recent
+        checkpoint and overwrites a previous checkpoint, if it exists.
+
+        Args:
+            config: The config to associate with the checkpoint.
+            checkpoint: The checkpoint to save.
+            metadata: Additional metadata to save with the checkpoint.
+            new_versions: New channel versions as of this write.
+
+        Returns:
+            RunnableConfig: Updated configuration after storing the checkpoint.
+        """
+        configurable = config["configurable"].copy()
+        thread_id = configurable.pop("thread_id")
+        checkpoint_ns = configurable.pop("checkpoint_ns")
+
+        copy = checkpoint.copy()
+        next_config = {
+            "configurable": {
+                "thread_id": thread_id,
+                "checkpoint_ns": checkpoint_ns,
+                "checkpoint_id": checkpoint["id"],
+            }
+        }
+
+        async with self._cursor(pipeline=True) as cur:
+            await cur.execute(
+                """DELETE FROM checkpoint_writes
+                WHERE thread_id = %s AND checkpoint_ns = %s AND checkpoint_id NOT IN (%s, %s)""",
+                (
+                    thread_id,
+                    checkpoint_ns,
+                    checkpoint["id"],
+                    configurable.get("checkpoint_id", ""),
+                ),
+            )
+            await cur.executemany(
+                self.UPSERT_CHECKPOINT_BLOBS_SQL,
+                _dump_blobs(
+                    self.serde,
+                    thread_id,
+                    checkpoint_ns,
+                    copy.pop("channel_values"),  # type: ignore[misc]
+                    new_versions,
+                ),
+            )
+            await cur.execute(
+                self.UPSERT_CHECKPOINTS_SQL,
+                (
+                    thread_id,
+                    checkpoint_ns,
+                    Jsonb(copy),
+                    Jsonb(get_checkpoint_metadata(config, metadata)),
+                ),
+            )
+        return next_config
+
+    async def aput_writes(
+        self,
+        config: RunnableConfig,
+        writes: Sequence[tuple[str, Any]],
+        task_id: str,
+        task_path: str = "",
+    ) -> None:
+        """Store intermediate writes linked to a checkpoint asynchronously.
+
+        This method saves intermediate writes associated with a checkpoint to the database.
+
+        Args:
+            config: Configuration of the related checkpoint.
+            writes: List of writes to store, each as (channel, value) pair.
+            task_id: Identifier for the task creating the writes.
+        """
+        query = (
+            self.UPSERT_CHECKPOINT_WRITES_SQL
+            if all(w[0] in WRITES_IDX_MAP for w in writes)
+            else self.INSERT_CHECKPOINT_WRITES_SQL
+        )
+        params = await asyncio.to_thread(
+            self._dump_writes,
+            config["configurable"]["thread_id"],
+            config["configurable"]["checkpoint_ns"],
+            config["configurable"]["checkpoint_id"],
+            task_id,
+            task_path,
+            writes,
+        )
+        async with self._cursor(pipeline=True) as cur:
+            await cur.executemany(query, params)
+
+    @asynccontextmanager
+    async def _cursor(
+        self, *, pipeline: bool = False
+    ) -> AsyncIterator[AsyncCursor[DictRow]]:
+        """Create a database cursor as a context manager.
+
+        Args:
+            pipeline: whether to use pipeline for the DB operations inside the context manager.
+                Will be applied regardless of whether the AsyncShallowPostgresSaver instance was initialized with a pipeline.
+                If pipeline mode is not supported, will fall back to using transaction context manager.
+        """
+        async with _ainternal.get_connection(self.conn) as conn:
+            if self.pipe:
+                # a connection in pipeline mode can be used concurrently
+                # in multiple threads/coroutines, but only one cursor can be
+                # used at a time
+                try:
+                    async with conn.cursor(binary=True, row_factory=dict_row) as cur:
+                        yield cur
+                finally:
+                    if pipeline:
+                        await self.pipe.sync()
+            elif pipeline:
+                # a connection not in pipeline mode can only be used by one
+                # thread/coroutine at a time, so we acquire a lock
+                if self.supports_pipeline:
+                    async with (
+                        self.lock,
+                        conn.pipeline(),
+                        conn.cursor(binary=True, row_factory=dict_row) as cur,
+                    ):
+                        yield cur
+                else:
+                    # Use connection's transaction context manager when pipeline mode not supported
+                    async with (
+                        self.lock,
+                        conn.transaction(),
+                        conn.cursor(binary=True, row_factory=dict_row) as cur,
+                    ):
+                        yield cur
+            else:
+                async with (
+                    self.lock,
+                    conn.cursor(binary=True, row_factory=dict_row) as cur,
+                ):
+                    yield cur
+
+    def list(
+        self,
+        config: Optional[RunnableConfig],
+        *,
+        filter: Optional[dict[str, Any]] = None,
+        before: Optional[RunnableConfig] = None,
+        limit: Optional[int] = None,
+    ) -> Iterator[CheckpointTuple]:
+        """List checkpoints from the database.
+
+        This method retrieves a list of checkpoint tuples from the Postgres database based
+        on the provided config. For ShallowPostgresSaver, this method returns a list with
+        ONLY the most recent checkpoint.
+        """
+        aiter_ = self.alist(config, filter=filter, before=before, limit=limit)
+        while True:
+            try:
+                yield asyncio.run_coroutine_threadsafe(
+                    anext(aiter_),  # noqa: F821
+                    self.loop,
+                ).result()
+            except StopAsyncIteration:
+                break
+
+    def get_tuple(self, config: RunnableConfig) -> Optional[CheckpointTuple]:
+        """Get a checkpoint tuple from the database.
+
+        This method retrieves a checkpoint tuple from the Postgres database based on the
+        provided config (matching the thread ID in the config).
+
+        Args:
+            config: The config to use for retrieving the checkpoint.
+
+        Returns:
+            Optional[CheckpointTuple]: The retrieved checkpoint tuple, or None if no matching checkpoint was found.
+        """
+        try:
+            # check if we are in the main thread, only bg threads can block
+            # we don't check in other methods to avoid the overhead
+            if asyncio.get_running_loop() is self.loop:
+                raise asyncio.InvalidStateError(
+                    "Synchronous calls to AsyncShallowPostgresSaver are only allowed from a "
+                    "different thread. From the main thread, use the async interface."
+                    "For example, use `await checkpointer.aget_tuple(...)` or `await "
+                    "graph.ainvoke(...)`."
+                )
+        except RuntimeError:
+            pass
+        return asyncio.run_coroutine_threadsafe(
+            self.aget_tuple(config), self.loop
+        ).result()
+
+    def put(
+        self,
+        config: RunnableConfig,
+        checkpoint: Checkpoint,
+        metadata: CheckpointMetadata,
+        new_versions: ChannelVersions,
+    ) -> RunnableConfig:
+        """Save a checkpoint to the database.
+
+        This method saves a checkpoint to the Postgres database. The checkpoint is associated
+        with the provided config. For AsyncShallowPostgresSaver, this method saves ONLY the most recent
+        checkpoint and overwrites a previous checkpoint, if it exists.
+
+        Args:
+            config: The config to associate with the checkpoint.
+            checkpoint: The checkpoint to save.
+            metadata: Additional metadata to save with the checkpoint.
+            new_versions: New channel versions as of this write.
+
+        Returns:
+            RunnableConfig: Updated configuration after storing the checkpoint.
+        """
+        return asyncio.run_coroutine_threadsafe(
+            self.aput(config, checkpoint, metadata, new_versions), self.loop
+        ).result()
+
+    def put_writes(
+        self,
+        config: RunnableConfig,
+        writes: Sequence[tuple[str, Any]],
+        task_id: str,
+        task_path: str = "",
+    ) -> None:
+        """Store intermediate writes linked to a checkpoint.
+
+        This method saves intermediate writes associated with a checkpoint to the database.
+
+        Args:
+            config: Configuration of the related checkpoint.
+            writes: List of writes to store, each as (channel, value) pair.
+            task_id: Identifier for the task creating the writes.
+            task_path: Path of the task creating the writes.
+        """
+        return asyncio.run_coroutine_threadsafe(
+            self.aput_writes(config, writes, task_id, task_path), self.loop
+        ).result()
@@ -1,53 +0,0 @@
-from __future__ import annotations
-
-from collections.abc import Mapping
-from datetime import datetime, timezone
-from typing import Any, Protocol
-
-from langgraph.checkpoint.base import Checkpoint, EmptyChannelError
-from langgraph.checkpoint.base.id import uuid6
-
-
-class ChannelProtocol(Protocol):
-    def checkpoint(self) -> Any | None: ...
-
-
-def empty_checkpoint() -> Checkpoint:
-    return Checkpoint(
-        v=1,
-        id=str(uuid6(clock_seq=-2)),
-        ts=datetime.now(timezone.utc).isoformat(),
-        channel_values={},
-        channel_versions={},
-        versions_seen={},
-    )
-
-
-def create_checkpoint(
-    checkpoint: Checkpoint,
-    channels: Mapping[str, ChannelProtocol] | None,
-    step: int,
-    *,
-    id: str | None = None,
-) -> Checkpoint:
-    """Create a checkpoint for the given channels."""
-    ts = datetime.now(timezone.utc).isoformat()
-    if channels is None:
-        values = checkpoint["channel_values"]
-    else:
-        values = {}
-        for k, v in channels.items():
-            if k not in checkpoint["channel_versions"]:
-                continue
-            try:
-                values[k] = v.checkpoint()
-            except EmptyChannelError:
-                pass
-    return Checkpoint(
-        v=1,
-        ts=ts,
-        id=id or str(uuid6(clock_seq=step)),
-        channel_values=values,
-        channel_versions=checkpoint["channel_versions"],
-        versions_seen=checkpoint["versions_seen"],
-    )
@@ -14,10 +14,14 @@ from langgraph.checkpoint.base import (
    EXCLUDED_METADATA_KEYS,
    Checkpoint,
    CheckpointMetadata,
+    create_checkpoint,
+    empty_checkpoint,
+)
+from langgraph.checkpoint.postgres.aio import (
+    AsyncPostgresSaver,
+    AsyncShallowPostgresSaver,
 )
-from langgraph.checkpoint.postgres.aio import AsyncPostgresSaver
 from langgraph.checkpoint.serde.types import TASKS
-from tests.checkpoint_utils import create_checkpoint, empty_checkpoint
 from tests.conftest import DEFAULT_POSTGRES_URI


@@ -108,11 +112,41 @@ async def _base_saver():
            await conn.execute(f"DROP DATABASE {database}")


+@asynccontextmanager
+async def _shallow_saver():
+    """Fixture for shallow connection mode testing."""
+    database = f"test_{uuid4().hex[:16]}"
+    # create unique db
+    async with await AsyncConnection.connect(
+        DEFAULT_POSTGRES_URI, autocommit=True
+    ) as conn:
+        await conn.execute(f"CREATE DATABASE {database}")
+    try:
+        async with await AsyncConnection.connect(
+            DEFAULT_POSTGRES_URI + database,
+            autocommit=True,
+            prepare_threshold=0,
+            row_factory=dict_row,
+        ) as conn:
+            checkpointer = AsyncShallowPostgresSaver(conn)
+            await checkpointer.setup()
+            yield checkpointer
+    finally:
+        # drop unique db
+        async with await AsyncConnection.connect(
+            DEFAULT_POSTGRES_URI, autocommit=True
+        ) as conn:
+            await conn.execute(f"DROP DATABASE {database}")
+
+
@asynccontextmanager
 async def _saver(name: str):
    if name == "base":
        async with _base_saver() as saver:
            yield saver
+    elif name == "shallow":
+        async with _shallow_saver() as saver:
+            yield saver
    elif name == "pool":
        async with _pool_saver() as saver:
            yield saver
@@ -172,7 +206,7 @@ def test_data():
    }


-@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe"])
+@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe", "shallow"])
 async def test_combined_metadata(saver_name: str, test_data) -> None:
    async with _saver(saver_name) as saver:
        config = {
@@ -199,7 +233,7 @@ async def test_combined_metadata(saver_name: str, test_data) -> None:
        }


-@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe"])
+@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe", "shallow"])
 async def test_asearch(saver_name: str, test_data) -> None:
    async with _saver(saver_name) as saver:
        configs = test_data["configs"]
@@ -250,7 +284,7 @@ async def test_asearch(saver_name: str, test_data) -> None:
        } == {"", "inner"}


-@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe"])
+@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe", "shallow"])
 async def test_null_chars(saver_name: str, test_data) -> None:
    async with _saver(saver_name) as saver:
        config = await saver.aput(
@@ -15,10 +15,11 @@ from langgraph.checkpoint.base import (
    EXCLUDED_METADATA_KEYS,
    Checkpoint,
    CheckpointMetadata,
+    create_checkpoint,
+    empty_checkpoint,
 )
-from langgraph.checkpoint.postgres import PostgresSaver
+from langgraph.checkpoint.postgres import PostgresSaver, ShallowPostgresSaver
 from langgraph.checkpoint.serde.types import TASKS
-from tests.checkpoint_utils import create_checkpoint, empty_checkpoint
 from tests.conftest import DEFAULT_POSTGRES_URI


@@ -97,11 +98,37 @@ def _base_saver():
            conn.execute(f"DROP DATABASE {database}")


+@contextmanager
+def _shallow_saver():
+    """Fixture for regular connection mode testing with a shallow checkpointer."""
+    database = f"test_{uuid4().hex[:16]}"
+    # create unique db
+    with Connection.connect(DEFAULT_POSTGRES_URI, autocommit=True) as conn:
+        conn.execute(f"CREATE DATABASE {database}")
+    try:
+        with Connection.connect(
+            DEFAULT_POSTGRES_URI + database,
+            autocommit=True,
+            prepare_threshold=0,
+            row_factory=dict_row,
+        ) as conn:
+            checkpointer = ShallowPostgresSaver(conn)
+            checkpointer.setup()
+            yield checkpointer
+    finally:
+        # drop unique db
+        with Connection.connect(DEFAULT_POSTGRES_URI, autocommit=True) as conn:
+            conn.execute(f"DROP DATABASE {database}")
+
+
@contextmanager
 def _saver(name: str):
    if name == "base":
        with _base_saver() as saver:
            yield saver
+    elif name == "shallow":
+        with _shallow_saver() as saver:
+            yield saver
    elif name == "pool":
        with _pool_saver() as saver:
            yield saver
@@ -161,7 +188,7 @@ def test_data():
    }


-@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe"])
+@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe", "shallow"])
 def test_combined_metadata(saver_name: str, test_data) -> None:
    with _saver(saver_name) as saver:
        config = {
@@ -188,7 +215,7 @@ def test_combined_metadata(saver_name: str, test_data) -> None:
        }


-@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe"])
+@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe", "shallow"])
 def test_search(saver_name: str, test_data) -> None:
    with _saver(saver_name) as saver:
        configs = test_data["configs"]
@@ -237,7 +264,7 @@ def test_search(saver_name: str, test_data) -> None:
        } == {"", "inner"}


-@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe"])
+@pytest.mark.parametrize("saver_name", ["base", "pool", "pipe", "shallow"])
 def test_null_chars(saver_name: str, test_data) -> None:
    with _saver(saver_name) as saver:
        config = saver.put(
@@ -328,6 +328,7 @@ dev = [
    { name = "mypy" },
    { name = "numpy" },
    { name = "pandas" },
+    { name = "pandas-stubs", specifier = ">=2.2.2.240807" },
    { name = "pytest" },
    { name = "pytest-asyncio" },
    { name = "pytest-mock" },
@@ -536,7 +536,7 @@ class SqliteSaver(BaseCheckpointSaver[str]):
        """
        raise NotImplementedError(_AIO_ERROR_MSG)

-    def get_next_version(self, current: str | None) -> str:
+    def get_next_version(self, current: str | None, channel: None) -> str:
        """Generate the next version ID for a channel.

        This method creates a new version identifier for a channel based on its current version.
@@ -591,7 +591,7 @@ class AsyncSqliteSaver(BaseCheckpointSaver[str]):
            )
            await self.conn.commit()

-    def get_next_version(self, current: str | None) -> str:
+    def get_next_version(self, current: str | None, channel: None) -> str:
        """Generate the next version ID for a channel.

        This method creates a new version identifier for a channel based on its current version.
@@ -1,53 +0,0 @@
-from __future__ import annotations
-
-from collections.abc import Mapping
-from datetime import datetime, timezone
-from typing import Any, Protocol
-
-from langgraph.checkpoint.base import Checkpoint, EmptyChannelError
-from langgraph.checkpoint.base.id import uuid6
-
-
-class ChannelProtocol(Protocol):
-    def checkpoint(self) -> Any | None: ...
-
-
-def empty_checkpoint() -> Checkpoint:
-    return Checkpoint(
-        v=1,
-        id=str(uuid6(clock_seq=-2)),
-        ts=datetime.now(timezone.utc).isoformat(),
-        channel_values={},
-        channel_versions={},
-        versions_seen={},
-    )
-
-
-def create_checkpoint(
-    checkpoint: Checkpoint,
-    channels: Mapping[str, ChannelProtocol] | None,
-    step: int,
-    *,
-    id: str | None = None,
-) -> Checkpoint:
-    """Create a checkpoint for the given channels."""
-    ts = datetime.now(timezone.utc).isoformat()
-    if channels is None:
-        values = checkpoint["channel_values"]
-    else:
-        values = {}
-        for k, v in channels.items():
-            if k not in checkpoint["channel_versions"]:
-                continue
-            try:
-                values[k] = v.checkpoint()
-            except EmptyChannelError:
-                pass
-    return Checkpoint(
-        v=1,
-        ts=ts,
-        id=id or str(uuid6(clock_seq=step)),
-        channel_values=values,
-        channel_versions=checkpoint["channel_versions"],
-        versions_seen=checkpoint["versions_seen"],
-    )
@@ -6,9 +6,10 @@ from langchain_core.runnables import RunnableConfig
 from langgraph.checkpoint.base import (
    Checkpoint,
    CheckpointMetadata,
+    create_checkpoint,
+    empty_checkpoint,
 )
 from langgraph.checkpoint.sqlite.aio import AsyncSqliteSaver
-from tests.checkpoint_utils import create_checkpoint, empty_checkpoint


 class TestAsyncSqliteSaver:
@@ -6,10 +6,11 @@ from langchain_core.runnables import RunnableConfig
 from langgraph.checkpoint.base import (
    Checkpoint,
    CheckpointMetadata,
+    create_checkpoint,
+    empty_checkpoint,
 )
 from langgraph.checkpoint.sqlite import SqliteSaver
 from langgraph.checkpoint.sqlite.utils import _metadata_predicate, search_where
-from tests.checkpoint_utils import create_checkpoint, empty_checkpoint


 class TestSqliteSaver:
@@ -340,6 +340,7 @@ dev = [
    { name = "mypy" },
    { name = "numpy" },
    { name = "pandas" },
+    { name = "pandas-stubs", specifier = ">=2.2.2.240807" },
    { name = "pytest" },
    { name = "pytest-asyncio" },
    { name = "pytest-mock" },
@@ -1,6 +1,6 @@
 from __future__ import annotations

-from collections.abc import AsyncIterator, Iterator, Sequence
+from collections.abc import AsyncIterator, Iterator, Mapping, Sequence
 from typing import (  # noqa: UP035
    Any,
    Generic,
@@ -13,6 +13,7 @@ from typing import (  # noqa: UP035

 from langchain_core.runnables import RunnableConfig

+from langgraph.checkpoint.base.id import uuid6
 from langgraph.checkpoint.serde.base import SerializerProtocol, maybe_add_typed_methods
 from langgraph.checkpoint.serde.jsonplus import JsonPlusSerializer
 from langgraph.checkpoint.serde.types import (
@@ -20,6 +21,7 @@ from langgraph.checkpoint.serde.types import (
    INTERRUPT,
    RESUME,
    SCHEDULED,
+    ChannelProtocol,
 )

 V = TypeVar("V", int, float, str)
@@ -89,6 +91,7 @@ def copy_checkpoint(checkpoint: Checkpoint) -> Checkpoint:
        channel_values=checkpoint["channel_values"].copy(),
        channel_versions=checkpoint["channel_versions"].copy(),
        versions_seen={k: v.copy() for k, v in checkpoint["versions_seen"].items()},
+        pending_sends=checkpoint.get("pending_sends", []).copy(),
    )


@@ -125,6 +128,15 @@ class BaseCheckpointSaver(Generic[V]):
    ) -> None:
        self.serde = maybe_add_typed_methods(serde or self.serde)

+    @property
+    def config_specs(self) -> list:
+        """Define the configuration options for the checkpoint saver.
+
+        Returns:
+            list: List of configuration field specs.
+        """
+        return []
+
    def get(self, config: RunnableConfig) -> Checkpoint | None:
        """Fetch a checkpoint using the given configuration.

@@ -334,7 +346,7 @@ class BaseCheckpointSaver(Generic[V]):
        """
        raise NotImplementedError

-    def get_next_version(self, current: V | None) -> V:
+    def get_next_version(self, current: V | None, channel: None) -> V:
        """Generate the next version ID for a channel.

        Default is to use integer versions, incrementing by 1. If you override, you can use str/int/float versions,
@@ -342,6 +354,7 @@ class BaseCheckpointSaver(Generic[V]):

        Args:
            current: The current version identifier (int, float, or str).
+            channel: Deprecated argument, kept for backwards compatibility.

        Returns:
            V: The next version identifier, which must be increasing.
@@ -404,3 +417,54 @@ EXCLUDED_METADATA_KEYS = {
    "checkpoint_ns",
    "checkpoint_map",
 }
+
+# --- below are deprecated utilities used by past versions of LangGraph ---
+
+LATEST_VERSION = 2
+
+
+def empty_checkpoint() -> Checkpoint:
+    from datetime import datetime, timezone
+
+    return Checkpoint(
+        v=LATEST_VERSION,
+        id=str(uuid6(clock_seq=-2)),
+        ts=datetime.now(timezone.utc).isoformat(),
+        channel_values={},
+        channel_versions={},
+        versions_seen={},
+        pending_sends=[],
+    )
+
+
+def create_checkpoint(
+    checkpoint: Checkpoint,
+    channels: Mapping[str, ChannelProtocol] | None,
+    step: int,
+    *,
+    id: str | None = None,
+) -> Checkpoint:
+    """Create a checkpoint for the given channels."""
+    from datetime import datetime, timezone
+
+    ts = datetime.now(timezone.utc).isoformat()
+    if channels is None:
+        values = checkpoint["channel_values"]
+    else:
+        values = {}
+        for k, v in channels.items():
+            if k not in checkpoint["channel_versions"]:
+                continue
+            try:
+                values[k] = v.checkpoint()
+            except EmptyChannelError:
+                pass
+    return Checkpoint(
+        v=LATEST_VERSION,
+        ts=ts,
+        id=id or str(uuid6(clock_seq=step)),
+        channel_values=values,
+        channel_versions=checkpoint["channel_versions"],
+        versions_seen=checkpoint["versions_seen"],
+        pending_sends=checkpoint.get("pending_sends", []),
+    )
@@ -512,7 +512,7 @@ class InMemorySaver(
        """
        return self.delete_thread(thread_id)

-    def get_next_version(self, current: str | None) -> str:
+    def get_next_version(self, current: str | None, channel: None) -> str:
        if current is None:
            current_v = 0
        elif isinstance(current, int):
@@ -1,4 +1,13 @@
-from typing import Any, Protocol, TypeVar, runtime_checkable
+from collections.abc import Sequence
+from typing import (
+    Any,
+    Optional,
+    Protocol,
+    TypeVar,
+    runtime_checkable,
+)
+
+from typing_extensions import Self

 ERROR = "__error__"
 SCHEDULED = "__scheduled__"
@@ -11,6 +20,25 @@ Update = TypeVar("Update", contravariant=True)
 C = TypeVar("C")


+class ChannelProtocol(Protocol[Value, Update, C]):
+    # Mirrors langgraph.channels.base.BaseChannel
+    @property
+    def ValueType(self) -> Any: ...
+
+    @property
+    def UpdateType(self) -> Any: ...
+
+    def checkpoint(self) -> Optional[C]: ...
+
+    def from_checkpoint(self, checkpoint: Optional[C]) -> Self: ...
+
+    def update(self, values: Sequence[Update]) -> bool: ...
+
+    def get(self) -> Value: ...
+
+    def consume(self) -> bool: ...
+
+
@runtime_checkable
 class SendProtocol(Protocol):
    # Mirrors langgraph.constants.Send
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"

 [project]
 name = "langgraph-checkpoint"
-version = "2.0.26"
+version = "2.1.0"
 description = "Library with base interfaces for LangGraph checkpoint savers."
 authors = []
 requires-python = ">=3.9"
@@ -31,6 +31,7 @@ dev = [
  "dataclasses-json",
  "numpy",
  "pandas",
+  "pandas-stubs>=2.2.2.240807",
 ]

 [tool.hatch.build.targets.wheel]
@@ -1,53 +0,0 @@
-from __future__ import annotations
-
-from collections.abc import Mapping
-from datetime import datetime, timezone
-from typing import Any, Protocol
-
-from langgraph.checkpoint.base import Checkpoint, EmptyChannelError
-from langgraph.checkpoint.base.id import uuid6
-
-
-class ChannelProtocol(Protocol):
-    def checkpoint(self) -> Any | None: ...
-
-
-def empty_checkpoint() -> Checkpoint:
-    return Checkpoint(
-        v=1,
-        id=str(uuid6(clock_seq=-2)),
-        ts=datetime.now(timezone.utc).isoformat(),
-        channel_values={},
-        channel_versions={},
-        versions_seen={},
-    )
-
-
-def create_checkpoint(
-    checkpoint: Checkpoint,
-    channels: Mapping[str, ChannelProtocol] | None,
-    step: int,
-    *,
-    id: str | None = None,
-) -> Checkpoint:
-    """Create a checkpoint for the given channels."""
-    ts = datetime.now(timezone.utc).isoformat()
-    if channels is None:
-        values = checkpoint["channel_values"]
-    else:
-        values = {}
-        for k, v in channels.items():
-            if k not in checkpoint["channel_versions"]:
-                continue
-            try:
-                values[k] = v.checkpoint()
-            except EmptyChannelError:
-                pass
-    return Checkpoint(
-        v=1,
-        ts=ts,
-        id=id or str(uuid6(clock_seq=step)),
-        channel_values=values,
-        channel_versions=checkpoint["channel_versions"],
-        versions_seen=checkpoint["versions_seen"],
-    )
@@ -6,12 +6,10 @@ from langchain_core.runnables import RunnableConfig
 from langgraph.checkpoint.base import (
    Checkpoint,
    CheckpointMetadata,
-)
-from langgraph.checkpoint.memory import InMemorySaver
-from tests.checkpoint_utils import (
    create_checkpoint,
    empty_checkpoint,
 )
+from langgraph.checkpoint.memory import InMemorySaver


 class TestMemorySaver:
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"

 [project]
 name = "langgraph-cli"
-version = "0.3.2"
+version = "0.3.3"
 description = "CLI for interacting with LangGraph API"
 authors = []
 requires-python = ">=3.9"
@@ -501,7 +501,7 @@ wheels = [

 [[package]]
 name = "langgraph-cli"
-version = "0.3.2"
+version = "0.3.3"
 source = { editable = "." }
 dependencies = [
    { name = "click" },
@@ -1,14 +1,15 @@
 from langgraph.channels.any_value import AnyValue
 from langgraph.channels.binop import BinaryOperatorAggregate
 from langgraph.channels.ephemeral_value import EphemeralValue
-from langgraph.channels.last_value import LastValue, LastValueAfterFinish
+from langgraph.channels.last_value import LastValue
 from langgraph.channels.topic import Topic
+from langgraph.channels.untracked_value import UntrackedValue

 __all__ = [
    "LastValue",
-    "LastValueAfterFinish",
    "Topic",
    "BinaryOperatorAggregate",
+    "UntrackedValue",
    "EphemeralValue",
    "AnyValue",
 ]
@@ -0,0 +1,66 @@
+from collections.abc import Sequence
+from typing import Generic
+
+from typing_extensions import Self
+
+from langgraph.channels.base import BaseChannel, Value
+from langgraph.constants import MISSING
+from langgraph.errors import EmptyChannelError, InvalidUpdateError
+
+
+class UntrackedValue(Generic[Value], BaseChannel[Value, Value, Value]):
+    """Stores the last value received, never checkpointed."""
+
+    __slots__ = ("value", "guard")
+
+    def __init__(self, typ: type[Value], guard: bool = True) -> None:
+        super().__init__(typ)
+        self.guard = guard
+        self.value = MISSING
+
+    def __eq__(self, value: object) -> bool:
+        return isinstance(value, UntrackedValue) and value.guard == self.guard
+
+    @property
+    def ValueType(self) -> type[Value]:
+        """The type of the value stored in the channel."""
+        return self.typ
+
+    @property
+    def UpdateType(self) -> type[Value]:
+        """The type of the update received by the channel."""
+        return self.typ
+
+    def copy(self) -> Self:
+        """Return a copy of the channel."""
+        empty = self.__class__(self.typ, self.guard)
+        empty.key = self.key
+        empty.value = self.value
+        return empty
+
+    def checkpoint(self) -> Value:
+        return MISSING
+
+    def from_checkpoint(self, checkpoint: Value) -> Self:
+        empty = self.__class__(self.typ, self.guard)
+        empty.key = self.key
+        return empty
+
+    def update(self, values: Sequence[Value]) -> bool:
+        if len(values) == 0:
+            return False
+        if len(values) != 1 and self.guard:
+            raise InvalidUpdateError(
+                f"At key '{self.key}': UntrackedValue(guard=True) can receive only one value per step. Use guard=False if you want to store any one of multiple values."
+            )
+
+        self.value = values[-1]
+        return True
+
+    def get(self) -> Value:
+        if self.value is MISSING:
+            raise EmptyChannelError()
+        return self.value
+
+    def is_available(self) -> bool:
+        return self.value is not MISSING
@@ -38,7 +38,7 @@ from langgraph.pregel.read import PregelNode
 from langgraph.pregel.write import ChannelWrite, ChannelWriteEntry
 from langgraph.store.base import BaseStore
 from langgraph.types import _DC_KWARGS, CachePolicy, RetryPolicy, StreamMode
-from langgraph.warnings import LangGraphDeprecatedSinceV10
+from langgraph.warnings import LangGraphDeprecatedSinceV05


 class TaskFunction(Generic[P, T]):
@@ -179,7 +179,7 @@ def task(
    if (retry := kwargs.get("retry", UNSET)) is not UNSET:
        warnings.warn(
            "`retry` is deprecated and will be removed. Please use `retry_policy` instead.",
-            category=LangGraphDeprecatedSinceV10,
+            category=LangGraphDeprecatedSinceV05,
        )
        if retry_policy is None:
            retry_policy = retry  # type: ignore[assignment]
@@ -383,7 +383,7 @@ class entrypoint:
        if (retry := kwargs.get("retry", UNSET)) is not UNSET:
            warnings.warn(
                "`retry` is deprecated and will be removed. Please use `retry_policy` instead.",
-                category=LangGraphDeprecatedSinceV10,
+                category=LangGraphDeprecatedSinceV05,
            )
            if retry_policy is None:
                retry_policy = retry  # type: ignore[assignment]
@@ -1,11 +1,12 @@
 from langgraph.constants import END, START
-from langgraph.graph.message import MessagesState, add_messages
+from langgraph.graph.message import MessageGraph, MessagesState, add_messages
 from langgraph.graph.state import StateGraph

 __all__ = [
    "END",
    "START",
    "StateGraph",
+    "MessageGraph",
    "add_messages",
    "MessagesState",
 ]
@@ -25,6 +25,7 @@ from langchain_core.messages import (
 from typing_extensions import TypedDict

 from langgraph.constants import CONF, CONFIG_KEY_SEND
+from langgraph.graph.state import StateGraph

 Messages = Union[list[MessageLikeRepresentation], MessageLikeRepresentation]

@@ -226,6 +227,57 @@ def add_messages(
    return merged


+class MessageGraph(StateGraph):
+    """A StateGraph where every node receives a list of messages as input and returns one or more messages as output.
+
+    MessageGraph is a subclass of StateGraph whose entire state is a single, append-only* list of messages.
+    Each node in a MessageGraph takes a list of messages as input and returns zero or more
+    messages as output. The `add_messages` function is used to merge the output messages from each node
+    into the existing list of messages in the graph's state.
+
+    Examples:
+        ```pycon
+        >>> from langgraph.graph.message import MessageGraph
+        ...
+        >>> builder = MessageGraph()
+        >>> builder.add_node("chatbot", lambda state: [("assistant", "Hello!")])
+        >>> builder.set_entry_point("chatbot")
+        >>> builder.set_finish_point("chatbot")
+        >>> builder.compile().invoke([("user", "Hi there.")])
+        [HumanMessage(content="Hi there.", id='...'), AIMessage(content="Hello!", id='...')]
+        ```
+
+        ```pycon
+        >>> from langchain_core.messages import AIMessage, HumanMessage, ToolMessage
+        >>> from langgraph.graph.message import MessageGraph
+        ...
+        >>> builder = MessageGraph()
+        >>> builder.add_node(
+        ...     "chatbot",
+        ...     lambda state: [
+        ...         AIMessage(
+        ...             content="Hello!",
+        ...             tool_calls=[{"name": "search", "id": "123", "args": {"query": "X"}}],
+        ...         )
+        ...     ],
+        ... )
+        >>> builder.add_node(
+        ...     "search", lambda state: [ToolMessage(content="Searching...", tool_call_id="123")]
+        ... )
+        >>> builder.set_entry_point("chatbot")
+        >>> builder.add_edge("chatbot", "search")
+        >>> builder.set_finish_point("search")
+        >>> builder.compile().invoke([HumanMessage(content="Hi there. Can you search for X?")])
+        {'messages': [HumanMessage(content="Hi there. Can you search for X?", id='b8b7d8f4-7f4d-4f4d-9c1d-f8b8d8f4d9c1'),
+                     AIMessage(content="Hello!", id='f4d9c1d8-8d8f-4d9c-b8b7-d8f4f4d9c1d8'),
+                     ToolMessage(content="Searching...", id='d8f4f4d9-c1d8-4f4d-b8b7-d8f4f4d9c1d8', tool_call_id="123")]}
+        ```
+    """
+
+    def __init__(self) -> None:
+        super().__init__(Annotated[list[AnyMessage], add_messages])  # type: ignore[arg-type]
+
+
 class MessagesState(TypedDict):
    messages: Annotated[list[AnyMessage], add_messages]

@@ -86,7 +86,7 @@ from langgraph.utils.fields import (
 )
 from langgraph.utils.pydantic import create_model
 from langgraph.utils.runnable import coerce_to_runnable
-from langgraph.warnings import LangGraphDeprecatedSinceV10
+from langgraph.warnings import LangGraphDeprecatedSinceV05

 logger = logging.getLogger(__name__)

@@ -160,6 +160,7 @@ StateNode: TypeAlias = Union[
    _NodeWithConfigWriter[StateT_contra],
    _NodeWithConfigStore[StateT_contra],
    _NodeWithConfigWriterStore[StateT_contra],
+    Runnable[StateT_contra, Any],
 ]


@@ -261,7 +262,7 @@ class StateGraph(Generic[StateT, InputT, OutputT]):
        if (input_ := kwargs.get("input", UNSET)) is not UNSET:
            warnings.warn(
                "`input` is deprecated and will be removed. Please use `input_schema` instead.",
-                category=LangGraphDeprecatedSinceV10,
+                category=LangGraphDeprecatedSinceV05,
                stacklevel=2,
            )
            if input_schema is None:
@@ -270,7 +271,7 @@ class StateGraph(Generic[StateT, InputT, OutputT]):
        if (output := kwargs.get("output", UNSET)) is not UNSET:
            warnings.warn(
                "`output` is deprecated and will be removed. Please use `output_schema` instead.",
-                category=LangGraphDeprecatedSinceV10,
+                category=LangGraphDeprecatedSinceV05,
                stacklevel=2,
            )
            if output_schema is None:
@@ -436,7 +437,7 @@ class StateGraph(Generic[StateT, InputT, OutputT]):
        if (retry := kwargs.get("retry", UNSET)) is not UNSET:
            warnings.warn(
                "`retry` is deprecated and will be removed. Please use `retry_policy` instead.",
-                category=LangGraphDeprecatedSinceV10,
+                category=LangGraphDeprecatedSinceV05,
            )
            if retry_policy is None:
                retry_policy = retry  # type: ignore[assignment]
@@ -444,7 +445,7 @@ class StateGraph(Generic[StateT, InputT, OutputT]):
        if (input_ := kwargs.get("input", UNSET)) is not UNSET:
            warnings.warn(
                "`input` is deprecated and will be removed. Please use `input_schema` instead.",
-                category=LangGraphDeprecatedSinceV10,
+                category=LangGraphDeprecatedSinceV05,
            )
            if input_schema is None:
                input_schema = cast(Union[type[InputT], None], input_)
@@ -535,7 +536,7 @@ class StateGraph(Generic[StateT, InputT, OutputT]):
        if input_schema is not None:
            self._add_schema(input_schema)
        self.nodes[node] = StateNodeSpec(
-            coerce_to_runnable(action, name=node, trace=False),  # type: ignore
+            coerce_to_runnable(action, name=node, trace=False),
            metadata,
            input=input_schema or self.state_schema,
            retry_policy=retry_policy,
@@ -1101,6 +1102,7 @@ class CompiledStateGraph(

    def _migrate_checkpoint(self, checkpoint: Checkpoint) -> None:
        """Migrate a checkpoint to new channel layout."""
+        super()._migrate_checkpoint(checkpoint)

        values = checkpoint["channel_values"]
        versions = checkpoint["channel_versions"]
@@ -32,7 +32,6 @@ from langgraph.checkpoint.base import (
    BaseCheckpointSaver,
    Checkpoint,
    CheckpointTuple,
-    copy_checkpoint,
 )
 from langgraph.config import get_config
 from langgraph.constants import (
@@ -79,6 +78,7 @@ from langgraph.pregel.algo import (
 from langgraph.pregel.call import identifier
 from langgraph.pregel.checkpoint import (
    channels_from_checkpoint,
+    copy_checkpoint,
    create_checkpoint,
    empty_checkpoint,
 )
@@ -908,7 +908,12 @@ class Pregel(PregelProtocol[StateT, InputT, OutputT], Generic[StateT, InputT, Ou

    def _migrate_checkpoint(self, checkpoint: Checkpoint) -> None:
        """Migrate a saved checkpoint to new channel layout."""
-        pass
+        if checkpoint["v"] < 4 and checkpoint.get("pending_sends"):
+            pending_sends: list[Send] = checkpoint.pop("pending_sends")
+            checkpoint["channel_values"][TASKS] = pending_sends
+            checkpoint["channel_versions"][TASKS] = max(
+                checkpoint["channel_versions"].values()
+            )

    def _prepare_state_snapshot(
        self,
@@ -2298,7 +2303,8 @@ class Pregel(PregelProtocol[StateT, InputT, OutputT], Generic[StateT, InputT, Ou
                - `"custom"`: Emit custom data from inside nodes or tasks using `StreamWriter`.
                - `"messages"`: Emit LLM messages token-by-token together with metadata for any LLM invocations inside nodes or tasks.
                    Will be emitted as 2-tuples `(LLM token, metadata)`.
-                - `"debug"`: Emit debug events with as much information as possible for each step.
+                - `"checkpoints"`: Emit an event when a checkpoint is created, in the same format as returned by get_state().
+                - `"tasks"`: Emit events when tasks start and finish, including their results and errors.

                You can pass a list as the `stream_mode` parameter to stream multiple modes at once.
                The streamed outputs will be tuples of `(mode, data)`.
@@ -83,7 +83,7 @@ from langgraph.types import (
 )
 from langgraph.utils.config import merge_configs, patch_config

-GetNextVersion = Callable[[Optional[V]], V]
+GetNextVersion = Callable[[Optional[V], None], V]
 SUPPORTS_EXC_NOTES = sys.version_info >= (3, 11)


@@ -214,7 +214,7 @@ def local_read(
    return values


-def increment(current: int | None) -> int:
+def increment(current: int | None, channel: None) -> int:
    """Default channel versioning function, increments the current int version."""
    return current + 1 if current is not None else 1

@@ -265,7 +265,8 @@ def apply_writes(
        next_version = get_next_version(
            max(checkpoint["channel_versions"].values())
            if checkpoint["channel_versions"]
-            else None
+            else None,
+            None,
        )

    # Consume all channels that were read
@@ -71,3 +71,14 @@ def channels_from_checkpoint(
        },
        managed_specs,
    )
+
+
+def copy_checkpoint(checkpoint: Checkpoint) -> Checkpoint:
+    return Checkpoint(
+        v=checkpoint["v"],
+        ts=checkpoint["ts"],
+        id=checkpoint["id"],
+        channel_values=checkpoint["channel_values"].copy(),
+        channel_versions=checkpoint["channel_versions"].copy(),
+        versions_seen={k: v.copy() for k, v in checkpoint["versions_seen"].items()},
+    )
@@ -3,13 +3,8 @@ from __future__ import annotations
 from collections import defaultdict
 from collections.abc import Iterable, Iterator, Mapping, Sequence
 from dataclasses import asdict
-from datetime import datetime, timezone
 from pprint import pformat
-from typing import (
-    Any,
-    Literal,
-    Union,
-)
+from typing import Any
 from uuid import UUID

 from langchain_core.runnables.config import RunnableConfig
@@ -17,7 +12,7 @@ from langchain_core.utils.input import get_bolded_text, get_colored_text
 from typing_extensions import TypedDict

 from langgraph.channels.base import BaseChannel
-from langgraph.checkpoint.base import Checkpoint, CheckpointMetadata, PendingWrite
+from langgraph.checkpoint.base import CheckpointMetadata, PendingWrite
 from langgraph.constants import (
    CONF,
    CONFIG_KEY_CHECKPOINT_NS,
@@ -66,82 +61,43 @@ class CheckpointPayload(TypedDict):
    tasks: list[CheckpointTask]


-class DebugOutputBase(TypedDict):
-    timestamp: str
-    step: int
-
-
-class DebugOutputTask(DebugOutputBase):
-    type: Literal["task"]
-    payload: TaskPayload
-
-
-class DebugOutputTaskResult(DebugOutputBase):
-    type: Literal["task_result"]
-    payload: TaskResultPayload
-
-
-class DebugOutputCheckpoint(DebugOutputBase):
-    type: Literal["checkpoint"]
-    payload: CheckpointPayload
-
-
-DebugOutput = Union[DebugOutputTask, DebugOutputTaskResult, DebugOutputCheckpoint]
-
-
 TASK_NAMESPACE = UUID("6ba7b831-9dad-11d1-80b4-00c04fd430c8")


-def map_debug_tasks(
-    step: int, tasks: Iterable[PregelExecutableTask]
-) -> Iterator[DebugOutputTask]:
+def map_debug_tasks(tasks: Iterable[PregelExecutableTask]) -> Iterator[TaskPayload]:
    """Produce "task" events for stream_mode=debug."""
-    ts = datetime.now(timezone.utc).isoformat()
    for task in tasks:
        if task.config is not None and TAG_HIDDEN in task.config.get("tags", []):
            continue

        yield {
-            "type": "task",
-            "timestamp": ts,
-            "step": step,
-            "payload": {
-                "id": task.id,
-                "name": task.name,
-                "input": task.input,
-                "triggers": task.triggers,
-            },
+            "id": task.id,
+            "name": task.name,
+            "input": task.input,
+            "triggers": task.triggers,
        }


 def map_debug_task_results(
-    step: int,
    task_tup: tuple[PregelExecutableTask, Sequence[tuple[str, Any]]],
    stream_keys: str | Sequence[str],
-) -> Iterator[DebugOutputTaskResult]:
+) -> Iterator[TaskResultPayload]:
    """Produce "task_result" events for stream_mode=debug."""
    stream_channels_list = (
        [stream_keys] if isinstance(stream_keys, str) else stream_keys
    )
    task, writes = task_tup
    yield {
-        "type": "task_result",
-        "timestamp": datetime.now(timezone.utc).isoformat(),
-        "step": step,
-        "payload": {
-            "id": task.id,
-            "name": task.name,
-            "error": next((w[1] for w in writes if w[0] == ERROR), None),
-            "result": [
-                w for w in writes if w[0] in stream_channels_list or w[0] == RETURN
-            ],
-            "interrupts": [
-                asdict(v)
-                for w in writes
-                if w[0] == INTERRUPT
-                for v in (w[1] if isinstance(w[1], Sequence) else [w[1]])
-            ],
-        },
+        "id": task.id,
+        "name": task.name,
+        "error": next((w[1] for w in writes if w[0] == ERROR), None),
+        "result": [w for w in writes if w[0] in stream_channels_list or w[0] == RETURN],
+        "interrupts": [
+            asdict(v)
+            for w in writes
+            if w[0] == INTERRUPT
+            for v in (w[1] if isinstance(w[1], Sequence) else [w[1]])
+        ],
    }


@@ -159,17 +115,15 @@ def rm_pregel_keys(config: RunnableConfig | None) -> RunnableConfig | None:


 def map_debug_checkpoint(
-    step: int,
    config: RunnableConfig,
    channels: Mapping[str, BaseChannel],
    stream_channels: str | Sequence[str],
    metadata: CheckpointMetadata,
-    checkpoint: Checkpoint,
    tasks: Iterable[PregelExecutableTask],
    pending_writes: list[PendingWrite],
    parent_config: RunnableConfig | None,
    output_keys: str | Sequence[str],
-) -> Iterator[DebugOutputCheckpoint]:
+) -> Iterator[CheckpointPayload]:
    """Produce "checkpoint" events for stream_mode=debug."""

    parent_ns = config[CONF].get(CONFIG_KEY_CHECKPOINT_NS, "")
@@ -193,42 +147,35 @@ def map_debug_checkpoint(
        }

    yield {
-        "type": "checkpoint",
-        "timestamp": checkpoint["ts"],
-        "step": step,
-        "payload": {
-            "config": rm_pregel_keys(patch_checkpoint_map(config, metadata)),
-            "parent_config": rm_pregel_keys(
-                patch_checkpoint_map(parent_config, metadata)
-            ),
-            "values": read_channels(channels, stream_channels),
-            "metadata": metadata,
-            "next": [t.name for t in tasks],
-            "tasks": [
-                {
-                    "id": t.id,
-                    "name": t.name,
-                    "error": t.error,
-                    "state": t.state,
-                }
-                if t.error
-                else {
-                    "id": t.id,
-                    "name": t.name,
-                    "result": t.result,
-                    "interrupts": tuple(asdict(i) for i in t.interrupts),
-                    "state": t.state,
-                }
-                if t.result
-                else {
-                    "id": t.id,
-                    "name": t.name,
-                    "interrupts": tuple(asdict(i) for i in t.interrupts),
-                    "state": t.state,
-                }
-                for t in tasks_w_writes(tasks, pending_writes, task_states, output_keys)
-            ],
-        },
+        "config": rm_pregel_keys(patch_checkpoint_map(config, metadata)),
+        "parent_config": rm_pregel_keys(patch_checkpoint_map(parent_config, metadata)),
+        "values": read_channels(channels, stream_channels),
+        "metadata": metadata,
+        "next": [t.name for t in tasks],
+        "tasks": [
+            {
+                "id": t.id,
+                "name": t.name,
+                "error": t.error,
+                "state": t.state,
+            }
+            if t.error
+            else {
+                "id": t.id,
+                "name": t.name,
+                "result": t.result,
+                "interrupts": tuple(asdict(i) for i in t.interrupts),
+                "state": t.state,
+            }
+            if t.result
+            else {
+                "id": t.id,
+                "name": t.name,
+                "interrupts": tuple(asdict(i) for i in t.interrupts),
+                "state": t.state,
+            }
+            for t in tasks_w_writes(tasks, pending_writes, task_states, output_keys)
+        ],
    }


@@ -11,6 +11,7 @@ from contextlib import (
    AsyncExitStack,
    ExitStack,
 )
+from datetime import datetime, timezone
 from inspect import signature
 from types import TracebackType
 from typing import (
@@ -37,7 +38,6 @@ from langgraph.checkpoint.base import (
    CheckpointMetadata,
    CheckpointTuple,
    PendingWrite,
-    copy_checkpoint,
 )
 from langgraph.constants import (
    CONF,
@@ -84,6 +84,7 @@ from langgraph.pregel.algo import (
 )
 from langgraph.pregel.checkpoint import (
    channels_from_checkpoint,
+    copy_checkpoint,
    create_checkpoint,
    empty_checkpoint,
 )
@@ -118,6 +119,7 @@ from langgraph.types import (
    PregelScratchpad,
    RetryPolicy,
    StreamChunk,
+    StreamMode,
    StreamProtocol,
 )
 from langgraph.utils.config import patch_configurable
@@ -421,7 +423,7 @@ class PregelLoop:
            ),
        ):
            # produce debug output
-            self._emit("debug", map_debug_tasks, self.step, [pushed])
+            self._emit("tasks", map_debug_tasks, [pushed])
            # debug flag
            if self.debug:
                print_step_tasks(self.step, [pushed])
@@ -471,9 +473,8 @@ class PregelLoop:
        # produce debug output
        if self._checkpointer_put_after_previous is not None:
            self._emit(
-                "debug",
+                "checkpoints",
                map_debug_checkpoint,
-                self.step - 1,  # printing checkpoint for previous step
                {
                    **self.checkpoint_config,
                    CONF: {
@@ -484,7 +485,6 @@ class PregelLoop:
                self.channels,
                self.stream_keys,
                self.checkpoint_metadata,
-                self.checkpoint,
                self.tasks.values(),
                self.checkpoint_pending_writes,
                self.prev_checkpoint_config,
@@ -508,7 +508,7 @@ class PregelLoop:
            raise GraphInterrupt()

        # produce debug output
-        self._emit("debug", map_debug_tasks, self.step, self.tasks.values())
+        self._emit("tasks", map_debug_tasks, self.tasks.values())

        # debug flag
        if self.debug:
@@ -833,17 +833,39 @@ class PregelLoop:

    def _emit(
        self,
-        mode: str,
+        mode: StreamMode,
        values: Callable[P, Iterator[Any]],
        *args: P.args,
        **kwargs: P.kwargs,
    ) -> None:
        if self.stream is None:
            return
-        if mode not in self.stream.modes:
+        debug_remap = mode in ("checkpoints", "tasks") and "debug" in self.stream.modes
+        if mode not in self.stream.modes and not debug_remap:
            return
        for v in values(*args, **kwargs):
-            self.stream((self.checkpoint_ns, mode, v))
+            if mode in self.stream.modes:
+                self.stream((self.checkpoint_ns, mode, v))
+            # "debug" mode is "checkpoints" or "tasks" with a wrapper dict
+            if debug_remap:
+                self.stream(
+                    (
+                        self.checkpoint_ns,
+                        "debug",
+                        {
+                            "step": self.step - 1
+                            if mode == "checkpoints"
+                            else self.step,
+                            "timestamp": datetime.now(timezone.utc).isoformat(),
+                            "type": "checkpoint"
+                            if mode == "checkpoints"
+                            else "task_result"
+                            if "result" in v
+                            else "task",
+                            "payload": v,
+                        },
+                    )
+                )

    def output_writes(
        self, task_id: str, writes: WritesT, *, cached: bool = False
@@ -884,9 +906,8 @@ class PregelLoop:
                )
            if not cached:
                self._emit(
-                    "debug",
+                    "tasks",
                    map_debug_task_results,
-                    self.step,
                    (task, writes),
                    self.stream_keys,
                )
@@ -46,7 +46,9 @@ Checkpointer = Union[None, bool, BaseCheckpointSaver]
 - False disables checkpointing, even if the parent graph has a checkpointer.
 - None inherits checkpointer from the parent graph."""

-StreamMode = Literal["values", "updates", "debug", "messages", "custom"]
+StreamMode = Literal[
+    "values", "updates", "checkpoints", "tasks", "debug", "messages", "custom"
+]
 """How the stream method should emit outputs.

 - `"values"`: Emit all values in the state after each step, including interrupts.
@@ -55,7 +57,9 @@ StreamMode = Literal["values", "updates", "debug", "messages", "custom"]
    If multiple updates are made in the same step (e.g. multiple nodes are run) then those updates are emitted separately.
 - `"custom"`: Emit custom data using from inside nodes or tasks using `StreamWriter`.
 - `"messages"`: Emit LLM messages token-by-token together with metadata for any LLM invocations inside nodes or tasks.
- `"debug"`: Emit debug events with as much information as possible for each step.
+- `"checkpoints"`: Emit an event when a checkpoint is created, in the same format as returned by get_state().
+- `"tasks"`: Emit events when tasks start and finish, including their results and errors.
+- `"debug"`: Emit "checlkpoints" and "tasks" events, for debugging purposes.
 """

 StreamWriter = Callable[[Any], None]
@@ -41,8 +41,8 @@ class LangGraphDeprecationWarning(DeprecationWarning):
        return message


-class LangGraphDeprecatedSinceV10(LangGraphDeprecationWarning):
-    """A specific `LangGraphDeprecationWarning` subclass defining functionality deprecated since LangGraph v1.0.0"""
+class LangGraphDeprecatedSinceV05(LangGraphDeprecationWarning):
+    """A specific `LangGraphDeprecationWarning` subclass defining functionality deprecated since LangGraph v0.5.0"""

    def __init__(self, message: str, *args: object) -> None:
-        super().__init__(message, *args, since=(1, 0), expected_removal=(2, 0))
+        super().__init__(message, *args, since=(0, 5), expected_removal=(2, 0))
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"

 [project]
 name = "langgraph"
-version = "0.4.7"
+version = "0.5.0rc1"
 description = "Building stateful, multi-actor applications with LLMs"
 authors = []
 requires-python = ">=3.9"
@@ -13,9 +13,9 @@ license = "MIT"
 license-files = ['LICENSE']
 dependencies = [
    "langchain-core>=0.1",
-    "langgraph-checkpoint>=2.0.26",
+    "langgraph-checkpoint>=2.1.0",
    "langgraph-sdk>=0.1.42",
-    "langgraph-prebuilt>=0.2.0",
+    "langgraph-prebuilt>=0.5.0rc0",
    "xxhash>=3.5.0",
    "pydantic>=2.7.4",
 ]
@@ -12,6 +12,7 @@ from langgraph.checkpoint.base import BaseCheckpointSaver
 from langgraph.store.base import BaseStore
 from tests.conftest_checkpointer import (
    _checkpointer_memory,
+    _checkpointer_memory_migrate_sends,
    _checkpointer_postgres,
    _checkpointer_postgres_aio,
    _checkpointer_postgres_aio_pipe,
@@ -125,6 +126,7 @@ async def async_store(request: pytest.FixtureRequest) -> AsyncIterator[BaseStore
    if NO_DOCKER
    else [
        "memory",
+        "memory_migrate_sends",
        "sqlite",
        "sqlite_aes",
        "postgres",
@@ -139,6 +141,9 @@ def sync_checkpointer(
    if checkpointer_name == "memory":
        with _checkpointer_memory() as checkpointer:
            yield checkpointer
+    elif checkpointer_name == "memory_migrate_sends":
+        with _checkpointer_memory_migrate_sends() as checkpointer:
+            yield checkpointer
    elif checkpointer_name == "sqlite":
        with _checkpointer_sqlite() as checkpointer:
            yield checkpointer
@@ -14,7 +14,10 @@ from langgraph.checkpoint.sqlite.aio import AsyncSqliteSaver

 pytest.register_assert_rewrite("tests.memory_assert")

-from tests.memory_assert import MemorySaverAssertImmutable  # noqa: E402
+from tests.memory_assert import (  # noqa: E402
+    MemorySaverAssertImmutable,
+    MemorySaverNeedsPendingSendsMigration,
+)

 DEFAULT_POSTGRES_URI = "postgres://postgres:postgres@localhost:5442/"

@@ -24,6 +27,11 @@ def _checkpointer_memory():
    yield MemorySaverAssertImmutable()


+@contextmanager
+def _checkpointer_memory_migrate_sends():
+    yield MemorySaverNeedsPendingSendsMigration()
+
+
@contextmanager
 def _checkpointer_sqlite():
    with SqliteSaver.from_conn_string(":memory:") as checkpointer:
@@ -187,6 +195,7 @@ async def _checkpointer_postgres_aio_pool():

 __all__ = [
    "_checkpointer_memory",
+    "_checkpointer_memory_migrate_sends",
    "_checkpointer_sqlite",
    "_checkpointer_sqlite_aes",
    "_checkpointer_postgres",
@@ -7,6 +7,7 @@ from typing import Any, Optional
 from langchain_core.runnables import RunnableConfig

 from langgraph.checkpoint.base import (
+    BaseCheckpointSaver,
    ChannelVersions,
    Checkpoint,
    CheckpointMetadata,
@@ -14,6 +15,7 @@ from langgraph.checkpoint.base import (
    SerializerProtocol,
 )
 from langgraph.checkpoint.memory import InMemorySaver, PersistentDict
+from langgraph.constants import TASKS


 class NoopSerializer(SerializerProtocol):
@@ -24,6 +26,28 @@ class NoopSerializer(SerializerProtocol):
        return "type", obj


+class MemorySaverNeedsPendingSendsMigration(BaseCheckpointSaver):
+    def __init__(self) -> None:
+        self.saver = InMemorySaver()
+
+    def __getattribute__(self, name):
+        if name in ("saver", "__class__", "get_tuple"):
+            return object.__getattribute__(self, name)
+        return getattr(self.saver, name)
+
+    def get_tuple(self, config):
+        if tup := self.saver.get_tuple(config):
+            if tup.checkpoint["v"] == 4 and tup.checkpoint["channel_values"].get(TASKS):
+                tup.checkpoint["v"] = 3
+                tup.checkpoint["pending_sends"] = tup.checkpoint["channel_values"].pop(
+                    TASKS
+                )
+                tup.checkpoint["channel_versions"].pop(TASKS)
+                for seen in tup.checkpoint["versions_seen"].values():
+                    seen.pop(TASKS, None)
+        return tup
+
+
 class MemorySaverAssertImmutable(InMemorySaver):
    storage_for_copies: defaultdict[str, dict[str, dict[str, Checkpoint]]]

@@ -7,12 +7,9 @@ from typing import Annotated, Literal, Optional, Union
 import pytest
 from typing_extensions import TypedDict

-from langgraph.checkpoint.base import (
-    BaseCheckpointSaver,
-    CheckpointTuple,
-    copy_checkpoint,
-)
+from langgraph.checkpoint.base import BaseCheckpointSaver, CheckpointTuple
 from langgraph.graph.state import StateGraph
+from langgraph.pregel.checkpoint import copy_checkpoint
 from langgraph.types import Command, Interrupt, PregelTask, StateSnapshot, interrupt
 from langgraph.utils.config import patch_configurable
 from tests.any_int import AnyInt
@@ -4,7 +4,7 @@ from typing_extensions import TypedDict
 from langgraph.func import entrypoint, task
 from langgraph.graph import StateGraph
 from langgraph.types import RetryPolicy
-from langgraph.warnings import LangGraphDeprecatedSinceV10
+from langgraph.warnings import LangGraphDeprecatedSinceV05


 class PlainState(TypedDict): ...
@@ -14,7 +14,7 @@ def test_add_node_retry_arg() -> None:
    builder = StateGraph(PlainState)

    with pytest.warns(
-        LangGraphDeprecatedSinceV10,
+        LangGraphDeprecatedSinceV05,
        match="`retry` is deprecated and will be removed. Please use `retry_policy` instead.",
    ):
        builder.add_node("test_node", lambda state: state, retry=RetryPolicy())  # type: ignore[arg-type]
@@ -22,7 +22,7 @@ def test_add_node_retry_arg() -> None:

 def test_task_retry_arg() -> None:
    with pytest.warns(
-        LangGraphDeprecatedSinceV10,
+        LangGraphDeprecatedSinceV05,
        match="`retry` is deprecated and will be removed. Please use `retry_policy` instead.",
    ):

@@ -33,7 +33,7 @@ def test_task_retry_arg() -> None:

 def test_entrypoint_retry_arg() -> None:
    with pytest.warns(
-        LangGraphDeprecatedSinceV10,
+        LangGraphDeprecatedSinceV05,
        match="`retry` is deprecated and will be removed. Please use `retry_policy` instead.",
    ):

@@ -44,7 +44,7 @@ def test_entrypoint_retry_arg() -> None:

 def test_state_graph_input_schema() -> None:
    with pytest.warns(
-        LangGraphDeprecatedSinceV10,
+        LangGraphDeprecatedSinceV05,
        match="`input` is deprecated and will be removed. Please use `input_schema` instead.",
    ):
        StateGraph(PlainState, input=PlainState)  # type: ignore[arg-type]
@@ -52,7 +52,7 @@ def test_state_graph_input_schema() -> None:

 def test_state_graph_output_schema() -> None:
    with pytest.warns(
-        LangGraphDeprecatedSinceV10,
+        LangGraphDeprecatedSinceV05,
        match="`output` is deprecated and will be removed. Please use `output_schema` instead.",
    ):
        StateGraph(PlainState, output=PlainState)  # type: ignore[arg-type]
@@ -62,7 +62,7 @@ def test_add_node_input_schema() -> None:
    builder = StateGraph(PlainState)

    with pytest.warns(
-        LangGraphDeprecatedSinceV10,
+        LangGraphDeprecatedSinceV05,
        match="`input` is deprecated and will be removed. Please use `input_schema` instead.",
    ):
        builder.add_node("test_node", lambda state: state, input=PlainState)  # type: ignore[arg-type]
@@ -16,13 +16,14 @@ from langchain_core.runnables import RunnableConfig, RunnablePick
 from pytest_mock import MockerFixture
 from typing_extensions import TypedDict

-from langgraph.channels.ephemeral_value import EphemeralValue
 from langgraph.channels.last_value import LastValue
+from langgraph.channels.untracked_value import UntrackedValue
 from langgraph.checkpoint.base import BaseCheckpointSaver
 from langgraph.constants import END, PULL, PUSH, START
-from langgraph.graph.message import add_messages
+from langgraph.graph.message import MessageGraph, add_messages
 from langgraph.graph.state import StateGraph
 from langgraph.prebuilt.chat_agent_executor import create_react_agent
+from langgraph.prebuilt.tool_node import ToolNode
 from langgraph.pregel import NodeBuilder, Pregel
 from langgraph.types import PregelTask, Send, StateSnapshot, StreamWriter
 from tests.any_int import AnyInt
@@ -499,7 +500,7 @@ async def test_conditional_graph_state(async_checkpointer: BaseCheckpointSaver)
    from langchain_core.tools import tool

    class AgentState(TypedDict):
-        input: Annotated[str, EphemeralValue]
+        input: Annotated[str, UntrackedValue]
        agent_outcome: Optional[Union[AgentAction, AgentFinish]]
        intermediate_steps: Annotated[list[tuple[AgentAction, str]], operator.add]

@@ -574,6 +575,7 @@ async def test_conditional_graph_state(async_checkpointer: BaseCheckpointSaver)
    app = workflow.compile()

    assert await app.ainvoke({"input": "what is weather in sf"}) == {
+        "input": "what is weather in sf",
        "intermediate_steps": [
            [
                AgentAction(
@@ -2082,7 +2084,417 @@ async def test_state_graph_packets(async_checkpointer: BaseCheckpointSaver) -> N
    )


+async def test_message_graph(async_checkpointer: BaseCheckpointSaver) -> None:
+    from langchain_core.language_models.fake_chat_models import (
+        FakeMessagesListChatModel,
+    )
+    from langchain_core.messages import AIMessage, HumanMessage
+    from langchain_core.tools import tool
+
+    class FakeFuntionChatModel(FakeMessagesListChatModel):
+        def bind_functions(self, functions: list):
+            return self
+
+    @tool()
+    def search_api(query: str) -> str:
+        """Searches the API for the query."""
+        return f"result for {query}"
+
+    tools = [search_api]
+
+    model = FakeFuntionChatModel(
+        responses=[
+            AIMessage(
+                content="",
+                tool_calls=[
+                    {
+                        "id": "tool_call123",
+                        "name": "search_api",
+                        "args": {"query": "query"},
+                    }
+                ],
+                id="ai1",
+            ),
+            AIMessage(
+                content="",
+                tool_calls=[
+                    {
+                        "id": "tool_call456",
+                        "name": "search_api",
+                        "args": {"query": "another"},
+                    }
+                ],
+                id="ai2",
+            ),
+            AIMessage(content="answer", id="ai3"),
+        ]
+    )
+
+    # Define the function that determines whether to continue or not
+    def should_continue(messages):
+        last_message = messages[-1]
+        # If there is no function call, then we finish
+        if not last_message.tool_calls:
+            return "end"
+        # Otherwise if there is, we continue
+        else:
+            return "continue"
+
+    # Define a new graph
+    workflow = MessageGraph()
+
+    # Define the two nodes we will cycle between
+    workflow.add_node("agent", model)
+    workflow.add_node("tools", ToolNode(tools))
+
+    # Set the entrypoint as `agent`
+    # This means that this node is the first one called
+    workflow.set_entry_point("agent")
+
+    # We now add a conditional edge
+    workflow.add_conditional_edges(
+        # First, we define the start node. We use `agent`.
+        # This means these are the edges taken after the `agent` node is called.
+        "agent",
+        # Next, we pass in the function that will determine which node is called next.
+        should_continue,
+        # Finally we pass in a mapping.
+        # The keys are strings, and the values are other nodes.
+        # END is a special node marking that the graph should finish.
+        # What will happen is we will call `should_continue`, and then the output of that
+        # will be matched against the keys in this mapping.
+        # Based on which one it matches, that node will then be called.
+        {
+            # If `tools`, then we call the tool node.
+            "continue": "tools",
+            # Otherwise we finish.
+            "end": END,
+        },
+    )
+
+    # We now add a normal edge from `tools` to `agent`.
+    # This means that after `tools` is called, `agent` node is called next.
+    workflow.add_edge("tools", "agent")
+
+    # Finally, we compile it!
+    # This compiles it into a LangChain Runnable,
+    # meaning you can use it as you would any other runnable
+    app = workflow.compile()
+
+    assert await app.ainvoke(HumanMessage(content="what is weather in sf")) == [
+        _AnyIdHumanMessage(
+            content="what is weather in sf",
+        ),
+        AIMessage(
+            content="",
+            tool_calls=[
+                {
+                    "id": "tool_call123",
+                    "name": "search_api",
+                    "args": {"query": "query"},
+                }
+            ],
+            id="ai1",  # respects ids passed in
+        ),
+        _AnyIdToolMessage(
+            content="result for query",
+            name="search_api",
+            tool_call_id="tool_call123",
+        ),
+        AIMessage(
+            content="",
+            tool_calls=[
+                {
+                    "id": "tool_call456",
+                    "name": "search_api",
+                    "args": {"query": "another"},
+                }
+            ],
+            id="ai2",
+        ),
+        _AnyIdToolMessage(
+            content="result for another",
+            name="search_api",
+            tool_call_id="tool_call456",
+        ),
+        AIMessage(content="answer", id="ai3"),
+    ]
+
+    assert [
+        c async for c in app.astream([HumanMessage(content="what is weather in sf")])
+    ] == [
+        {
+            "agent": AIMessage(
+                content="",
+                tool_calls=[
+                    {
+                        "id": "tool_call123",
+                        "name": "search_api",
+                        "args": {"query": "query"},
+                    }
+                ],
+                id="ai1",
+            )
+        },
+        {
+            "tools": [
+                _AnyIdToolMessage(
+                    content="result for query",
+                    name="search_api",
+                    tool_call_id="tool_call123",
+                )
+            ]
+        },
+        {
+            "agent": AIMessage(
+                content="",
+                tool_calls=[
+                    {
+                        "id": "tool_call456",
+                        "name": "search_api",
+                        "args": {"query": "another"},
+                    }
+                ],
+                id="ai2",
+            )
+        },
+        {
+            "tools": [
+                _AnyIdToolMessage(
+                    content="result for another",
+                    name="search_api",
+                    tool_call_id="tool_call456",
+                )
+            ]
+        },
+        {"agent": AIMessage(content="answer", id="ai3")},
+    ]
+
+    app_w_interrupt = workflow.compile(
+        checkpointer=async_checkpointer,
+        interrupt_after=["agent"],
+    )
+    config = {"configurable": {"thread_id": "1"}}
+
+    assert [
+        c
+        async for c in app_w_interrupt.astream(
+            HumanMessage(content="what is weather in sf"), config
+        )
+    ] == [
+        {
+            "agent": AIMessage(
+                content="",
+                tool_calls=[
+                    {
+                        "id": "tool_call123",
+                        "name": "search_api",
+                        "args": {"query": "query"},
+                    }
+                ],
+                id="ai1",
+            )
+        },
+        {"__interrupt__": ()},
+    ]
+
+    tup = await app_w_interrupt.checkpointer.aget_tuple(config)
+    assert await app_w_interrupt.aget_state(config) == StateSnapshot(
+        values=[
+            _AnyIdHumanMessage(content="what is weather in sf"),
+            AIMessage(
+                content="",
+                tool_calls=[
+                    {
+                        "id": "tool_call123",
+                        "name": "search_api",
+                        "args": {"query": "query"},
+                    }
+                ],
+                id="ai1",
+            ),
+        ],
+        tasks=(PregelTask(AnyStr(), "tools", (PULL, "tools")),),
+        next=("tools",),
+        config=tup.config,
+        created_at=tup.checkpoint["ts"],
+        metadata={
+            "parents": {},
+            "source": "loop",
+            "step": 1,
+            "thread_id": "1",
+        },
+        parent_config=None,
+        interrupts=(),
+    )
+
+    # modify ai message
+    last_message = (await app_w_interrupt.aget_state(config)).values[-1]
+    last_message.tool_calls[0]["args"] = {"query": "a different query"}
+    await app_w_interrupt.aupdate_state(config, last_message)
+
+    # message was replaced instead of appended
+    tup = await app_w_interrupt.checkpointer.aget_tuple(config)
+    assert await app_w_interrupt.aget_state(config) == StateSnapshot(
+        values=[
+            _AnyIdHumanMessage(content="what is weather in sf"),
+            AIMessage(
+                content="",
+                id="ai1",
+                tool_calls=[
+                    {
+                        "id": "tool_call123",
+                        "name": "search_api",
+                        "args": {"query": "a different query"},
+                    }
+                ],
+            ),
+        ],
+        tasks=(PregelTask(AnyStr(), "tools", (PULL, "tools")),),
+        next=("tools",),
+        config=tup.config,
+        created_at=tup.checkpoint["ts"],
+        metadata={
+            "parents": {},
+            "source": "update",
+            "step": 2,
+            "thread_id": "1",
+        },
+        parent_config=(
+            [c async for c in app_w_interrupt.checkpointer.alist(config, limit=2)][
+                -1
+            ].config
+        ),
+        interrupts=(),
+    )
+
+    assert [c async for c in app_w_interrupt.astream(None, config)] == [
+        {
+            "tools": [
+                _AnyIdToolMessage(
+                    content="result for a different query",
+                    name="search_api",
+                    tool_call_id="tool_call123",
+                )
+            ]
+        },
+        {
+            "agent": AIMessage(
+                content="",
+                tool_calls=[
+                    {
+                        "id": "tool_call456",
+                        "name": "search_api",
+                        "args": {"query": "another"},
+                    }
+                ],
+                id="ai2",
+            )
+        },
+        {"__interrupt__": ()},
+    ]
+
+    tup = await app_w_interrupt.checkpointer.aget_tuple(config)
+    assert await app_w_interrupt.aget_state(config) == StateSnapshot(
+        values=[
+            _AnyIdHumanMessage(content="what is weather in sf"),
+            AIMessage(
+                content="",
+                id="ai1",
+                tool_calls=[
+                    {
+                        "id": "tool_call123",
+                        "name": "search_api",
+                        "args": {"query": "a different query"},
+                    }
+                ],
+            ),
+            _AnyIdToolMessage(
+                content="result for a different query",
+                name="search_api",
+                tool_call_id="tool_call123",
+            ),
+            AIMessage(
+                content="",
+                tool_calls=[
+                    {
+                        "id": "tool_call456",
+                        "name": "search_api",
+                        "args": {"query": "another"},
+                    }
+                ],
+                id="ai2",
+            ),
+        ],
+        tasks=(PregelTask(AnyStr(), "tools", (PULL, "tools")),),
+        next=("tools",),
+        config=tup.config,
+        created_at=tup.checkpoint["ts"],
+        metadata={
+            "parents": {},
+            "source": "loop",
+            "step": 4,
+            "thread_id": "1",
+        },
+        parent_config=(
+            [c async for c in app_w_interrupt.checkpointer.alist(config, limit=2)][
+                -1
+            ].config
+        ),
+        interrupts=(),
+    )
+
+    await app_w_interrupt.aupdate_state(
+        config,
+        AIMessage(content="answer", id="ai2"),
+    )
+
+    # replaces message even if object identity is different, as long as id is the same
+    tup = await app_w_interrupt.checkpointer.aget_tuple(config)
+    assert await app_w_interrupt.aget_state(config) == StateSnapshot(
+        values=[
+            _AnyIdHumanMessage(content="what is weather in sf"),
+            AIMessage(
+                content="",
+                id="ai1",
+                tool_calls=[
+                    {
+                        "id": "tool_call123",
+                        "name": "search_api",
+                        "args": {"query": "a different query"},
+                    }
+                ],
+            ),
+            _AnyIdToolMessage(
+                content="result for a different query",
+                name="search_api",
+                tool_call_id="tool_call123",
+            ),
+            AIMessage(content="answer", id="ai2"),
+        ],
+        tasks=(),
+        next=(),
+        config=tup.config,
+        created_at=tup.checkpoint["ts"],
+        metadata={
+            "parents": {},
+            "source": "update",
+            "step": 5,
+            "thread_id": "1",
+        },
+        parent_config=(
+            [c async for c in app_w_interrupt.checkpointer.alist(config, limit=2)][
+                -1
+            ].config
+        ),
+        interrupts=(),
+    )
+
+
 async def test_in_one_fan_out_out_one_graph_state() -> None:
+    def sorted_add(x: list[str], y: list[str]) -> list[str]:
+        return sorted(operator.add(x, y))
+
    class State(TypedDict, total=False):
        query: str
        answer: str
@@ -46,7 +46,7 @@ from langgraph.constants import CONFIG_KEY_NODE_FINISHED, ERROR, PULL, START
 from langgraph.errors import InvalidUpdateError, ParentCommand
 from langgraph.func import entrypoint, task
 from langgraph.graph import END, StateGraph
-from langgraph.graph.message import MessagesState, add_messages
+from langgraph.graph.message import MessageGraph, MessagesState, add_messages
 from langgraph.prebuilt.tool_node import ToolNode
 from langgraph.pregel import (
    GraphRecursionError,
@@ -159,7 +159,7 @@ def test_checkpoint_errors() -> None:
            raise ValueError("Faulty put_writes")

    class FaultyVersionCheckpointer(InMemorySaver):
-        def get_next_version(self, current: Optional[int]) -> int:
+        def get_next_version(self, current: Optional[int], channel: None) -> int:
            raise ValueError("Faulty get_next_version")

    def logic(inp: str) -> str:
@@ -3994,14 +3994,9 @@ def test_checkpoint_metadata(sync_checkpointer: BaseCheckpointSaver) -> None:
 def test_remove_message_via_state_update(
    sync_checkpointer: BaseCheckpointSaver,
 ) -> None:
-    from langchain_core.messages import (
-        AIMessage,
-        AnyMessage,
-        HumanMessage,
-        RemoveMessage,
-    )
+    from langchain_core.messages import AIMessage, HumanMessage, RemoveMessage

-    workflow = StateGraph(Annotated[list[AnyMessage], add_messages])
+    workflow = MessageGraph()
    workflow.add_node(
        "chatbot",
        lambda state: [
@@ -4032,14 +4027,9 @@ def test_remove_message_via_state_update(


 def test_remove_message_from_node():
-    from langchain_core.messages import (
-        AIMessage,
-        AnyMessage,
-        HumanMessage,
-        RemoveMessage,
-    )
+    from langchain_core.messages import AIMessage, HumanMessage, RemoveMessage

-    workflow = StateGraph(Annotated[list[AnyMessage], add_messages])
+    workflow = MessageGraph()
    workflow.add_node(
        "chatbot",
        lambda state: [
@@ -103,7 +103,7 @@ async def test_checkpoint_errors() -> None:
            raise ValueError("Faulty put_writes")

    class FaultyVersionCheckpointer(InMemorySaver):
-        def get_next_version(self, current: Optional[int]) -> int:
+        def get_next_version(self, current: Optional[int], channel: None) -> int:
            raise ValueError("Faulty get_next_version")

    def logic(inp: str) -> str:
@@ -1201,7 +1201,7 @@ wheels = [

 [[package]]
 name = "langgraph"
-version = "0.4.7"
+version = "0.5.0rc1"
 source = { editable = "." }
 dependencies = [
    { name = "langchain-core" },
@@ -1310,7 +1310,7 @@ wheels = [

 [[package]]
 name = "langgraph-checkpoint"
-version = "2.0.26"
+version = "2.1.0"
 source = { editable = "../checkpoint" }
 dependencies = [
    { name = "langchain-core" },
@@ -1330,6 +1330,7 @@ dev = [
    { name = "mypy" },
    { name = "numpy" },
    { name = "pandas" },
+    { name = "pandas-stubs", specifier = ">=2.2.2.240807" },
    { name = "pytest" },
    { name = "pytest-asyncio" },
    { name = "pytest-mock" },
@@ -1422,7 +1423,7 @@ inmem = [

 [[package]]
 name = "langgraph-prebuilt"
-version = "0.2.2"
+version = "0.5.0rc0"
 source = { editable = "../prebuilt" }
 dependencies = [
    { name = "langchain-core" },
@@ -591,7 +591,7 @@ def create_react_agent(
        workflow = StateGraph(state_schema, config_schema=config_schema)
        workflow.add_node(
            "agent",
-            RunnableCallable(call_model, acall_model),  # type: ignore[call-overload]
+            RunnableCallable(call_model, acall_model),
            input_schema=input_schema,
        )
        if pre_model_hook is not None:
@@ -610,7 +610,7 @@ def create_react_agent(
        if response_format is not None:
            workflow.add_node(
                "generate_structured_response",
-                RunnableCallable(  # type: ignore[call-overload]
+                RunnableCallable(
                    generate_structured_response,
                    agenerate_structured_response,
                ),
@@ -660,10 +660,10 @@ def create_react_agent(
    # Define the two nodes we will cycle between
    workflow.add_node(
        "agent",
-        RunnableCallable(call_model, acall_model),  # type: ignore[call-overload]
+        RunnableCallable(call_model, acall_model),
        input_schema=input_schema,
    )
-    workflow.add_node("tools", tool_node)  # type: ignore[call-overload]
+    workflow.add_node("tools", tool_node)

    # Optionally add a pre-model hook node that will be called
    # every time before the "agent" (LLM-calling node)
@@ -693,7 +693,7 @@ def create_react_agent(
    if response_format is not None:
        workflow.add_node(
            "generate_structured_response",
-            RunnableCallable(  # type: ignore[call-overload]
+            RunnableCallable(
                generate_structured_response,
                agenerate_structured_response,
            ),
@@ -629,7 +629,7 @@ def tools_condition(

    Args:
        state: The state to check for
-            tool calls. Must have a list of messages or have the
+            tool calls. Must have a list of messages (MessageGraph) or have the
            "messages" key (StateGraph).

    Returns:
@@ -2,7 +2,7 @@
 in a langchain graph. It applies a pydantic schema to tool_calls in the models' outputs,
 and returns a ToolMessage with the validated content. If the schema is not valid, it
 returns a ToolMessage with the error message. The ValidationNode can be used in a
-StateGraph with a "messages" key. If multiple tool calls are
+StateGraph with a "messages" key or in a MessageGraph. If multiple tool calls are
 requested, they will be run in parallel.
 """

@@ -49,7 +49,7 @@ def _default_format_error(
 class ValidationNode(RunnableCallable):
    """A node that validates all tools requests from the last AIMessage.

-    It can be used in StateGraph with a "messages" key.
+    It can be used either in StateGraph with a "messages" key or in MessageGraph.

    !!! note

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"

 [project]
 name = "langgraph-prebuilt"
-version = "0.2.2"
+version = "0.5.0rc0"
 description = "Library with high-level APIs for creating and executing LangGraph agents and tools."
 authors = []
 requires-python = ">=3.9"
@@ -12,7 +12,7 @@ readme = "README.md"
 license = "MIT"
 license-files = ['LICENSE']
 dependencies = [
-    "langgraph-checkpoint>=2.0.10",
+    "langgraph-checkpoint>=2.1.0",
    "langchain-core>=0.3.22",
 ]

@@ -13,9 +13,9 @@ from langgraph.checkpoint.base import (
    CheckpointMetadata,
    CheckpointTuple,
    SerializerProtocol,
-    copy_checkpoint,
 )
 from langgraph.checkpoint.memory import InMemorySaver, PersistentDict
+from langgraph.pregel.checkpoint import copy_checkpoint


 class NoopSerializer(SerializerProtocol):
@@ -320,7 +320,7 @@ wheels = [

 [[package]]
 name = "langgraph"
-version = "0.4.7"
+version = "0.5.0rc0"
 source = { editable = "../langgraph" }
 dependencies = [
    { name = "langchain-core" },
@@ -371,7 +371,7 @@ dev = [

 [[package]]
 name = "langgraph-checkpoint"
-version = "2.0.26"
+version = "2.1.0"
 source = { editable = "../checkpoint" }
 dependencies = [
    { name = "langchain-core" },
@@ -391,6 +391,7 @@ dev = [
    { name = "mypy" },
    { name = "numpy" },
    { name = "pandas" },
+    { name = "pandas-stubs", specifier = ">=2.2.2.240807" },
    { name = "pytest" },
    { name = "pytest-asyncio" },
    { name = "pytest-mock" },
@@ -463,7 +464,7 @@ dev = [

 [[package]]
 name = "langgraph-prebuilt"
-version = "0.2.2"
+version = "0.5.0rc0"
 source = { editable = "." }
 dependencies = [
    { name = "langchain-core" },
@@ -36,7 +36,15 @@ Represents the status of a thread:
 """

 StreamMode = Literal[
-    "values", "messages", "updates", "events", "debug", "custom", "messages-tuple"
+    "values",
+    "messages",
+    "updates",
+    "events",
+    "tasks",
+    "checkpoints",
+    "debug",
+    "custom",
+    "messages-tuple",
 ]
 """
 Defines the mode of streaming:
@@ -44,6 +52,8 @@ Defines the mode of streaming:
 - "messages": Stream complete messages.
 - "updates": Stream updates to the state.
 - "events": Stream events occurring during execution.
+- "checkpoints": Stream checkpoints as they are created.
+- "tasks": Stream task start and finish events.
 - "debug": Stream detailed debug information.
 - "custom": Stream custom events.
 """
Author	SHA1	Message	Date
Eugene Yurtsev	0b8634b4c6	x	2025-06-25 10:25:06 -04:00
Eugene Yurtsev	bc3ef7f913	x	2025-06-24 14:47:20 -04:00
Eugene Yurtsev	0dda1b4b1e	x	2025-06-24 12:35:31 -04:00
Eugene Yurtsev	d22c2c4dac	x	2025-06-24 11:17:21 -04:00
Eugene Yurtsev	db03dccb2b	x	2025-06-24 09:56:19 -04:00
Eugene Yurtsev	c79c9ea733	x	2025-06-20 17:06:13 -04:00
Eugene Yurtsev	b115e1dcde	tools	2025-06-20 17:04:41 -04:00
Eugene Yurtsev	3b59213311	fix tools	2025-06-20 16:45:24 -04:00
Eugene Yurtsev	0c0e5a299d	Replace notebook with markdown file	2025-06-20 16:33:42 -04:00
Lauren Hirata Singh	69d4c37d25	Consolidate assistant conceptual guides	2025-06-18 19:05:34 -04:00
Lauren Hirata Singh	0e7554a1a1	Update navigation	2025-06-18 10:58:18 -04:00
Lauren Hirata Singh	596c60a65c	Move LGP to platform section	2025-06-18 10:58:12 -04:00
Lauren Hirata Singh	e45797ce19	Fix titles based on feedback	2025-06-18 10:49:55 -04:00
Lauren Hirata Singh	e746b54a57	Change titles	2025-06-17 16:30:35 -04:00
Lauren Hirata SinghandGitHub	c88e22ffa7	Merge branch 'main' into get-started	2025-06-17 16:16:36 -04:00
Lauren Hirata Singh	dfdeb6a6f1	Edit stream modes	2025-06-17 15:41:57 -04:00
lc-arjun GitHub Marco Perini	1309243b29	docs: studio evals (#5129 ) * docs: studio evals * docs: added studio evals images (#5076) * docs: added studio evals images * Update docs/docs/cloud/how-tos/studio/run_evals.md Co-authored-by: lc-arjun <arjun@langchain.dev> * Update docs/docs/cloud/how-tos/studio/run_evals.md Co-authored-by: lc-arjun <arjun@langchain.dev> * Update docs/docs/cloud/how-tos/studio/run_evals.md Co-authored-by: lc-arjun <arjun@langchain.dev> * docs: updated studio evals * Update docs/docs/cloud/how-tos/studio/run_evals.md Co-authored-by: lc-arjun <arjun@langchain.dev> * docs: removed images --------- Co-authored-by: lc-arjun <arjun@langchain.dev> * final changes * i think its this --------- Co-authored-by: Marco Perini <perinim.98@gmail.com>	2025-06-17 12:32:04 -07:00
Lauren Hirata Singh	ba08acb71c	Fix broken links	2025-06-17 15:21:50 -04:00
Lauren Hirata Singh	dd64636ca8	Remove agents/streaming	2025-06-17 15:14:48 -04:00
Lauren Hirata Singh	664475887d	Consolidate streaming	2025-06-17 15:07:07 -04:00
Lauren Hirata Singh	3d3a2bfacd	edits	2025-06-16 21:07:12 -04:00
Nuno Campos	771c6150a4	langgraph 0.5.0rc1	2025-06-16 17:52:13 -07:00
Nuno Campos	edfb65fd3a	langgraph-prebuilt 0.5.0rc0	2025-06-16 17:47:21 -07:00
Lauren Hirata SinghandGitHub	0f92470e49	docs: Remove cookie consent (#5123 )	2025-06-16 18:41:55 -04:00
Nuno Campos	dfcaf97c73	langgraph 0.5.0rc0	2025-06-16 15:17:56 -07:00
Nuno Campos	63a0028372	langgraph-checkpoint 2.1.0	2025-06-16 14:58:50 -07:00
Nuno CamposandGitHub	1134017d07	Preparation for 0.5 release: langgraph-checkpoint (#5124 ) Prepare langgraph-checkpoint for 0.5 - Given we have no upper bound on langgraph-checkpoint dep need to undo all changes in langgraph-checkpoint that might break previous versions of langgraph	2025-06-16 21:57:11 +00:00
Lauren Hirata Singh	bbe90e04ca	Remove cookie consent popup	2025-06-16 16:35:27 -04:00
Lauren Hirata Singh	33feba4877	Remove cookie consent	2025-06-16 16:33:32 -04:00
Lauren Hirata Singh	905fcb3d02	Fix broken links	2025-06-16 16:25:31 -04:00
Lauren Hirata Singh	543d7d85af	Organize existing content differently	2025-06-16 16:17:52 -04:00
Nuno CamposandGitHub	4fec8e9dec	Preparation for 0.5 release (#5121 )	2025-06-16 13:14:25 -07:00
Nuno Campos	c137169325	Preparation for 0.5 release - Update deprecation warnings to mention 0.5, no 1.0 - Add back type hint support for Runnable arg to add_node	2025-06-16 13:07:52 -07:00
Nuno CamposandGitHub	1e2672e63d	Restore shallow checkpointer (#5105 )	2025-06-16 11:23:27 -07:00
Nuno CamposandGitHub	06803ab683	Add migration for pending_sends (#5106 )	2025-06-16 11:23:17 -07:00
hari-dhanushkodiandGitHub	3488ee47e0	chore: add docs for lgp deployment monitoring (#5104 )	2025-06-16 10:21:42 -07:00
Nuno CamposandGitHub	289bdd0cea	Introduce "tasks" and "checkpoints" stream modes (#5117 )	2025-06-16 10:14:18 -07:00
Nuno Campos	417103066b	Lint	2025-06-16 09:29:03 -07:00
Nuno Campos	25a59447c1	Introduce "tasks" and "checkpoints" stream modes - These are split out of "debug" stream mode, which is now an alias for ["tasks", "checkpoints"]	2025-06-16 08:47:45 -07:00
Nuno Campos	21906d2b7b	Add migration for pending_sends - Checkpoints saved on older versions of langgraph will be compatible with langgraph 0.5 and 1.0	2025-06-13 17:42:14 -07:00
Nuno Campos	0cad7019cb	Restore shallow checkpointer - This should definitely be removed soon, but let's give people more time to update	2025-06-13 17:37:40 -07:00
Nuno CamposandGitHub	7e735672bf	Restore compatibility with custom checkpointer classes created in prior versions (#5103 )	2025-06-13 16:36:36 -07:00
Nuno Campos	5498893780	Restore compatibility with custom checkpointer classes created in prior versions - Ensure existing custom checkpointer classes are compatible with new langgraph-checkpoint release	2025-06-13 16:29:55 -07:00
Nuno CamposandGitHub	e80f47aa01	Revert removals of APIs that were slated for removal in 1.0 (#5101 )	2025-06-13 16:09:20 -07:00
Nuno Campos	a0b2f742a3	Revert "Remove UntrackedValue channel" This reverts commit `05f3904d09`.	2025-06-13 15:36:47 -07:00
William FHandGitHub	b7973d65db	fix: Update lockfile (#5102 )	2025-06-13 14:53:40 -07:00
Nuno Campos	3fa3a586b5	Revert "Remove `MessageGraph` (#4875 )" This reverts commit `a5e6223569`.	2025-06-13 14:21:05 -07:00
William FHandGitHub	053b606b46	cli: 0.3.3 (#5100 )	2025-06-13 13:15:26 -07:00