feat(cli): add internal docker tag support (#4333)

- Deletes all data associated with a thread_id
- Implemented in InMemory, Sqlite and Postgres checkpointers

Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>

README.md

@@ -40,6 +40,9 @@ agent.invoke(
 )
 ```
 
+> [!TIP]
+> Check out [this guide](https://langchain-ai.github.io/langgraph/tutorials/workflows/) that walks through implementing common patterns (workflows and agents) in LangGraph. 
+
 ## Why use LangGraph?
 
 LangGraph is built for developers who want to build powerful, adaptable AI agents. Developers choose LangGraph for:

docs/Makefile

@@ -19,7 +19,7 @@ build-prebuilt:
 		poetry run python -m _scripts.third_party_page.get_download_stats --fake stats.yml; \
 		set +x; \
 	fi
-	poetry run python -m _scripts.third_party_page.create_third_party_page stats.yml docs/prebuilt.md --language python
+	poetry run python -m _scripts.third_party_page.create_third_party_page stats.yml docs/agents/prebuilt.md --language python
 
 build-docs: build-typedoc build-prebuilt
 	poetry run python -m mkdocs build --clean -f mkdocs.yml --strict

docs/_scripts/generate_api_reference_links.py

@@ -45,6 +45,8 @@ MANUAL_API_REFERENCES_LANGGRAPH = [
     (["langgraph.constants"], "langgraph.types", "Interrupt", "types"),
     (["langgraph.constants"], "langgraph.types", "interrupt", "types"),
     (["langgraph.constants"], "langgraph.types", "Command", "types"),
+    (["langgraph.config"], "langgraph.config", "get_stream_writer", "config"),
+    (["langgraph.config"], "langgraph.config", "get_store", "config"),
     (["langgraph.func"], "langgraph.func", "entrypoint", "func"),
     (["langgraph.func"], "langgraph.func", "task", "func"),
     (["langgraph.types"], "langgraph.types", "RetryPolicy", "types"),
@@ -56,6 +58,7 @@ MANUAL_API_REFERENCES_LANGGRAPH = [
     ([], "langgraph.checkpoint.base", "SerializerProtocol", "checkpoints"),
     ([], "langgraph.checkpoint.serde.jsonplus", "JsonPlusSerializer", "checkpoints"),
     ([], "langgraph.checkpoint.memory", "MemorySaver", "checkpoints"),
+    ([], "langgraph.checkpoint.memory", "InMemorySaver", "checkpoints"),
     ([], "langgraph.checkpoint.sqlite.aio", "AsyncSqliteSaver", "checkpoints"),
     ([], "langgraph.checkpoint.sqlite", "SqliteSaver", "checkpoints"),
     ([], "langgraph.checkpoint.postgres.aio", "AsyncPostgresSaver", "checkpoints"),
@@ -214,7 +217,7 @@ def update_markdown_with_imports(markdown: str, path: str) -> str:
         path: The path of the file where the markdown content originated.
 
     Returns:
-        Updated markdown with API reference links appended to Python code blocks.
+        Updated markdown with API reference links prepended to Python code blocks.
 
     Example:
         Given a markdown with a Python code block:
@@ -237,7 +240,7 @@ def update_markdown_with_imports(markdown: str, path: str) -> str:
             match (re.Match): The regex match object containing the code block.
 
         Returns:
-            str: The modified code block with API reference links appended if applicable.
+            str: The modified code block with API reference links prepended if applicable.
         """
         indent = match.group("indent")
         code_block = match.group("code")
@@ -253,8 +256,8 @@ def update_markdown_with_imports(markdown: str, path: str) -> str:
         api_links = " | ".join(
             f'<a href="{imp["docs"]}">{imp["imported"]}</a>' for imp in imports
         )
-        # Return the code block with appended API reference links
-        return f"{original_code_block}\n\n{indent}API Reference: {api_links}"
+        # Return the code block with prepended API reference links
+        return f"{indent}API Reference: {api_links}\n\n{original_code_block}"
 
     # Apply the replace_code_block function to all matches in the markdown
     updated_markdown = code_block_pattern.sub(replace_code_block, markdown)

docs/_scripts/notebook_hooks.py

@@ -31,6 +31,8 @@ REDIRECT_MAP = {
     "cloud/concepts/api.md": "concepts/langgraph_server.md",
     "cloud/concepts/cloud.md": "concepts/langgraph_cloud.md",
     "cloud/faq/studio.md": "concepts/langgraph_studio.md#studio-faqs",
+    # misc
+    "prebuilt.md": "agents/prebuilt.md"
 }
 
 

docs/_scripts/third_party_page/create_third_party_page.py

@@ -9,10 +9,7 @@ import yaml
 
 MARKDOWN = """\
 [//]: # (This file is automatically generated using a script in docs/_scripts. Do not edit this file directly!)
-# 🚀 Prebuilt Agents
-
-LangGraph includes a prebuilt React agent. For more information on how to use it, 
-check out our [how-to guides](https://langchain-ai.github.io/langgraph/how-tos/#prebuilt-react-agent).
+# Community Agents
 
 If you’re looking for other prebuilt libraries, explore the community-built options 
 below. These libraries can extend LangGraph's functionality in various ways.

docs/docs/agents/agents.md

@@ -0,0 +1,205 @@
+# Agents
+
+## What is an agent?
+
+An *agent* consists of three components: a **large language model (LLM)**, a set of **tools** it can use, and a **prompt** that provides instructions.
+
+The LLM operates in a loop. In each iteration, it selects a tool to invoke, provides input, receives the result (an observation), and uses that observation to inform the next action. The loop continues until a stopping condition is met — typically when the agent has gathered enough information to respond to the user.
+
+<figure markdown="1">
+![image](./assets/agent.png){: style="max-height:400px"}
+<figcaption>Agent loop: the LLM selects tools and uses their outputs to fulfill a user request.</figcaption>
+</figure>
+
+## Basic configuration
+
+Use [`create_react_agent`](https://python.langchain.com/docs/api_reference/langgraph.prebuilt.chat_agent_executor/#create-react-agent) to instantiate an agent:
+
+```python
+from langgraph.prebuilt import create_react_agent
+
+def get_weather(city: str) -> str:  # (1)!
+    """Get weather for a given city."""
+    return f"It's always sunny in {city}!"
+
+agent = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",  # (2)!
+    tools=[get_weather],  # (3)!
+    prompt="You are a helpful assistant"  # (4)!
+)
+
+# Run the agent
+agent.invoke({"messages": "what is the weather in sf"})
+```
+
+1. Define a tool for the agent to use. Tools can be defined as vanilla Python functions. For more advanced tool usage and customization, check the [tools](./tools.md) page.
+2. Provide a language model for the agent to use. To learn more about configuring language models for the agents, check the [models](./models.md) page.
+3. Provide a list of tools for the model to use.
+4. Provide a system prompt (instructions) to the language model used by the agent.
+
+## LLM configuration
+
+Use [init_chat_model](https://python.langchain.com/api_reference/langchain/chat_models/langchain.chat_models.base.init_chat_model.html) to configure an LLM with specific parameters,
+such as temperature:
+
+```python
+from langchain.chat_models import init_chat_model
+from langgraph.prebuilt import create_react_agent
+
+# highlight-next-line
+model = init_chat_model(
+    "anthropic:claude-3-7-sonnet-latest",
+    # highlight-next-line
+    temperature=0
+)
+
+agent = create_react_agent(
+    # highlight-next-line
+    model=model,
+    tools=[get_weather],
+)
+```
+
+See the [models](./models.md) page for more information on how to configure LLMs.
+
+## Custom Prompts
+
+Prompts instruct the LLM how to behave. They can be:
+
+* **Static**: A string is interpreted as a **system message**
+* **Dynamic**: a list of messages generated at **runtime** based on input or configuration
+
+### Static prompts
+
+Define a fixed prompt string or list of messages.
+
+```python
+from langgraph.prebuilt import create_react_agent
+
+agent = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",
+    tools=[get_weather],
+    # A static prompt that never changes
+    # highlight-next-line
+    prompt="Never answer questions about the weather."
+)
+
+agent.invoke(
+    {"messages": "what is the weather in sf"},
+)
+```
+
+### Dynamic prompts
+
+Define a function that returns a message list based on the agent's state and configuration:
+
+```python
+from langchain_core.messages import AnyMessage
+from langchain_core.runnables import RunnableConfig
+from langgraph.prebuilt.chat_agent_executor import AgentState
+from langgraph.prebuilt import create_react_agent
+
+# highlight-next-line
+def prompt(state: AgentState, config: RunnableConfig) -> list[AnyMessage]:  # (1)!
+    user_name = config.get("configurable", {}).get("user_name")
+    system_msg = f"You are a helpful assistant. Address the user as {user_name}."
+    return [{"role": "system", "content": system_msg}] + state["messages"]
+
+agent = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",
+    tools=[get_weather],
+    # highlight-next-line
+    prompt=prompt
+)
+
+agent.invoke(
+    {"messages": "what is the weather in sf"},
+    # highlight-next-line
+    config={"configurable": {"user_name": "John Smith"}}
+)
+```
+
+1. Dynamic prompts allow including non-message [context](./context.md) when constructing an input to the LLM, such as:
+
+    - Information passed at runtime, like a `user_id` or API credentials (using `config`).
+    - Internal agent state updated during a multi-step reasoning process (using `state`).
+
+    Dynamic prompts can be defined as functions that take `state` and `config` and return a list of messages to send to the LLM.
+
+See the [context](./context.md) page for more information.
+
+## Memory
+
+To allow multi-turn conversations with an agent, you need to enable [persistence](../concepts/persistence.md) by providing a `checkpointer` when creating an agent. At runtime you need to provide a config containing `thread_id` — a unique identifier for the conversation (session):
+
+```python
+from langgraph.prebuilt import create_react_agent
+from langgraph.checkpoint.memory import InMemorySaver
+
+# highlight-next-line
+checkpointer = InMemorySaver()
+
+agent = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",
+    tools=[get_weather],
+    # highlight-next-line
+    checkpointer=checkpointer  # (1)!
+)
+
+# Run the agent
+# highlight-next-line
+config = {"configurable": {"thread_id": "1"}}
+sf_response = agent.invoke(
+    {"messages": "what is the weather in sf"},
+    # highlight-next-line
+    config  # (2)!
+)
+ny_response = agent.invoke(
+    {"messages": "what about new york?"},
+    # highlight-next-line
+    config
+)
+```
+
+1. `checkpointer` allows the agent to store its state at every step in the tool calling loop. This enables [short-term memory](./memory.md#short-term-memory) and [human-in-the-loop](./human-in-the-loop.md) capabilities.
+2. Pass configuration with `thread_id` to be able to resume the same conversation on future agent invocations.
+
+When you enable the checkpointer, it stores agent state at every step in the provided checkpointer database (or in memory, if using `InMemorySaver`).
+
+Note that in the above example, when the agent is invoked the second time with the same `thread_id`, the original message history from the first conversation is automatically included, together with the new user input.
+
+Please see the [memory guide](./memory.md) for more details on how to work with memory.
+
+
+## Structured output
+
+To produce structured responses conforming to a schema, use the `response_format` parameter. The schema can be defined with a `Pydantic` model or `TypedDict`. The result will be accessible via the `structured_response` field.
+
+```python
+from pydantic import BaseModel
+from langgraph.prebuilt import create_react_agent
+
+class WeatherResponse(BaseModel):
+    conditions: str
+
+agent = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",
+    tools=[get_weather],
+    # highlight-next-line
+    response_format=WeatherResponse  # (1)!
+)
+
+response = agent.invoke({"messages": "what is the weather in sf"})
+
+# highlight-next-line
+response["structured_response"]
+```
+
+1. When `response_format` is provided, a separate step is added at the end of the agent loop: agent message history is passed to an LLM with structured output to generate a structured response.
+
+    To provide a system prompt to this LLM, use a tuple `(prompt, schema)`, e.g., `response_format=(prompt, WeatherResponse)`.
+
+!!! Note "LLM post-processing"
+
+    Structured output requires an additional call to the LLM to format the response according to the schema.
+

docs/docs/agents/context.md

@@ -0,0 +1,287 @@
+# Context
+
+Agents often require more than a list of messages to function effectively. They need **context**.
+
+Context includes *any* data outside the message list that can shape agent behavior or tool execution. This can be:
+
+- Information passed at runtime, like a `user_id` or API credentials.
+- Internal state updated during a multi-step reasoning process.
+- Persistent memory or facts from previous interactions.
+
+LangGraph provides **three** primary ways to supply context:
+
+| Type                                                                         | Description                                   | Mutable? | Lifetime                |
+|------------------------------------------------------------------------------|-----------------------------------------------|----------|-------------------------|
+| [**Config**](#config-static-context)                                         | data passed at the start of a run             | ❌        | per run                 |
+| [**State**](#state-mutable-context)                                          | dynamic data that can change during execution | ✅        | per run or conversation |
+| [**Long-term Memory (Store)**](#long-term-memory-cross-conversation-context) | data that can be shared between conversations | ✅        | across conversations    |
+
+You can use context to:
+
+- Adjust the system prompt the model sees
+- Feed tools with necessary inputs
+- Track facts during an ongoing conversation
+
+## Providing Runtime Context
+
+Use this when you need to inject data into an agent at runtime.
+
+### Config (static context)
+
+Config is for immutable data like user metadata or API keys. Use
+when you have values that don't change mid-run.
+
+Specify configuration using a key called **"configurable"** which is reserved
+for this purpose:
+
+```python
+agent.invoke(
+    {"messages": "hi!"},
+    # highlight-next-line
+    config={"configurable": {"user_id": "user_123"}}
+)
+```
+
+### State (mutable context)
+
+State acts as short-term memory during a run. It holds dynamic data that can evolve during execution, such as values derived from tools or LLM outputs.
+
+```python
+class CustomState(AgentState):
+    # highlight-next-line
+    user_name: str
+
+agent = create_react_agent(
+    # Other agent parameters...
+    # highlight-next-line
+    state_schema=CustomState,
+)
+
+agent.invoke({
+    "messages": "hi!",
+    "user_name": "Jane"
+})
+```
+
+!!! tip "Turning on memory"
+
+    Please see the [memory guide](./memory.md) for more details on how to enable memory. This is a powerful feature that allows you to persist the agent's state across multiple invocations.
+    Otherwise, the state is scoped only to a single agent run.
+
+
+
+### Long-Term Memory (cross-conversation context)
+
+For context that spans *across* conversations or sessions, LangGraph allows access to **long-term memory** via a `store`. This can be used to read or update persistent facts (e.g., user profiles, preferences, prior interactions). For more, see the [Memory guide](./memory.md).
+
+## Customizing Prompts with Context
+
+Prompts define how the agent behaves. To incorporate runtime context, you can dynamically generate prompts based on the agent's state or config.
+
+Common use cases:
+
+- Personalization
+- Role or goal customization
+- Conditional behavior (e.g., user is admin)
+
+=== "Using config"
+
+    ```python
+    from langchain_core.messages import AnyMessage
+    from langchain_core.runnables import RunnableConfig
+    from langgraph.prebuilt import create_react_agent
+    from langgraph.prebuilt.chat_agent_executor import AgentState
+
+    def prompt(
+        state: AgentState,
+        # highlight-next-line
+        config: RunnableConfig,
+    ) -> list[AnyMessage]:
+        # highlight-next-line
+        user_name = config.get("configurable", {}).get("user_name")
+        system_msg = f"You are a helpful assistant. User's name is {user_name}"
+        return [{"role": "system", "content": system_msg}] + state["messages"]
+
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_weather],
+        # highlight-next-line
+        prompt=prompt
+    )
+
+    agent.invoke(
+        ...,
+        # highlight-next-line
+        config={"configurable": {"user_name": "John Smith"}}
+    )
+    ```
+
+=== "Using state"
+
+    ```python
+    from langchain_core.messages import AnyMessage
+    from langchain_core.runnables import RunnableConfig
+    from langgraph.prebuilt import create_react_agent
+    from langgraph.prebuilt.chat_agent_executor import AgentState
+
+    class CustomState(AgentState):
+        # highlight-next-line
+        user_name: str
+
+    def prompt(
+        # highlight-next-line
+        state: CustomState
+    ) -> list[AnyMessage]:
+        # highlight-next-line
+        user_name = state["user_name"]
+        system_msg = f"You are a helpful assistant. User's name is {user_name}"
+        return [{"role": "system", "content": system_msg}] + state["messages"]
+
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[...],
+        # highlight-next-line
+        state_schema=CustomState,
+        # highlight-next-line
+        prompt=prompt
+    )
+
+    agent.invoke({
+        "messages": "hi!",
+        # highlight-next-line
+        "user_name": "John Smith"
+    })
+    ```
+
+## Tools
+
+Tools can access context through special parameter **annotations**.
+
+* Use `RunnableConfig` for config access
+* Use `Annotated[StateSchema, InjectedState]` for agent state
+
+
+!!! tip 
+
+    These annotations prevent LLMs from attempting to fill in the values. These parameters will be **hidden** from the LLM.
+
+=== "Using config"
+
+    ```python
+    def get_user_info(
+        # highlight-next-line
+        config: RunnableConfig,
+    ) -> str:
+        """Look up user info."""
+        # highlight-next-line
+        user_id = config.get("configurable", {}).get("user_id")
+        return "User is John Smith" if user_id == "user_123" else "Unknown user"
+
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_user_info],
+    )
+
+    agent.invoke(
+        {"messages": "look up user information"},
+        # highlight-next-line
+        config={"configurable": {"user_id": "user_123"}}
+    )
+    ```
+
+=== "Using State"
+
+    ```python
+    from typing import Annotated
+    from langgraph.prebuilt import InjectedState
+
+    class CustomState(AgentState):
+        # highlight-next-line
+        user_id: str
+
+    def get_user_info(
+        # highlight-next-line
+        state: Annotated[CustomState, InjectedState]
+    ) -> str:
+        """Look up user info."""
+        # highlight-next-line
+        user_id = state["user_id"]
+        return "User is John Smith" if user_id == "user_123" else "Unknown user"
+
+    agent = create_react_agent(
+        model="anthropic:claude-3-7-sonnet-latest",
+        tools=[get_user_info],
+        # highlight-next-line
+        state_schema=CustomState,
+    )
+
+    agent.invoke({
+        "messages": "look up user information",
+        # highlight-next-line
+        "user_id": "user_123"
+    })
+    ```
+
+
+## Update context from tools
+
+Tools can modify the agent's state during execution. This is useful for persisting intermediate results or making information accessible to subsequent tools or prompts.
+
+```python
+from typing import Annotated
+from langchain_core.tools import InjectedToolCallId
+from langchain_core.messages import ToolMessage
+from langgraph.prebuilt import InjectedState
+from langgraph.types import Command
+
+class CustomState(AgentState):
+    # highlight-next-line
+    user_name: str
+
+def get_user_info(
+    # highlight-next-line
+    tool_call_id: Annotated[str, InjectedToolCallId],
+    # highlight-next-line
+    config: RunnableConfig
+) -> Command:
+    """Look up user info."""
+    # highlight-next-line
+    user_id = config.get("configurable", {}).get("user_id")
+    name = "John Smith" if user_id == "user_123" else "Unknown user"
+    return Command(update={
+        # highlight-next-line
+        "user_name": name,
+        # update the message history
+        # highlight-next-line
+        "messages": [
+            ToolMessage(
+                "Successfully looked up user information",
+                # highlight-next-line
+                tool_call_id=tool_call_id
+            )
+        ]
+    })
+
+def greet(
+    # highlight-next-line
+    state: Annotated[CustomState, InjectedState]
+) -> str:
+    """Use this to greet the user once you found their info."""
+    user_name = state["user_name"]
+    return f"Hello {user_name}!"
+
+agent = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",
+    tools=[get_user_info, greet],
+    # highlight-next-line
+    state_schema=CustomState
+)
+
+agent.invoke(
+    {"messages": "greet the user"},
+    # highlight-next-line
+    config={"configurable": {"user_id": "user_123"}}
+)
+```
+
+For more details, see [how to update state from tools](../how-tos/update-state-from-tools.ipynb).

docs/docs/agents/deployment.md

@@ -0,0 +1,83 @@
+# Deployment
+
+To deploy your LangGraph agent, create and configure a LangGraph app. This setup supports both local development and production deployments.
+
+Features: 
+
+* 🖥️ Local server for development
+* 🧩 Studio Web UI for visual debugging
+* ☁️ Cloud and 🔧 self-hosted deployment options
+* 📊 LangSmith integration for tracing and observability
+
+!!! info "Requirements" 
+
+    - ✅ You **must** have a [LangSmith account](https://www.langchain.com/langsmith). You can sign up for **free** and get started with the free tier.
+
+## Create a LangGraph app
+
+```bash
+pip install -U "langgraph-cli[inmem]"
+langgraph new path/to/your/app --template new-langgraph-project-python
+```
+
+This will create an empty LangGraph project. You can modify it by replacing the code in `src/agent/graph.py` with your agent code. For example:
+
+```python
+from langgraph.prebuilt import create_react_agent
+
+def get_weather(city: str) -> str:
+    """Get weather for a given city."""
+    return f"It's always sunny in {city}!"
+
+graph = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",
+    tools=[get_weather],
+    prompt="You are a helpful assistant"
+)
+```
+
+### Install dependencies
+
+In the root of your new LangGraph app, install the dependencies in `edit` mode so your local changes are used by the server:
+
+```shell
+pip install -e .
+```
+
+### Create an `.env` file
+
+You will find a `.env.example` in the root of your new LangGraph app. Create
+a `.env` file in the root of your new LangGraph app and copy the contents of the `.env.example` file into it, filling in the necessary API keys:
+
+```bash
+LANGSMITH_API_KEY=lsv2...
+ANTHROPIC_API_KEY=sk-
+```
+
+## Launch LangGraph server locally
+
+```shell
+langgraph dev
+```
+
+This will start up the LangGraph API server locally. If this runs successfully, you should see something like:
+
+>    Ready!
+> 
+>    - API: [http://localhost:2024](http://localhost:2024/)
+>     
+>    - Docs: http://localhost:2024/docs
+>     
+>    - LangGraph Studio Web UI: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024
+
+See this [tutorial](https://langchain-ai.github.io/langgraph/tutorials/langgraph-platform/local-server/) to learn more about running LangGraph app locally.
+
+## LangGraph Studio Web UI
+
+LangGraph Studio Web is a specialized UI that you can connect to LangGraph API server to enable visualization, interaction, and debugging of your application locally. Test your graph in the LangGraph Studio Web UI by visiting the URL provided in the output of the `langgraph dev` command.
+
+>    - LangGraph Studio Web UI: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024
+
+## Deployment
+
+Once your LangGraph app is running locally, you can deploy it using LangGraph Cloud or self-hosted options. Refer to the [deployment options guide](../tutorials/deployment.md) for detailed instructions on all supported deployment models.

docs/docs/agents/evals.md

@@ -0,0 +1,119 @@
+# Evals
+
+To evaluate your agent's performance you can use `LangSmith` [evaluations](https://docs.smith.langchain.com/evaluation). You would need to first define an evaluator function to judge the results from an agent, such as final outputs or trajectory. Depending on your evaluation technique, this may or may not involve a reference output:
+
+```python
+def evaluator(*, outputs: dict, reference_outputs: dict):
+    # compare agent outputs against reference outputs
+    output_messages = outputs["messages"]
+    reference_messages = reference["messages"]
+    score = compare_messages(output_messages, reference_messages)
+    return {"key": "evaluator_score", "score": score}
+```
+
+To get started, you can use prebuilt evaluators from `AgentEvals` package:
+
+```bash
+pip install -U agentevals
+```
+
+## Create evaluator
+
+A common way to evaluate agent performance is by comparing its trajectory (the order in which it calls its tools) against a reference trajectory:
+
+```python
+import json
+# highlight-next-line
+from agentevals.trajectory.match import create_trajectory_match_evaluator
+
+outputs = [
+    {
+        "role": "assistant",
+        "tool_calls": [
+            {
+                "function": {
+                    "name": "get_weather",
+                    "arguments": json.dumps({"city": "san francisco"}),
+                }
+            },
+            {
+                "function": {
+                    "name": "get_directions",
+                    "arguments": json.dumps({"destination": "presidio"}),
+                }
+            }
+        ],
+    }
+]
+reference_outputs = [
+    {
+        "role": "assistant",
+        "tool_calls": [
+            {
+                "function": {
+                    "name": "get_weather",
+                    "arguments": json.dumps({"city": "san francisco"}),
+                }
+            },
+        ],
+    }
+]
+
+# Create the evaluator
+evaluator = create_trajectory_match_evaluator(
+    # highlight-next-line
+    trajectory_match_mode="superset",  # (1)!
+)
+
+# Run the evaluator
+result = evaluator(
+    outputs=outputs, reference_outputs=reference_outputs
+)
+```
+
+1. Specify how the trajectories will be compared. `superset` will accept output trajectory as valid if it's a superset of the reference one. Other options include: [strict](https://github.com/langchain-ai/agentevals?tab=readme-ov-file#strict-match), [unordered](https://github.com/langchain-ai/agentevals?tab=readme-ov-file#unordered-match) and [subset](https://github.com/langchain-ai/agentevals?tab=readme-ov-file#subset-and-superset-match)
+
+
+As a next step, learn more about how to [customize trajectory match evaluator](https://github.com/langchain-ai/agentevals?tab=readme-ov-file#agent-trajectory-match).
+
+### LLM-as-a-judge
+
+You can use LLM-as-a-judge evaluator that uses an LLM to compare the trajectory against the reference outputs and output a score:
+
+```python
+import json
+from agentevals.trajectory.llm import (
+    # highlight-next-line
+    create_trajectory_llm_as_judge,
+    TRAJECTORY_ACCURACY_PROMPT_WITH_REFERENCE
+)
+
+evaluator = create_trajectory_llm_as_judge(
+    prompt=TRAJECTORY_ACCURACY_PROMPT_WITH_REFERENCE,
+    model="openai:o3-mini"
+)
+```
+
+## Run evaluator
+
+To run an evaluator, you will first need to create a [LangSmith dataset](https://docs.smith.langchain.com/evaluation/concepts#datasets). To use the prebuilt AgentEvals evaluators, you will need a dataset with the following schema:
+
+- **input**: `{"messages": [...]}` input messages to call the agent with.
+- **output**: `{"messages": [...]}` expected message history in the agent output. For trajectory evaluation, you can choose to keep only assistant messages.
+
+```python
+from langsmith import Client
+from langgraph.prebuilt import create_react_agent
+from agentevals.trajectory.match import create_trajectory_match_evaluator
+
+client = Client()
+agent = create_react_agent(...)
+evaluator = create_trajectory_match_evaluator(...)
+
+experiment_results = client.evaluate(
+    lambda inputs: agent.invoke(inputs),
+    # replace with your dataset name
+    data="<Name of your dataset>",
+    evaluators=[evaluator]
+)
+```

docs/docs/agents/human-in-the-loop.md

@@ -0,0 +1,227 @@
+# Human-in-the-loop
+
+To review, edit and approve tool calls in an agent you can use LangGraph's built-in [human-in-the-loop](../concepts/human_in_the_loop.md) features, specifically the [`interrupt()`][langgraph.types.interrupt] primitive.
+
+LangGraph allows you to pause execution **indefinitely** — for minutes, hours, or even days—until human input is received.
+
+This is possible because the agent state is **checkpointed into a database**, which allows the system to persist execution context and later resume the workflow, continuing from where it left off.
+
+For a deeper dive into the **human-in-the-loop** concept, see the [concept guide](../concepts/human_in_the_loop.md).
+
+<figure markdown="1">
+![image](../concepts/img/human_in_the_loop/tool-call-review.png){: style="max-height:400px"}
+<figcaption>
+A human can review and edit the output from the agent before proceeding. This is particularly critical in applications where the tool calls requested may be sensitive or require human oversight.
+</figcaption>
+</figure>
+
+
+## Review tool calls
+
+To add a human approval step to a tool:
+
+1. Use `interrupt()` in the tool to pause execution.
+2. Resume with a `Command(resume=...)` to continue based on human input.
+
+```python
+from langgraph.checkpoint.memory import InMemorySaver
+from langgraph.types import interrupt
+from langgraph.prebuilt import create_react_agent
+
+# An example of a sensitive tool that requires human review / approval
+def book_hotel(hotel_name: str):
+    """Book a hotel"""
+    # highlight-next-line
+    response = interrupt(  # (1)!
+        f"Trying to call `book_hotel` with args {{'hotel_name': {hotel_name}}}. "
+        "Please approve or suggest edits."
+    )
+    if response["type"] == "accept":
+        pass
+    elif response["type"] == "edit":
+        hotel_name = response["args"]["hotel_name"]
+    else:
+        raise ValueError(f"Unknown response type: {response['type']}")
+    return f"Successfully booked a stay at {hotel_name}."
+
+# highlight-next-line
+checkpointer = InMemorySaver() # (2)!
+
+agent = create_react_agent(
+    model="anthropic:claude-3-5-sonnet-latest",
+    tools=[book_hotel],
+    # highlight-next-line
+    checkpointer=checkpointer, # (3)!
+)
+```
+
+1. The [`interrupt` function][langgraph.types.interrupt] pauses the agent graph at a specific node. In this case, we call `interrupt()` at the beginning of the tool function, which pauses the graph at the node that executes the tool. The information inside `interrupt()` (e.g., tool calls) can be presented to a human, and the graph can be resumed with the user input (tool call approval, edit or feedback).
+2. The `InMemorySaver` is used to store the agent state at every step in the tool calling loop. This enables [short-term memory](./memory.md#short-term-memory) and [human-in-the-loop](./human-in-the-loop.md) capabilities. In this example, we use `InMemorySaver` to store the agent state in memory. In a production application, the agent state will be stored in a database.
+3. Initialize the agent with the `checkpointer`.
+
+Run the agent with the `stream()` method, passing the `config` object to specify the thread ID. This allows the agent to resume the same conversation on future invocations.
+
+```python
+config = {
+   "configurable": {
+      # highlight-next-line
+      "thread_id": "1"
+   }
+}
+
+for chunk in agent.stream(
+    {"messages": "book a stay at McKittrick hotel"},
+    # highlight-next-line
+    config
+):
+    print(chunk)
+    print("\n")
+```
+
+> You should see that the agent runs until it reaches the `interrupt()` call, at which point it pauses and waits for human input.
+
+Resume the agent with a `Command(resume=...)` to continue based on human input.
+
+```python
+from langgraph.types import Command
+
+for chunk in agent.stream(
+    # highlight-next-line
+    Command(resume={"type": "accept"}),  # (1)!
+    # Command(resume={"type": "edit", "args": {"hotel_name": "McKittrick Hotel"}}),
+    config
+):
+    print(chunk)
+    print("\n")
+```
+
+1. The [`interrupt` function][langgraph.types.interrupt] is used in conjunction with the [`Command`](../reference/types.md#langgraph.types.Command) object to resume the graph with a value provided by the human.
+
+## Using with Agent Inbox
+
+You can create a wrapper to add interrupts to *any* tool. 
+
+The example below provides a reference implementation compatible with [Agent Inbox UI](https://github.com/langchain-ai/agent-inbox) and [Agent Chat UI](https://github.com/langchain-ai/agent-chat-ui).
+
+```python title="Wrapper that adds human-in-the-loop to any tool"
+from typing import Callable
+from langchain_core.tools import BaseTool, tool as create_tool
+from langchain_core.runnables import RunnableConfig
+from langgraph.types import interrupt 
+from langgraph.prebuilt.interrupt import HumanInterruptConfig, HumanInterrupt
+
+def add_human_in_the_loop(
+    tool: Callable | BaseTool,
+    *,
+    interrupt_config: HumanInterruptConfig = None,
+) -> BaseTool:
+   """Wrap a tool to support human-in-the-loop review.""" 
+    if not isinstance(tool, BaseTool):
+        tool = create_tool(tool)
+
+    if interrupt_config is None:
+        interrupt_config = {
+            "allow_accept": True,
+            "allow_edit": True,
+            "allow_respond": True,
+        }
+
+    @create_tool(  # (1)!
+        tool.name,
+        description=tool.description,
+        args_schema=tool.args_schema
+    )
+    def call_tool_with_interrupt(config: RunnableConfig, **tool_input):
+        request: HumanInterrupt = {
+            "action_request": {
+                "action": tool.name,
+                "args": tool_input
+            },
+            "config": interrupt_config,
+            "description": "Please review the tool call"
+        }
+        # highlight-next-line
+        response = interrupt([request])[0]  # (2)!
+        # approve the tool call
+        if response["type"] == "accept":
+            tool_response = tool.invoke(tool_input, config)
+        # update tool call args
+        elif response["type"] == "edit":
+            tool_input = response["args"]["args"]
+            tool_response = tool.invoke(tool_input, config)
+        # respond to the LLM with user feedback
+        elif response["type"] == "response":
+            user_feedback = response["args"]
+            tool_response = user_feedback
+        else:
+            raise ValueError(f"Unsupported interrupt response type: {response['type']}")
+
+        return tool_response
+
+    return call_tool_with_interrupt
+```
+
+1. This wrapper creates a new tool that calls `interrupt()` **before** executing the wrapped tool.
+2. `interrupt()` is using special input and output format that's expected by [Agent Inbox UI](https://github.com/langchain-ai/agent-inbox):
+    - a list of [`HumanInterrupt`][langgraph.prebuilt.interrupt.HumanInterrupt] objects is sent to `AgentInbox` render interrupt information to the end user
+    - resume value is provided by `AgentInbox` as a list (i.e., `Command(resume=[...])`)
+
+You can use the `add_human_in_the_loop` wrapper to add `interrupt()` to any tool without having to add it *inside* the tool:
+
+```python
+from langgraph.checkpoint.memory import InMemorySaver
+from langgraph.prebuilt import create_react_agent
+
+# highlight-next-line
+checkpointer = InMemorySaver()
+
+def book_hotel(hotel_name: str):
+   """Book a hotel"""
+   return f"Successfully booked a stay at {hotel_name}."
+
+
+agent = create_react_agent(
+    model="anthropic:claude-3-5-sonnet-latest",
+    tools=[
+        # highlight-next-line
+        add_human_in_the_loop(book_hotel), # (1)!
+    ],
+    # highlight-next-line
+    checkpointer=checkpointer,
+)
+
+config = {"configurable": {"thread_id": "1"}}
+
+# Run the agent
+for chunk in agent.stream(
+    {"messages": "book a stay at McKittrick hotel"},
+    # highlight-next-line
+    config
+):
+    print(chunk)
+    print("\n")
+```
+
+1. The `add_human_in_the_loop` wrapper is used to add `interrupt()` to the tool. This allows the agent to pause execution and wait for human input before proceeding with the tool call.
+
+> You should see that the agent runs until it reaches the `interrupt()` call, 
+>  at which point it pauses and waits for human input.
+
+Resume the agent with a `Command(resume=...)`  to continue based on human input.
+
+```python
+from langgraph.types import Command 
+
+for chunk in agent.stream(
+    # highlight-next-line
+    Command(resume=[{"type": "accept"}]),
+    # Command(resume=[{"type": "edit", "args": {"args": {"hotel_name": "McKittrick Hotel"}}}]),
+    config
+):
+    print(chunk)
+    print("\n")
+```
+
+## Additional resources
+
+* [Human-in-the-loop in LangGraph](../concepts/human_in_the_loop.md)

docs/docs/agents/mcp.md

@@ -0,0 +1,94 @@
+# MCP Integration
+
+[Model Context Protocol (MCP)](https://modelcontextprotocol.io/introduction) is an open protocol that standardizes how applications provide tools and context to language models. LangGraph agents can use tools defined on MCP servers through the `langchain-mcp-adapters` library.
+
+![MCP](./assets/mcp.png)
+
+Install the `langchain-mcp-adapters` library to use MCP tools in LangGraph:
+
+```bash
+pip install langchain-mcp-adapters
+```
+
+## Use MCP tools
+
+The `langchain-mcp-adapters` package enables agents to use tools defined across one or more MCP servers.
+
+```python title="Agent using tools defined on MCP servers"
+# highlight-next-line
+from langchain_mcp_adapters.client import MultiServerMCPClient
+from langgraph.prebuilt import create_react_agent
+
+# highlight-next-line
+async with MultiServerMCPClient(
+    {
+        "math": {
+            "command": "python",
+            # Replace with absolute path to your math_server.py file
+            "args": ["/path/to/math_server.py"],
+            "transport": "stdio",
+        },
+        "weather": {
+            # Ensure your start your weather server on port 8000
+            "url": "http://localhost:8000/sse",
+            "transport": "sse",
+        }
+    }
+) as client:
+    agent = create_react_agent(
+        "anthropic:claude-3-7-sonnet-latest",
+        # highlight-next-line
+        client.get_tools()
+    )
+    math_response = await agent.ainvoke({"messages": "what's (3 + 5) x 12?"})
+    weather_response = await agent.ainvoke({"messages": "what is the weather in nyc?"})
+```
+
+## Custom MCP servers
+
+To create your own MCP servers, you can use the `mcp` library. This library provides a simple way to define tools and run them as servers.
+
+Install the MCP library:
+
+```bash
+pip install mcp
+```
+Use the following reference implementations to test your agent with MCP tool servers.
+
+```python title="Example Math Server (stdio transport)"
+from mcp.server.fastmcp import FastMCP
+
+mcp = FastMCP("Math")
+
+@mcp.tool()
+def add(a: int, b: int) -> int:
+    """Add two numbers"""
+    return a + b
+
+@mcp.tool()
+def multiply(a: int, b: int) -> int:
+    """Multiply two numbers"""
+    return a * b
+
+if __name__ == "__main__":
+    mcp.run(transport="stdio")
+```
+
+```python title="Example Weather Server (SSE transport)"
+from mcp.server.fastmcp import FastMCP
+
+mcp = FastMCP("Weather")
+
+@mcp.tool()
+async def get_weather(location: str) -> str:
+    """Get weather for location."""
+    return "It's always sunny in New York"
+
+if __name__ == "__main__":
+    mcp.run(transport="sse")
+```
+
+## Additional resources
+
+- [MCP documentation](https://modelcontextprotocol.io/introduction)
+- [MCP Transport documentation](https://modelcontextprotocol.io/docs/concepts/transports)

docs/docs/agents/memory.md

@@ -0,0 +1,262 @@
+# Memory
+
+LangGraph supports two types of memory essential for building conversational agents:
+
+- **[Short-term memory](#short-term-memory)**: Tracks the ongoing conversation by maintaining message history within a session.
+- **[Long-term memory](#long-term-memory)**: Stores user-specific or application-level data across sessions.
+
+This guide demonstrates how to use both memory types with agents in LangGraph. For a deeper
+understanding of memory concepts, refer to the [LangGraph memory documentation](../concepts/memory.md).
+
+<figure markdown="1">
+![image](./assets/memory.png){: style="max-height:400px"}
+<figcaption>Both <strong>short-term</strong> and <strong>long-term</strong> memory require persistent storage to maintain continuity across LLM interactions. In production environments, this data is typically stored in a database.</figcaption>
+</figure>
+
+!!! note "Terminology"
+
+    In LangGraph:
+
+    - *Short-term memory* is also referred to as **thread-level memory**.
+    - *Long-term memory* is also called **cross-thread memory**.
+
+    A [thread](../concepts/persistence.md#threads) represents a sequence of related runs
+    grouped by the same `thread_id`.
+
+## Short-term memory
+
+Short-term memory enables agents to track multi-turn conversations. To use it, you must:
+
+1. Provide a `checkpointer` when creating the agent. The `checkpointer` enables [persistence](../concepts/persistence.md) of the agent's state.
+2. Supply a `thread_id` in the config when running the agent. The `thread_id` is a unique identifier for the conversation session.
+
+```python
+from langgraph.prebuilt import create_react_agent
+from langgraph.checkpoint.memory import InMemorySaver
+
+# highlight-next-line
+checkpointer = InMemorySaver() # (1)!
+
+
+def get_weather(city: str) -> str:
+    """Get weather for a given city."""
+    return f"It's always sunny in {city}!"
+
+
+agent = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",
+    tools=[get_weather],
+    # highlight-next-line
+    checkpointer=checkpointer # (2)!
+)
+
+# Run the agent
+config = {
+    "configurable": {
+        # highlight-next-line
+        "thread_id": "1"  # (3)!
+    }
+}
+
+sf_response = agent.invoke(
+    {"messages": "what is the weather in sf"},
+    # highlight-next-line
+    config
+)
+
+# Continue the conversation using the same thread_id
+ny_response = agent.invoke(
+    {"messages": "what about new york?"},
+    # highlight-next-line
+    config # (4)!
+)
+```
+
+1. The `InMemorySaver` is a checkpointer that stores the agent's state in memory. In a production setting, you would typically use a database or other persistent storage. Please review the [checkpointer documentation](../reference/checkpoints.md) for more options. If you're deploying with **LangGraph Platform**, the platform will provide a production-ready checkpointer for you.
+2. The `checkpointer` is passed to the agent. This enables the agent to persist its state across invocations. Please note that 
+3. A unique `thread_id` is provided in the config. This ID is used to identify the conversation session. The value is controlled by the user and can be any string.
+4. The agent will continue the conversation using the same `thread_id`. This will allow the agent to infer that the user is asking specifically about the **weather** in New York. 
+
+When the agent is invoked the second time with the same `thread_id`, the original message history from the first conversation is automatically included, allowing the agent to infer that the user is asking specifically about the **weather** in New York.
+
+!!! Note "LangGraph Platform providers a production-ready checkpointer"
+
+    If you're using [LangGraph Platform](./deployment.md), during deployment your checkpointer will be automatically configured to use a production-ready database.
+
+### Message history summarization 
+
+<figure markdown="1">
+![image](./assets/summary.png){: style="max-height:400px"}
+<figcaption>Message history can grow quickly and exceed the LLM's context window. A common solution is to maintain a running summary of the conversation. This allows the agent to keep track of the conversation without exceeding the LLM's context window.
+</figcaption>
+</figure>
+
+Long conversations can exceed the LLM's context window. To handle this, you can summarize older messages by specifying a [`pre_model_hook`][langgraph.prebuilt.chat_agent_executor.create_react_agent], such as the prebuilt [`SummarizationNode`](https://langchain-ai.github.io/langmem/reference/short_term/#langmem.short_term.SummarizationNode):
+
+```python
+from langchain_anthropic import ChatAnthropic
+from langmem.short_term import SummarizationNode
+from langchain_core.messages.utils import count_tokens_approximately
+from langgraph.prebuilt import create_react_agent
+from langgraph.prebuilt.chat_agent_executor import AgentState
+from langgraph.checkpoint.memory import InMemorySaver
+from typing import Any
+
+model = ChatAnthropic(model="claude-3-7-sonnet-latest")
+
+summarization_node = SummarizationNode( # (1)!
+    token_counter=count_tokens_approximately,
+    model=model,
+    max_tokens=384,
+    max_summary_tokens=128,
+    output_messages_key="llm_input_messages",
+)
+
+class State(AgentState):
+    # NOTE: we're adding this key to keep track of previous summary information
+    # to make sure we're not summarizing on every LLM call
+    # highlight-next-line
+    context: dict[str, Any]  # (2)!
+
+
+checkpointer = InMemorySaver() # (3)!
+
+agent = create_react_agent(
+    model=model,
+    tools=tools,
+    # highlight-next-line
+    pre_model_hook=summarization_node, # (4)!
+    # highlight-next-line
+    state_schema=State, # (5)!
+    checkpointer=checkpointer,
+)
+```
+
+1. The `InMemorySaver` is a checkpointer that stores the agent's state in memory. In a production setting, you would typically use a database or other persistent storage. Please review the [checkpointer documentation](../reference/checkpoints.md) for more options. If you're deploying with **LangGraph Platform**, the platform will provide a production-ready checkpointer for you.
+2. The `context` key is added to the agent's state. The key contains book-keeping information for the summarization node. It is used to keep track of the last summary information and ensure that the agent doesn't summarize on every LLM call, which can be inefficient.
+3. The `checkpointer` is passed to the agent. This enables the agent to persist its state across invocations.
+4. The `pre_model_hook` is set to the `SummarizationNode`. This node will summarize the message history before sending it to the LLM. The summarization node will automatically handle the summarization process and update the agent's state with the new summary. You can replace this with a custom implementation if you prefer. Please see the [create_react_agent][langgraph.prebuilt.chat_agent_executor.create_react_agent] API reference for more details.
+5. The `state_schema` is set to the `State` class, which is the custom state that contains an extra `context` key.
+
+To learn more about using `pre_model_hook` for managing message history, see this [how-to guide](../how-tos/create-react-agent-manage-message-history.ipynb)
+
+## Long-term memory
+
+Use long-term memory to store user-specific or application-specific data across conversations. This is useful for applications like chatbots, where you want to remember user preferences or other information.
+
+To use long-term memory, you need to:
+
+1. [Configure a store](../how-tos/cross-thread-persistence.ipynb) to persist data across invocations.
+2. Use the [`get_store`][langgraph.config.get_store] function to access the store from within tools or prompts.
+
+### Reading
+
+```python title="A tool the agent can use to look up user information"
+from langgraph.config import get_store
+from langgraph.prebuilt import create_react_agent
+from langgraph.store.memory import InMemoryStore
+
+# highlight-next-line
+store = InMemoryStore() # (1)!
+
+# highlight-next-line
+store.put(  # (2)!
+    ("users",),  # (3)!
+    "user_123",  # (4)!
+    {
+        "name": "John Smith",
+        "language": "English",
+    } # (5)!
+)
+
+def get_user_info(config: RunnableConfig) -> str:
+    """Look up user info."""
+    # Same as that provided to `create_react_agent`
+    # highlight-next-line
+    store = get_store() # (6)!
+    user_id = config.get("configurable", {}).get("user_id")
+    # highlight-next-line
+    user_info = store.get(("users",), user_id) # (7)!
+    return str(user_info.value) if user_info else "Unknown user"
+
+agent = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",
+    tools=[get_user_info],
+    # highlight-next-line
+    store=store # (8)!
+)
+
+# Run the agent
+agent.invoke(
+    {"messages": "look up user information"},
+    # highlight-next-line
+    config={"configurable": {"user_id": "user_123"}}
+)
+```
+
+1. The `InMemoryStore` is a store that stores data in memory. In a production setting, you would typically use a database or other persistent storage. Please review the [store documentation](../reference/stores.md) for more options. If you're deploying with **LangGraph Platform**, the platform will provide a production-ready store for you.
+2. For this example, we write some sample data to the store using the `put` method. Please see the [BaseStore.put][langgraph.store.base.BaseStore.put] API reference for more details.
+3. The first argument is the namespace. This is used to group related data together. In this case, we are using the `users` namespace to group user data.
+4. A key within the namespace. This example uses a user ID for the key.
+5. The data that we want to store for the given user.
+6. The `get_store` function is used to access the store. You can call it from anywhere in your code, including tools and prompts. This function returns the store that was passed to the agent when it was created.
+7. The `get` method is used to retrieve data from the store. The first argument is the namespace, and the second argument is the key. This will return a `StoreValue` object, which contains the value and metadata about the value.
+8. The `store` is passed to the agent. This enables the agent to access the store when running tools. You can also use the `get_store` function to access the store from anywhere in your code.
+
+### Writing
+
+```python title="Example of a tool that updates user information"
+from typing import TypedDict
+
+from langgraph.config import get_store
+from langgraph.prebuilt import create_react_agent
+from langgraph.store.memory import InMemoryStore
+
+store = InMemoryStore() # (1)!
+
+class UserInfo(TypedDict): # (2)!
+    name: str
+
+def save_user_info(user_info: UserInfo, config: RunnableConfig) -> str: # (3)!
+    """Save user info."""
+    # Same as that provided to `create_react_agent`
+    # highlight-next-line
+    store = get_store() # (4)!
+    user_id = config.get("configurable", {}).get("user_id")
+    # highlight-next-line
+    store.put(("users",), user_id, user_info) # (5)!
+    return "Successfully saved user info."
+
+agent = create_react_agent(
+    model="anthropic:claude-3-7-sonnet-latest",
+    tools=[save_user_info],
+    # highlight-next-line
+    store=store
+)
+
+# Run the agent
+agent.invoke(
+    {"messages": "My name is John Smith"},
+    # highlight-next-line
+    config={"configurable": {"user_id": "user_123"}} # (6)!
+)
+
+# You can access the store directly to get the value
+store.get(("users",), "user_123").value
+```
+
+1. The `InMemoryStore` is a store that stores data in memory. In a production setting, you would typically use a database or other persistent storage. Please review the [store documentation](../reference/stores.md) for more options. If you're deploying with **LangGraph Platform**, the platform will provide a production-ready store for you.
+2. The `UserInfo` class is a `TypedDict` that defines the structure of the user information. The LLM will use this to format the response according to the schema.
+3. The `save_user_info` function is a tool that allows an agent to update user information. This could be useful for a chat application where the user wants to update their profile information.
+4. The `get_store` function is used to access the store. You can call it from anywhere in your code, including tools and prompts. This function returns the store that was passed to the agent when it was created.
+5. The `put` method is used to store data in the store. The first argument is the namespace, and the second argument is the key. This will store the user information in the store.
+6. The `user_id` is passed in the config. This is used to identify the user whose information is being updated.
+
+### Prebuilt memory tools
+
+**LangMem** is a LangChain-maintained library that offers tools for managing long-term memories in your agent. See the [LangMem documentation](https://langchain-ai.github.io/langmem/) for usage examples.
+
+
+## Additional resources
+
+* [Memory in LangGraph](../concepts/memory.md)

docs/docs/agents/models.md

@@ -0,0 +1,69 @@
+# Models
+
+This page describes how to configure the chat model used by an agent.
+
+## Tool calling support
+
+To enable tool-calling agents, the underlying LLM must support [tool calling](https://python.langchain.com/docs/concepts/tool_calling/).
+
+Compatible models can be found in the [LangChain integrations directory](https://python.langchain.com/docs/integrations/chat/).
+
+## Specifying a model by name
+
+You can configure an agent with a model name string:
+
+```python
+from langgraph.prebuilt import create_react_agent
+
+agent = create_react_agent(
+    # highlight-next-line
+    model="anthropic:claude-3-7-sonnet-latest",
+    # other parameters
+)
+```
+
+## Using `init_chat_model`
+
+The [`init_chat_model`](https://python.langchain.com/docs/how_to/chat_models_universal_init/) utility simplifies model initialization with configurable parameters:
+
+```python
+from langchain.chat_models import init_chat_model
+
+model = init_chat_model(
+    "anthropic:claude-3-7-sonnet-latest",
+    temperature=0,
+    max_tokens=2048
+)
+```
+
+Refer to the [API reference](https://python.langchain.com/api_reference/langchain/chat_models/langchain.chat_models.base.init_chat_model.html) for advanced options.
+
+## Using provider-specific LLMs 
+
+If a model provider is not available via `init_chat_model`, you can instantiate the provider's model class directly. The model must implement the [BaseChatModel interface](https://python.langchain.com/api_reference/core/language_models/langchain_core.language_models.chat_models.BaseChatModel.html) and support tool calling:
+
+```python
+from langchain_anthropic import ChatAnthropic
+from langgraph.prebuilt import create_react_agent
+
+model = ChatAnthropic(
+    model="claude-3-7-sonnet-latest",
+    temperature=0,
+    max_tokens=2048
+)
+
+agent = create_react_agent(
+    # highlight-next-line
+    model=model,
+    # other parameters
+)
+```
+
+!!! note "Illustrative example" 
+
+    The example above uses `ChatAnthropic`, which is already supported by `init_chat_model`. This pattern is shown to illustrate how to manually instantiate a model not available through init_chat_model.
+
+## Additional resources
+
+- [Model integration directory](https://python.langchain.com/docs/integrations/chat/)
+- [Universal initialization with `init_chat_model`](https://python.langchain.com/docs/how_to/chat_models_universal_init/)

docs/docs/agents/multi-agent.md

@@ -0,0 +1,275 @@
+# Multi-agent
+
+A single agent might struggle if it needs to specialize in multiple domains or manage many tools. To tackle this, you can break your agent into smaller, independent agents and composing them into a [multi-agent system](../concepts/multi_agent.md).
+
+In multi-agent systems, agents need to communicate between each other. They do so via [handoffs](#handoffs) — a primitive that describes which agent to hand control to and the payload to send to that agent.
+
+Two of the most popular multi-agent architectures are:
+
+- [supervisor](#supervisor) — individual agents are coordinated by a central supervisor agent. The supervisor controls all communication flow and task delegation, making decisions about which agent to invoke based on the current context and task requirements.
+- [swarm](#swarm) — agents dynamically hand off control to one another based on their specializations. The system remembers which agent was last active, ensuring that on subsequent interactions, the conversation resumes with that agent.
+
+## Supervisor
+
+![Supervisor](./assets/supervisor.png)
+
+Use [`langgraph-supervisor`](https://github.com/langchain-ai/langgraph-supervisor-py) library to create a supervisor multi-agent system:
+
+```bash
+pip install langgraph-supervisor
+```
+
+```python
+from langchain_openai import ChatOpenAI
+from langgraph.prebuilt import create_react_agent
+# highlight-next-line
+from langgraph_supervisor import create_supervisor
+
+def book_hotel(hotel_name: str):
+    """Book a hotel"""
+    return f"Successfully booked a stay at {hotel_name}."
+
+def book_flight(from_airport: str, to_airport: str):
+    """Book a flight"""
+    return f"Successfully booked a flight from {from_airport} to {to_airport}."
+
+flight_assistant = create_react_agent(
+    model="openai:gpt-4o",
+    tools=[book_flight],
+    prompt="You are a flight booking assistant",
+    # highlight-next-line
+    name="flight_assistant"
+)
+
+hotel_assistant = create_react_agent(
+    model="openai:gpt-4o",
+    tools=[book_hotel],
+    prompt="You are a hotel booking assistant",
+    # highlight-next-line
+    name="hotel_assistant"
+)
+
+# highlight-next-line
+supervisor = create_supervisor(
+    agents=[flight_assistant, hotel_assistant],
+    model=ChatOpenAI(model="gpt-4o"),
+    prompt="You manage a hotel booking assistant and a flight booking assistant. Assign work to them."
+).compile()
+
+for chunk in supervisor.stream({
+    "messages": "book a flight from BOS to JFK and a stay at McKittrick Hotel"
+}):
+    print(chunk)
+    print("\n")
+```
+
+## Swarm
+
+![Swarm](./assets/swarm.png)
+
+Use [`langgraph-swarm`](https://github.com/langchain-ai/langgraph-swarm-py) library to create a swarm multi-agent system:
+
+```bash
+pip install langgraph-swarm
+```
+
+```python
+from langgraph.prebuilt import create_react_agent
+# highlight-next-line
+from langgraph_swarm import create_swarm, create_handoff_tool
+
+transfer_to_hotel_assistant = create_handoff_tool(
+    agent_name="hotel_assistant",
+    description="Transfer user to the hotel-booking assistant.",
+)
+transfer_to_flight_assistant = create_handoff_tool(
+    agent_name="flight_assistant",
+    description="Transfer user to the flight-booking assistant.",
+)
+
+flight_assistant = create_react_agent(
+    model="anthropic:claude-3-5-sonnet-latest",
+    # highlight-next-line
+    tools=[book_flight, transfer_to_hotel_assistant],
+    prompt="You are a flight booking assistant",
+    # highlight-next-line
+    name="flight_assistant"
+)
+hotel_assistant = create_react_agent(
+    model="anthropic:claude-3-5-sonnet-latest",
+    # highlight-next-line
+    tools=[book_hotel, transfer_to_flight_assistant],
+    prompt="You are a hotel booking assistant",
+    # highlight-next-line
+    name="hotel_assistant"
+)
+
+# highlight-next-line
+swarm = create_swarm(
+    agents=[flight_assistant, hotel_assistant],
+    default_active_agent="flight_assistant"
+).compile()
+
+for chunk in supervisor.stream({
+    "messages": "book a flight from BOS to JFK and a stay at McKittrick Hotel"
+}):
+    print(chunk)
+    print("\n")
+```
+
+## Handoffs
+
+A common pattern in multi-agent interactions is **handoffs**, where one agent *hands off* control to another. Handoffs allow you to specify:
+
+- **destination**: target agent to navigate to
+- **payload**: information to pass to that agent
+
+This is used both by `langgraph-supervisor` (supervisor hands off to individual agents) and `langgraph-swarm` (an individual agent can hand off to other agents).
+
+To implement handoffs with `create_react_agent`, you need to:
+
+1. Create a special tool that can transfer control to a different agent
+
+    ```python
+    def transfer_to_bob():
+        """Transfer to bob."""
+        return Command(
+            # name of the agent (node) to go to
+            # highlight-next-line
+            goto="bob",
+            # data to send to the agent
+            # highlight-next-line
+            update={"messages": [...]},
+            # indicate to LangGraph that we need to navigate to
+            # agent node in a parent graph
+            # highlight-next-line
+            graph=Command.PARENT,
+        )
+    ```
+
+1. Create individual agents that have access to handoff tools:
+
+    ```python
+    flight_assistant = create_react_agent(
+        ..., tools=[book_flight, transfer_to_hotel_assistant]
+    )
+    hotel_assistant = create_react_agent(
+        ..., tools=[book_hotel, transfer_to_flight_assistant]
+    )
+    ```
+
+1. Define a parent graph that contains individual agents as nodes:
+
+    ```python
+    from langgraph.graph import StateGraph, MessagesState
+    multi_agent_graph = (
+        StateGraph(MessagesState)
+        .add_node(flight_assistant)
+        .add_node(hotel_assistant)
+        ...
+    )
+    ```
+
+Putting this together, here is how you can implement a simple multi-agent system with two agents — a flight booking assistant and a hotel booking assistant:
+
+```python
+from typing import Annotated
+from langchain_core.tools import tool, InjectedToolCallId
+from langgraph.prebuilt import create_react_agent, InjectedState
+from langgraph.graph import StateGraph, START, MessagesState
+from langgraph.types import Command
+
+def create_handoff_tool(*, agent_name: str, description: str | None = None):
+    name = f"transfer_to_{agent_name}"
+    description = description or f"Transfer to {agent_name}"
+
+    @tool(name, description=description)
+    def handoff_tool(
+        # highlight-next-line
+        state: Annotated[MessagesState, InjectedState], # (1)!
+        # highlight-next-line
+        tool_call_id: Annotated[str, InjectedToolCallId],
+    ) -> Command:
+        tool_message = {
+            "role": "tool",
+            "content": f"Successfully transferred to {agent_name}",
+            "name": name,
+            "tool_call_id": tool_call_id,
+        }
+        return Command(  # (2)!
+            # highlight-next-line
+            goto=agent_name,  # (3)!
+            # highlight-next-line
+            update={"messages": state["messages"] + [tool_message]},  # (4)!
+            # highlight-next-line
+            graph=Command.PARENT,  # (5)!
+        )
+    return handoff_tool
+
+# Handoffs
+transfer_to_hotel_assistant = create_handoff_tool(
+    agent_name="hotel_assistant",
+    description="Transfer user to the hotel-booking assistant.",
+)
+transfer_to_flight_assistant = create_handoff_tool(
+    agent_name="flight_assistant",
+    description="Transfer user to the flight-booking assistant.",
+)
+
+# Simple agent tools
+def book_hotel(hotel_name: str):
+    """Book a hotel"""
+    return f"Successfully booked a stay at {hotel_name}."
+
+def book_flight(from_airport: str, to_airport: str):
+    """Book a flight"""
+    return f"Successfully booked a flight from {from_airport} to {to_airport}."
+
+# Define agents
+flight_assistant = create_react_agent(
+    model="anthropic:claude-3-5-sonnet-latest",
+    # highlight-next-line
+    tools=[book_flight, transfer_to_hotel_assistant],
+    prompt="You are a flight booking assistant",
+    # highlight-next-line
+    name="flight_assistant"
+)
+hotel_assistant = create_react_agent(
+    model="anthropic:claude-3-5-sonnet-latest",
+    # highlight-next-line
+    tools=[book_hotel, transfer_to_flight_assistant],
+    prompt="You are a hotel booking assistant",
+    # highlight-next-line
+    name="hotel_assistant"
+)
+
+# Define multi-agent graph
+multi_agent_graph = (
+    StateGraph(MessagesState)
+    .add_node(flight_assistant)
+    .add_node(hotel_assistant)
+    .add_edge(START, "flight_assistant")
+    .compile()
+)
+
+# Run the multi-agent graph
+for chunk in multi_agent_graph.stream({
+    "messages": "book a flight from BOS to JFK and a stay at McKittrick Hotel"
+}):
+    print(chunk)
+    print("\n")
+```
+
+1. Access agent's state
+2. The `Command` primitive allows specifying a state update and a node transition as a single operation, making it useful for implementing handoffs.
+3. Name of the agent or node to hand off to.
+4. Take the agent's messages and **add** them to the parent's **state** as part of the handoff. The next agent will see the parent state.
+5. Indicate to LangGraph that we need to navigate to agent node in a **parent** multi-agent graph.
+
+!!! Note
+    This handoff implementation assumes that:
+
+    - each agent receives overall message history (across all agents) in the multi-agent system as its input
+    - each agent outputs its internal messages history to the overall message history of the multi-agent system
+
+    Check out LangGraph [supervisor](https://github.com/langchain-ai/langgraph-supervisor-py#customizing-handoff-tools) and [swarm](https://github.com/langchain-ai/langgraph-swarm-py#customizing-handoff-tools) documentation to learn how to customize handoffs.

docs/docs/agents/overview.md

@@ -0,0 +1,38 @@
+---
+title: Overview
+---
+
+# Agent development with LangGraph
+
+**LangGraph** provides both low-level primitives and high-level prebuilt components for building agent-based applications. This section focuses on the **prebuilt**, **reusable** components designed to help you construct agentic systems quickly and reliably—without the need to implement orchestration, memory, or human feedback handling from scratch.
+
+## Key features
+
+LangGraph includes several capabilities essential for building robust, production-ready agentic systems:
+
+- [**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.
+- [**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.
+
+## High-level building blocks
+
+LangGraph comes with a set of prebuilt components that implement common agent behaviors and workflows. These abstractions are built on top of the LangGraph framework, offering a faster path to production while remaining flexible for advanced customization.
+
+Using LangGraph for agent development allows you to focus on your application's logic and behavior, instead of building and maintaining the supporting infrastructure for state, memory, and human feedback.
+
+## Package ecosystem
+
+The high-level components are organized into several packages, each with a specific focus.
+
+| Package                                    | Description                                                                 | Installation                            |
+|--------------------------------------------|-----------------------------------------------------------------------------|-----------------------------------------|
+| `langgraph-prebuilt` (part of `langgraph`) | Prebuilt components to [**create agents**](./agents.md)                     | `pip install -U langgraph langchain`    |
+| `langgraph-supervisor`                     | Tools for building [**supervisor**](./multi-agent.md#supervisor) agents     | `pip install -U langgraph-supervisor`   |
+| `langgraph-swarm`                          | Tools for building a [**swarm**](./multi-agent.md#swarm) multi-agent system | `pip install -U langgraph-swarm`        |
+| `langchain-mcp-adapters`                   | Interfaces to [**MCP servers**](./mcp.md) for tool and resource integration | `pip install -U langchain-mcp-adapters` |
+| `langmem`                                  | Agent memory management: [**short-term and long-term**](./memory.md)        | `pip install -U langmem`                |
+| `agentevals`                               | Utilities to [**evaluate agent performance**](./evals.md)                   | `pip install -U agentevals`             |
+

docs/docs/agents/prebuilt.md

@@ -1,4 +1,4 @@
-# 🚀 Prebuilt Agents
+# Community Agents
 
 To share your project, simply open a Pull Request adding an entry for your package in our [packages.yml](https://github.com/langchain-ai/langgraph/blob/main/docs/_scripts/third_party_page/packages.yml) file.
 

docs/docs/agents/run_agents.md

@@ -0,0 +1,159 @@
+# Running agents
+
+
+Agents support both synchronous and asynchronous execution using either `.invoke()` / `await .invoke()` for full responses, or `.stream()` / `.astream()` for **incremental** [streaming](#streaming) output. This section explains how to provide input, interpret output, enable streaming, and control execution limits.
+
+
+## Basic usage
+
+Agents can be executed in two primary modes:
+
+- **Synchronous** using `.invoke()` or `.stream()`
+- **Asynchronous** using `await .invoke()` or `async for` with `.astream()`
+
+=== "Sync invocation"
+    ```python
+    from langgraph.prebuilt import create_react_agent
+
+    agent = create_react_agent(...)
+
+    # highlight-next-line
+    response = agent.invoke({"messages": "what is the weather in sf"})
+    ```
+
+=== "Async invocation"
+    ```python
+    from langgraph.prebuilt import create_react_agent
+
+    agent = create_react_agent(...)
+    # highlight-next-line
+    response = await agent.ainvoke({"messages": "what is the weather in sf"})
+    ```
+
+## Inputs and outputs
+
+Agents use a language model that expects a list of `messages` as an input. Therefore, agent inputs and outputs are stored as a list of `messages` under the `messages` key in the agent [state](../concepts/low_level.md#working-with-messages-in-graph-state).
+
+## Input format
+
+Agent input must be a dictionary with a `messages` key. Supported formats are:
+
+| Format             | Example                                                                                                                       |
+|--------------------|-------------------------------------------------------------------------------------------------------------------------------|
+| String             | `{"messages": "Hello"}`  — Interpreted as a [HumanMessage](https://python.langchain.com/docs/concepts/messages/#humanmessage) |
+| Message dictionary | `{"messages": {"role": "user", "content": "Hello"}}`                                                                          |
+| List of messages   | `{"messages": [{"role": "user", "content": "Hello"}]}`                                                                        |
+| With custom state  | `{"messages": [{"role": "user", "content": "Hello"}], "user_name": "Alice"}` — If using a custom `state_schema`               |
+
+Messages are automatically converted into LangChain's internal message format. You can read
+more about [LangChain messages](https://python.langchain.com/docs/concepts/messages/#langchain-messages) in the LangChain documentation.
+
+!!! tip "Using custom agent state"
+
+    You can provide additional fields defined in your agent’s state schema directly in the input dictionary. This allows dynamic behavior based on runtime data or prior tool outputs.  
+    See the [context guide](./context.md) for full details.
+
+!!! note
+
+    A string input for `messages` is converted to a [HumanMessage](https://python.langchain.com/docs/concepts/messages/#humanmessage). This behavior differs from the `prompt` parameter in `create_react_agent`, which is interpreted as a [SystemMessage](https://python.langchain.com/docs/concepts/messages/#systemmessage) when passed as a string.
+
+
+## Output format
+
+Agent output is a dictionary containing:
+
+- `messages`: A list of all messages exchanged during execution (user input, assistant replies, tool invocations).
+- Optionally, `structured_response` if [structured output](./agents.md#structured-output) is configured.
+- If using a custom `state_schema`, additional keys corresponding to your defined fields may also be present in the output. These can hold updated state values from tool execution or prompt logic.
+
+See the [context guide](./context.md) for more details on working with custom state schemas and accessing context.
+
+## Streaming output
+
+Agents support streaming responses for more responsive applications. This includes:
+
+- **Progress updates** after each step
+- **LLM tokens** as they're generated
+- **Custom tool messages** during execution
+
+Streaming is available in both sync and async modes:
+
+=== "Sync streaming"
+
+    ```python
+    for chunk in agent.stream(
+        {"messages": "what is the weather in sf"},
+        stream_mode="updates"
+    ):
+        print(chunk)
+    ```
+
+=== "Async streaming"
+
+    ```python
+    async for chunk in agent.astream(
+        {"messages": "what is the weather in sf"},
+        stream_mode="updates"
+    ):
+        print(chunk)
+    ```
+
+!!! tip
+
+    For full details, see the [streaming guide](./streaming.md).
+
+## Max iterations
+
+To control agent execution and avoid infinite loops, set a recursion limit. This defines the maximum number of steps the agent can take before raising a `GraphRecursionError`. You can configure `recursion_limit` at runtime or when defining agent via `.with_config()`:
+
+=== "Runtime"
+
+    ```python
+    from langgraph.errors import GraphRecursionError
+    from langgraph.prebuilt import create_react_agent
+
+    max_iterations = 3
+    # highlight-next-line
+    recursion_limit = 2 * max_iterations + 1
+    agent = create_react_agent(
+        model="anthropic:claude-3-5-haiku-latest",
+        tools=[get_weather]
+    )
+
+    try:
+        response = agent.invoke(
+            {"messages": "what's the weather in sf"},
+            # highlight-next-line
+            {"recursion_limit": recursion_limit},
+        )
+    except GraphRecursionError:
+        print("Agent stopped due to max iterations.")
+    ```
+
+=== "`.with_config()`"
+
+    ```python
+    from langgraph.errors import GraphRecursionError
+    from langgraph.prebuilt import create_react_agent
+
+    max_iterations = 3
+    # highlight-next-line
+    recursion_limit = 2 * max_iterations + 1
+    agent = create_react_agent(
+        model="anthropic:claude-3-5-haiku-latest",
+        tools=[get_weather]
+    )
+    # highlight-next-line
+    agent_with_recursion_limit = agent.with_config(recursion_limit=recursion_limit)
+
+    try:
+        response = agent_with_recursion_limit.invoke(
+            {"messages": "what's the weather in sf"},
+        )
+    except GraphRecursionError:
+        print("Agent stopped due to max iterations.")
+    ```
+
+## Additional Resources
+
+* [Async programming in LangChain](https://python.langchain.com/docs/concepts/async)

docs/docs/agents/streaming.md

@@ -0,0 +1,208 @@
+# 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": "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": "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": "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": "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": "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": "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": "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": "what is the weather in sf"},
+        # highlight-next-line
+        stream_mode=["updates", "messages", "custom"]
+    ):
+        print(chunk)
+        print("\n")
+    ```
+
+## Additional resources
+
+* [Streaming in LangGraph](https://langchain-ai.github.io/langgraph/how-tos/streaming)

docs/docs/agents/tools.md

@@ -0,0 +1,268 @@
+# 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": "what's 3 + 5 and 4 * 7? make both calculations in parallel"})
+```
+
+## 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": "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": "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": "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": "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": "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.
+
+## 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.
+
+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.
+

docs/docs/agents/ui.md

@@ -0,0 +1,31 @@
+# UI
+
+You can use a prebuilt chat UI for interacting with any LangGraph agent through the [Agent Chat UI](https://github.com/langchain-ai/agent-chat-ui). Using the [deployed version](https://agentchat.vercel.app) is the quickest way to get started, and allows you to interact with both local and deployed graphs.
+
+## Run agent in UI
+
+First, set up LangGraph API server [locally](./deployment.md#launch-langgraph-server-locally) or deploy your agent on [LangGraph Cloud](https://langchain-ai.github.io/langgraph/cloud/quick_start/).
+
+Then, navigate to [Agent Chat UI](https://agentchat.vercel.app), or clone the repository and [run the dev server locally](https://github.com/langchain-ai/agent-chat-ui?tab=readme-ov-file#setup):
+
+<video controls src="../assets/base-chat-ui.mp4" type="video/mp4"></video>
+
+!!! Tip
+
+    UI has out-of-box support for rendering tool calls, and tool result messages. To customize what messages are shown, see the [Hiding Messages in the Chat](https://github.com/langchain-ai/agent-chat-ui?tab=readme-ov-file#hiding-messages-in-the-chat) section in the Agent Chat UI documentation.
+
+## Add human-in-the-loop
+
+Agent Chat UI has full support for [human-in-the-loop](../concepts/human_in_the_loop.md) workflows. To try it out, replace the agent code in `src/agent/graph.py` (from the [deployment](./deployment.md) guide) with this [agent implementation](./human-in-the-loop.md#using-with-agent-inbox):
+
+<video controls src="../assets/interrupt-chat-ui.mp4" type="video/mp4"></video>
+
+!!! Important
+
+    Agent Chat UI works best if your LangGraph agent interrupts using the [`HumanInterrupt` schema][langgraph.prebuilt.interrupt.HumanInterrupt]. If you do not use that schema, the Agent Chat UI will be able to render the input passed to the `interrupt` function, but it will not have full support for resuming your graph.
+
+## Generative UI
+
+You can also use generative UI in the Agent Chat UI.
+
+Generative UI allows you to define [React](https://react.dev/) components, and push them to the UI from the LangGraph server. For more documentation on building generative UI LangGraph agents, read [these docs](https://langchain-ai.github.io/langgraph/cloud/how-tos/generative_ui_react/).

docs/docs/cloud/deployment/cloud.md

@@ -1,4 +1,4 @@
-# How to Deploy to Cloud SaaS
+# How to Deploy to Cloud SaaS (Beta)
 
 Before deploying, review the [conceptual guide for the Cloud SaaS](../../concepts/langgraph_cloud.md) deployment option.
 

docs/docs/cloud/deployment/self_hosted_control_plane.md

@@ -1,4 +1,4 @@
-# How to Deploy Self-Hosted Control Plane
+# How to Deploy Self-Hosted Control Plane (Beta)
 
 Before deploying, review the [conceptual guide for the Self-Hosted Control Plane](../../concepts/langgraph_self_hosted_control_plane.md) deployment option.
 

docs/docs/cloud/deployment/self_hosted_data_plane.md

@@ -1,4 +1,4 @@
-# How to Deploy Self-Hosted Data Plane
+# How to Deploy Self-Hosted Data Plane (Beta)
 
 Before deploying, review the [conceptual guide for the Self-Hosted Data Plane](../../concepts/langgraph_self_hosted_data_plane.md) deployment option.
 

docs/docs/cloud/how-tos/use_stream_react.md

@@ -158,7 +158,7 @@ export default function HomePage() {
 }
 ```
 
-Under the hood, the `useStream()` hook will use the `streamMode: "messages-key"` to receive a stream of messages (i.e. individual LLM tokens) from any LangChain chat model invocations inside your graph nodes. Learn more about messages streaming in the [How to stream messages from your graph](./stream_messages.md) guide.
+Under the hood, the `useStream()` hook will use the `streamMode: "messages-tuple"` to receive a stream of messages (i.e. individual LLM tokens) from any LangChain chat model invocations inside your graph nodes. Learn more about messages streaming in the [How to stream messages from your graph](./stream_messages.md) guide.
 
 ### Interrupts
 

docs/docs/cloud/reference/env_var.md

@@ -1,6 +1,6 @@
 # Environment Variables
 
-The LangGraph Cloud Server supports specific environment variables for configuring a deployment.
+The LangGraph Server supports specific environment variables for configuring a deployment.
 
 ## `BG_JOB_ISOLATED_LOOPS`
 
@@ -32,7 +32,7 @@ See <a href="https://docs.smith.langchain.com/how_to_guides/tracing/sample_trace
 
 ## `LANGGRAPH_AUTH_TYPE`
 
-Type of authentication for the LangGraph Cloud Server deployment. Valid values: `langsmith`, `noop`.
+Type of authentication for the LangGraph Server deployment. Valid values: `langsmith`, `noop`.
 
 For deployments to LangGraph Cloud, this environment variable is set automatically. For local development or deployments where authentication is handled externally (e.g. self-hosted), set this environment variable to `noop`.
 
@@ -44,19 +44,27 @@ Set this environment variable to have a BYOC deployment send traces to a self-ho
 
 `SELF_HOSTED_LANGSMITH_HOSTNAME` is the hostname of the self-hosted LangSmith instance. It must be accessible to the BYOC deployment. `LANGSMITH_API_KEY` is a LangSmith API generated from the self-hosted LangSmith instance.
 
+## `LANGSMITH_TRACING`
+
+!!! info "Only for Self-Hosted Data Plane, Self-Hosted Control Plane, and Standalone Container"
+    Disabling LangSmith tracing is only available for [Self-Hosted Data Plane](../../concepts/langgraph_self_hosted_data_plane.md), [Self-Hosted Control Plane](../../concepts/langgraph_self_hosted_control_plane.md), and [Standalone Container](../../concepts/langgraph_standalone_container.md) deployments.
+
+Set `LANGSMITH_TRACING` to `false` to disable tracing to LangSmith.
+
 ## `LOG_LEVEL`
 
 Configure [log level](https://docs.python.org/3/library/logging.html#logging-levels). Defaults to `INFO`.
 
 ## `N_JOBS_PER_WORKER`
 
-Number of jobs per worker for the LangGraph Cloud task queue. Defaults to `10`.
+Number of jobs per worker for the LangGraph Server task queue. Defaults to `10`.
 
 ## `POSTGRES_URI_CUSTOM`
 
-For [Bring Your Own Cloud (BYOC)](../../concepts/bring_your_own_cloud.md) deployments only.
+!!! info "Only for Self-Hosted Data Plane and Self-Hosted Control Plane"
+    Custom Postgres instances are only available for [Self-Hosted Data Plane](../../concepts/langgraph_self_hosted_data_plane.md) and [Self-Hosted Control Plane](../../concepts/langgraph_self_hosted_control_plane.md) deployments.
 
-Specify `POSTGRES_URI_CUSTOM` to use an externally managed Postgres instance. The value of `POSTGRES_URI_CUSTOM` must be a valid [Postgres connection URI](https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNSTRING-URIS).
+Specify `POSTGRES_URI_CUSTOM` to use a custom Postgres instance. The value of `POSTGRES_URI_CUSTOM` must be a valid [Postgres connection URI](https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNSTRING-URIS).
 
 Postgres:
 
@@ -73,11 +81,11 @@ Control Plane Functionality:
 
 Database Connectivity:
 
-- The externally managed Postgres instance must be accessible by the LangGraph Server service in the ECS cluster. The BYOC user is responsible for ensuring connectivity.
-- For example, if an AWS RDS Postgres instance is provisioned, it can be provisioned in the same VPC (`langgraph-cloud-vpc`) as the ECS cluster with the `langgraph-cloud-service-sg` security group to ensure connectivity.
+- The custom Postgres instance must be accessible by the LangGraph Server. The user is responsible for ensuring connectivity.
 
 ## `REDIS_URI_CUSTOM`
 
-For [Bring Your Own Cloud (BYOC)](../../concepts/bring_your_own_cloud.md) deployments only.
+!!! info "Only for Self-Hosted Data Plane and Self-Hosted Control Plane"
+    Custom Redis instances are only available for [Self-Hosted Data Plane](../../concepts/langgraph_self_hosted_data_plane.md) and [Self-Hosted Control Plane](../../concepts/langgraph_self_hosted_control_plane.md) deployments.
 
-Specify `REDIS_URI_CUSTOM` to use an externally managed Redis instance. The value of `REDIS_URI_CUSTOM` must be a valid [Redis connection URI](https://redis-py.readthedocs.io/en/stable/connections.html#redis.Redis.from_url).
+Specify `REDIS_URI_CUSTOM` to use a custom Redis instance. The value of `REDIS_URI_CUSTOM` must be a valid [Redis connection URI](https://redis-py.readthedocs.io/en/stable/connections.html#redis.Redis.from_url).

docs/docs/concepts/deployment_options.md

@@ -10,11 +10,11 @@
 
 There are 4 main options for deploying with the LangGraph Platform:
 
-1. **[Cloud SaaS](#cloud-saas)**: Available for **Plus** and **Enterprise** plans.
+1. **<a href="#cloud-saas">Cloud SaaS<sup>(Beta)</sup></a>**: Available for **Plus** and **Enterprise** plans.
 
-1. **[Self-Hosted Data Plane](#self-hosted-data-plane)**: Available for the **Enterprise** plan.
+1. **<a href="#self-hosted-data-plane">Self-Hosted Data Plane<sup>(Beta)</sup></a>**: Available for the **Enterprise** plan.
 
-1. **[Self-Hosted Control Plane](#self-hosted-control-plane)**: Available for the **Enterprise** plan.
+1. **<a href="#self-hosted-control-plane">Self-Hosted Control Plane<sup>(Beta)</sup></a>**: Available for the **Enterprise** plan.
 
 1. **[Standalone Container](#standalone-container)**: Available for all plans.
 

docs/docs/concepts/index.md

@@ -76,7 +76,7 @@ The LangGraph Platform comprises several components that work together to suppor
 
 ### Deployment Options
 
-- [Cloud SaaS](../concepts/langgraph_cloud.md): Connect to your GitHub repositories and deploy LangGraph Servers to LangChain's cloud. We manage everything.
-- [Self-Hosted Data Plane](../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](../concepts/langgraph_self_hosted_control_plane.md#control-plane-ui): Create deployments from a self-hosted [Control Plane UI](../concepts/langgraph_control_plane.md) and deploy LangGraph Servers to your cloud. You manage everything.
+- <a href="./langgraph_cloud/">Cloud SaaS<sup>(Beta)</sup></a>: Connect to your GitHub repositories and deploy LangGraph Servers to LangChain's cloud. We manage everything.
+- <a href="./langgraph_self_hosted_data_plane/">Self-Hosted Data Plane<sup>(Beta)</sup></a>: 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.
+- <a href="./langgraph_self_hosted_control_plane/">Self-Hosted Control Plane<sup>(Beta)</sup></a>: 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.

docs/docs/concepts/langgraph_cli.md

@@ -54,7 +54,7 @@ pip install -U "langgraph-cli[inmem]"
 
 ### `up`
 
-The `langgraph up` command starts an instance of the [LangGraph API server](./langgraph_server.md) locally in a docker container. This requires thedocker server to be running locally. It also requires a LangSmith API key for local development or a license key for production use.
+The `langgraph up` command starts an instance of the [LangGraph API server](./langgraph_server.md) locally in a docker container. This requires the docker server to be running locally. It also requires a LangSmith API key for local development or a license key for production use.
 
 The server includes all API endpoints for your graph's runs, threads, assistants, etc. as well as the other services required to run your agent, including a managed database for checkpointing and storage.
 

docs/docs/concepts/langgraph_cloud.md

@@ -1,4 +1,4 @@
-# Cloud SaaS
+# Cloud SaaS (Beta)
 
 To deploy a [LangGraph Server](../concepts/langgraph_server.md), follow the how-to guide for [how to deploy to Cloud SaaS](../cloud/deployment/cloud.md).
 

docs/docs/concepts/langgraph_data_plane.md

@@ -44,21 +44,24 @@ Feature Differences:
 
 ### Autoscaling
 
-[`Production` type](../concepts/langgraph_control_plane.md#deployment-types) deployments automatically scale up to 10 containers. Scaling is based on the current request load for a single container. Specifically, the autoscaling implementation scales the deployment so that each container is processing about 10 concurrent requests. For example... 
+[`Production` type](../concepts/langgraph_control_plane.md#deployment-types) deployments automatically scale up to 10 containers. Scaling is based on 3 metrics:
 
-- If the deployment is processing 20 concurrent requests, the deployment will scale up from 1 container to 2 containers (20 requests / 2 containers = 10 requests per container).
-- If a deployment of 2 containers is processing 10 requests, the deployment will scale down from 2 containers to 1 container (10 requests / 1 container = 10 requests per container).
+1. CPU utilization
+1. Memory utilization
+1. Number of pending (in progress) [runs](../concepts/langgraph_server.md#runs)
 
-10 concurrent requests per container is the target threshold. However, 10 concurrent requests per container is not a hard limit. The number of concurrent requests can exceed 10 if there is a sudden burst of requests.
+For CPU utilization, the autoscaler targets 75% utilization. This means the autoscaler will scale the number of containers up or down to ensure that CPU utilization is at or near 75%. For memory utilization, the autoscaler targets 75% utilization as well.
 
-Scale down actions are delayed for 30 minutes before any action is taken. In other words, if the autoscaling implementation decides to scale down a deployment, it will first wait for 30 minutes before scaling down. After 30 minutes, the concurrency metric is recomputed and the deployment will scale down if the concurrency metric has met the target threshold. Otherwise, the deployment remains scaled up. This "cool down" period ensures that deployments do not scale up and down too frequently.
+For number of pending runs, the autoscaler targets 10 pending runs. For example, if the current number of containers is 1, but the number of pending runs in 20, the autoscaler will scale up the deployment to 2 containers (20 pending runs / 2 containers = 10 pending runs per container).
 
-In the future, the autoscaling implementation may evolve to accommodate other metrics such as background run queue size.
+Each metric is computed independently and the autoscaler will determine the scaling action based on the metric that results in the most number of containers.
+
+Scale down actions are delayed for 30 minutes before any action is taken. In other words, if the autoscaler decides to scale down a deployment, it will first wait for 30 minutes before scaling down. After 30 minutes, the metrics are recomputed and the deployment will scale down if the recomputed metrics result in a lower number of containers than the current number. Otherwise, the deployment remains scaled up. This "cool down" period ensures that deployments do not scale up and down too frequently.
 
 ### Static IP Addresses
 
 !!! info "Only for Cloud SaaS"
-    Static IP addresses are only available for [Cloud SaaS](../concepts/langgraph_cloud.md).
+    Static IP addresses are only available for [Cloud SaaS](../concepts/langgraph_cloud.md) deployments.
 
 All traffic from deployments created after January 6th 2025 will come through a NAT gateway. This NAT gateway will have several static IP addresses depending on the data region. Refer to the table below for the list of static IP addresses:
 
@@ -72,3 +75,46 @@ All traffic from deployments created after January 6th 2025 will come through a
 | 34.169.88.30   | 34.91.238.184  |
 | 34.19.93.202   | 35.204.101.241 |
 | 34.19.34.50    | 35.204.48.32   |
+
+### Custom Postgres
+
+!!! info "Only for Self-Hosted Data Plane and Self-Hosted Control Plane"
+    Custom Postgres instances are only available for [Self-Hosted Data Plane](../concepts/langgraph_self_hosted_data_plane.md) and [Self-Hosted Control Plane](../concepts/langgraph_self_hosted_control_plane.md) deployments.
+
+A custom Postgres instance can be used instead of the [one automatically created by the control plane](./langgraph_control_plane.md#database-provisioning). Specify the [`POSTGRES_URI_CUSTOM`](../cloud/reference/env_var.md#postgres_uri_custom) environment variable to use a custom Postgres instance.
+
+Multiple deployments can share the same Postgres instance. For example, for `Deployment A`, `POSTGRES_URI_CUSTOM` can be set to `postgres://<user>:<password>@/<database_name_1>?host=<hostname_1>` and for `Deployment B`, `POSTGRES_URI_CUSTOM` can be set to `postgres://<user>:<password>@/<database_name_2>?host=<hostname_1>`. `<database_name_1>` and `database_name_2` are different databases within the same instance, but `<hostname_1>` is shared. **The same database cannot be used for separate deployments**.
+
+### Custom Redis
+
+!!! info "Only for Self-Hosted Data Plane and Self-Hosted Control Plane"
+    Custom Redis instances are only available for [Self-Hosted Data Plane](../concepts/langgraph_self_hosted_data_plane.md) and [Self-Hosted Control Plane](../concepts/langgraph_self_hosted_control_plane.md) deployments.
+
+A custom Redis instance can be used instead of the one automatically created by the control plane. Specify the [REDIS_URI_CUSTOM](../cloud/reference/env_var.md#redis_uri_custom) environment variable to use a custom Redis instance.
+
+
+Multiple deployments can share the same Redis instance. For example, for `Deployment A`, `REDIS_URI_CUSTOM` can be set to `redis://<hostname_1>:<port>/1` and for `Deployment B`, `REDIS_URI_CUSTOM` can be set to `redis://<hostname_1>:<port>/2`. `1` and `2` are different database numbers within the same instance, but `<hostname_1>` is shared. **The same database number cannot be used for separate deployments**.
+
+### LangSmith Tracing
+
+LangGraph Server is automatically configured to send traces to LangSmith. See the table below for details with respect to each deployment option.
+
+| Cloud SaaS | Self-Hosted Data Plane | Self-Hosted Control Plane | Standalone Container |
+|------------|------------------------|---------------------------|----------------------|
+| Required<br><br>Trace to LangSmith SaaS. | Optional<br><br>Disable tracing or trace to LangSmith SaaS. | Optional<br><br>Disable tracing or trace to Self-Hosted LangSmith. | Optional<br><br>Disable tracing, trace to LangSmith SaaS, or trace to Self-Hosted LangSmith. |
+
+### Telemetry
+
+LangGraph Server is automatically configured to report telemetry metadata for billing purposes. See the table below for details with respect to each deployment option.
+
+| Cloud SaaS | Self-Hosted Data Plane | Self-Hosted Control Plane | Standalone Container |
+|------------|------------------------|---------------------------|----------------------|
+| Telemetry sent to LangSmith SaaS. | Telemetry sent to LangSmith SaaS. | Self-reported usage (audit) for air-gapped license key.<br><br>Telemetry sent to LangSmith SaaS for LangGraph Platform License Key. | Self-reported usage (audit) for air-gapped license key.<br><br>Telemetry sent to LangSmith SaaS for LangGraph Platform License Key. |
+
+### Licensing
+
+LangGraph Server is automatically configured to perform license key validation. See the table below for details with respect to each deployment option.
+
+| Cloud SaaS | Self-Hosted Data Plane | Self-Hosted Control Plane | Standalone Container |
+|------------|------------------------|---------------------------|----------------------|
+| LangSmith API Key validated against LangSmith SaaS. | LangSmith API Key validated against LangSmith SaaS. | Air-gapped license key or LangGraph Platform License Key validated against LangSmith SaaS. | Air-gapped license key or LangGraph Platform License Key validated against LangSmith SaaS. |

docs/docs/concepts/langgraph_platform.md

@@ -5,6 +5,10 @@ search:
 
 # LangGraph Platform
 
+Watch this 4-minute overview of LangGraph Platform to see how it helps you build, deploy, and evaluate agentic applications.
+
+<iframe width="560" height="315" src="https://www.youtube.com/embed/pfAQxBS5z88?si=XGS6Chydn6lhSO1S" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe>
+
 ## Overview
 
 LangGraph Platform is a commercial solution for deploying agentic applications to production, built on the open-source [LangGraph framework](./high_level.md).

docs/docs/concepts/langgraph_self_hosted_control_plane.md

@@ -1,4 +1,4 @@
-# Self-Hosted Control Plane
+# Self-Hosted Control Plane (Beta)
 
 To deploy a [LangGraph Server](../concepts/langgraph_server.md), follow the how-to guide for [how to deploy the Self-Hosted Control Plane](../cloud/deployment/self_hosted_control_plane.md).
 

docs/docs/concepts/langgraph_self_hosted_data_plane.md

@@ -1,4 +1,4 @@
-# Self-Hosted Data Plane
+# Self-Hosted Data Plane (Beta)
 
 To deploy a [LangGraph Server](../concepts/langgraph_server.md), follow the how-to guide for [how to deploy the Self-Hosted Data Plane](../cloud/deployment/self_hosted_data_plane.md).
 

docs/docs/concepts/platform_architecture.md

@@ -4,7 +4,7 @@
 
 ## How we use Postgres
 
-Postgres is the persistence layer for all user and run data in LGP. This stores both checkpoints (see more info [here](./persistence.md)) as well as the server resources (threads, runs, assistants and crons).
+Postgres is the persistence layer for all user, run, and long-term memory data in LGP. This stores both checkpoints (see more info [here](./persistence.md)), server resources (threads, runs, assistants and crons), as well as items saved in the long-term memory store (see more info [here](./persistence.md#memory-store)).
 
 ## How we use Redis
 

docs/docs/how-tos/http/custom_routes.md

@@ -8,10 +8,6 @@ Defining a custom app object lets you add any routes you'd like, so you can do a
 
 Below is an example using FastAPI.
 
-???+ note "Python only"
-
-    We currently only support custom authentication and authorization in Python deployments with `langgraph-api>=0.0.26`.
-
 ## Create app
 
 Starting from an **existing** LangGraph Platform application, add the following custom route code to your `webapp.py` file. If you are starting from scratch, you can create a new app from a template using the CLI.

docs/docs/reference/prebuilt.md

@@ -17,3 +17,12 @@
     options:
       members:
         - ValidationNode
+
+
+::: langgraph.prebuilt.interrupt
+    options:
+      members:
+        - HumanInterruptConfig
+        - ActionRequest
+        - HumanInterrupt
+        - HumanResponse

docs/docs/tutorials/deployment.md

@@ -17,9 +17,9 @@ Get started deploying your LangGraph applications locally or on the cloud with
 
 ## Deployment Options
 
-- [Cloud SaaS](../concepts/langgraph_cloud.md): Connect to your GitHub repositories and deploy LangGraph Servers to LangChain's cloud. We manage everything.
-- [Self-Hosted Data Plane](../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](../concepts/langgraph_self_hosted_control_plane.md#control-plane-ui): Create deployments from a self-hosted [Control Plane UI](../concepts/langgraph_control_plane.md) and deploy LangGraph Servers to your cloud. You manage everything.
+- <a href="../../concepts/langgraph_cloud/">Cloud SaaS<sup>(Beta)</sup></a>: Connect to your GitHub repositories and deploy LangGraph Servers to LangChain's cloud. We manage everything.
+- <a href="../../concepts/langgraph_self_hosted_data_plane/">Self-Hosted Data Plane<sup>(Beta)</sup></a>: 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.
+- <a href="../../concepts/langgraph_self_hosted_control_plane/">Self-Hosted Control Plane<sup>(Beta)</sup></a>: 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.
 
 A quick comparison...

docs/docs/tutorials/workflows/index.md

@@ -222,7 +222,7 @@ As noted in the Anthropic blog on `Building Effective Agents`:
 
 
     @entrypoint()
-    def parallel_workflow(topic: str):
+    def prompt_chaining_workflow(topic: str):
         original_joke = generate_joke(topic).result()
         if check_punchline(original_joke) == "Pass":
             return original_joke
@@ -231,7 +231,7 @@ As noted in the Anthropic blog on `Building Effective Agents`:
         return polish_joke(improved_joke).result()
 
     # Invoke
-    for step in parallel_workflow.stream("cats", stream_mode="updates"):
+    for step in prompt_chaining_workflow.stream("cats", stream_mode="updates"):
         print(step)
         print("\n")
     ```

docs/mkdocs.yml

@@ -57,12 +57,16 @@ plugins:
       separator: '[\s\u200b\-,:!=\[\]()"`/]+|\.(?!\d)|&[lg]t;'
   - autorefs
   - mkdocstrings:
+      custom_templates: templates
       handlers:
         python:
           import:
             - https://docs.python.org/3/objects.inv
             - https://python.langchain.com/api_reference/objects.inv
           options:
+            preload_modules:
+              - langchain
+              - langchain_core
             enable_inventory: true
             members_order: source
             allow_inspection: true
@@ -75,7 +79,10 @@ plugins:
               docstring_style: google
             docstring_section_style: list
             show_root_toc_entry: false
+            show_signature: true
             show_signature_annotations: true
+            separate_signature: true
+            line_length: 60
             show_symbol_type_heading: true
             show_symbol_type_toc: true
             signature_crossrefs: true
@@ -84,7 +91,7 @@ plugins:
                 - "!^_"
 
 nav:
-  - Home:
+  - LangGraph:
     - index.md
     - Get started:
       - Learn the basics: tutorials/introduction.ipynb
@@ -213,8 +220,8 @@ nav:
                 - how-tos/ttl/configure_ttl.md
             - Authentication & Access Control:
                 - Authentication & Access Control: how-tos#authentication-access-control
-                - cloud/how-tos/auth/custom_auth_new.md
-                - cloud/how-tos/auth/openapi_security_new.md
+                - how-tos/auth/custom_auth.md
+                - how-tos/auth/openapi_security.md
             - Assistants:
                 - Assistants: how-tos#assistants
                 - cloud/how-tos/configuration_cloud.md
@@ -256,6 +263,11 @@ nav:
                 - cloud/how-tos/webhooks.md
             - Cron Jobs:
                 - cloud/how-tos/cron_jobs.md
+            - Modifying the API:
+                - Modifying the API: how-tos#modifying-the-api
+                - how-tos/http/custom_lifespan.md
+                - how-tos/http/custom_middleware.md
+                - how-tos/http/custom_routes.md
             - LangGraph Studio:
                 - LangGraph Studio: how-tos#langgraph-studio
                 - cloud/how-tos/test_deployment.md
@@ -265,6 +277,7 @@ nav:
                 - cloud/how-tos/datasets_studio.md
                 - cloud/how-tos/iterate_graph_studio.md
                 - cloud/how-tos/clone_traces_studio.md
+                - how-tos/local-studio.md
       - Concepts:
         - concepts/index.md
         - LangGraph:
@@ -273,8 +286,9 @@ nav:
           - concepts/low_level.md
           - concepts/agentic_concepts.md
           - concepts/multi_agent.md
-          - concepts/breakpoints
+          - concepts/breakpoints.md
           - concepts/human_in_the_loop.md
+          - concepts/v0-human-in-the-loop.md
           - concepts/time-travel.md
           - concepts/persistence.md
           - concepts/memory.md
@@ -287,7 +301,10 @@ nav:
           - High Level:
             - High Level: concepts#high-level
             - concepts/langgraph_platform.md
+            - concepts/platform_architecture.md
+            - concepts/scalability_and_resilience.md
             - concepts/deployment_options.md
+            - concepts/bring_your_own_cloud.md
             - concepts/plans.md
             - concepts/template_applications.md
           - Components:
@@ -311,6 +328,7 @@ nav:
             - concepts/langgraph_self_hosted_data_plane.md
             - concepts/langgraph_self_hosted_control_plane.md
             - concepts/langgraph_standalone_container.md
+            - concepts/self_hosted.md
       - Tutorials:
         - tutorials/index.md
         - Quick Start:
@@ -370,8 +388,6 @@ nav:
             - tutorials/auth/resource_auth.md
             - tutorials/auth/add_auth_server.md
     - Resources:
-      # NOTE: prebuilt.md is auto-generated by `make build-prebuilt`
-      - Prebuilt Agents: prebuilt.md
       - Companies using LangGraph: adopters.md
       - LLMS-txt: llms-txt-overview.md
       - FAQ: concepts/faq.md
@@ -386,7 +402,28 @@ nav:
           - troubleshooting/errors/INVALID_LICENSE.md
       - LangGraph Academy Course: https://academy.langchain.com/courses/intro-to-langgraph
 
+  - Agents:
+    - agents/overview.md
+    - Get started:
+      - agents/agents.md
+    - Documentation:
+      - 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
+    - Resources:
+      # NOTE: prebuilt.md is auto-generated by `make build-prebuilt`
+      - agents/prebuilt.md
   - API reference:
+    - reference/index.md
     - Library:
       - Graphs: reference/graphs.md
       - Checkpointing: reference/checkpoints.md
@@ -497,13 +534,8 @@ extra:
             Thanks for your feedback! Please help us improve this page by adding to the discussion below.
 validation:
   # https://www.mkdocs.org/user-guide/configuration/
-  # We're `ignoring` nav.omitted_files because we are going to rely
-  # on files being properly links to from the index pages of:
-  # - tutorials
-  # - concepts
-  # - how-tos
-  # - reference
-  omitted_files: ignore
+  # We are still raising for omitted files because they determine the breadcrumbs for pages.
+  omitted_files: warn 
   absolute_links: warn
   unrecognized_links: warn
   # TODO: figure out how to enable 'warn' for this

docs/pyproject.toml

@@ -11,6 +11,7 @@ python = "^3.10"
 aiohappyeyeballs = "2.4.3"
 hub = "^3.0.1"
 xxhash = "^3.5.0"
+black = "^25.1.0"
 
 [tool.poetry.group.docs.dependencies]
 langgraph = { path = "../libs/langgraph/", develop = true }

docs/templates/python/material/function.html.jinja

@@ -0,0 +1,121 @@
+{#- Template for Python functions.
+   
+This template renders a Python function or method.
+
+Context:
+  function (griffe.Function): The function to render.
+  root (bool): Whether this is the root object, injected with `:::` in a Markdown page.
+  heading_level (int): The HTML heading level to use.
+  config (dict): The configuration options.
+-#}
+
+{% block logs scoped %}
+  {{ log.debug("Rendering " + function.path) }}
+{% endblock logs %}
+
+{% import "language"|get_template as lang with context %}
+{#- Language module providing the `t` translation method. -#}
+
+<div class="doc doc-object doc-function">
+  {% with obj = function, html_id = function.path %}
+    {% if root %}
+      {% set show_full_path = config.show_root_full_path %}
+      {% set root_members = True %}
+    {% elif root_members %}
+      {% set show_full_path = config.show_root_members_full_path or config.show_object_full_path %}
+      {% set root_members = False %}
+    {% else %}
+      {% set show_full_path = config.show_object_full_path %}
+    {% endif %}
+
+    {% set function_name = function.path if show_full_path else function.name %}
+    {#- Brief or full function name depending on configuration. -#}
+    {% set symbol_type = "method" if function.parent.is_class else "function" %}
+    {#- Symbol type: method when parent is a class, function otherwise. -#}
+
+    {% if not root or config.show_root_heading %}
+      {% filter heading(
+          heading_level,
+          role="function",
+          id=html_id,
+          class="doc doc-heading",
+          toc_label=(('<code class="doc-symbol doc-symbol-toc doc-symbol-' + symbol_type + '"></code>&nbsp;')|safe if config.show_symbol_type_toc else '') + function.name,
+        ) %}
+      
+        {% block heading scoped %}
+          {% if config.show_symbol_type_heading %}<code class="doc-symbol doc-symbol-heading doc-symbol-{{ symbol_type }}"></code>{% endif %}
+          {% if config.separate_signature %}
+            <span class="doc doc-object-name doc-function-name">{{ config.heading if config.heading and root else function_name }}</span>
+          {% else %}
+            {%+ filter highlight(language="python", inline=True) %}
+              {{ function_name }}{% include "signature"|get_template with context %}
+            {% endfilter %}
+          {% endif %}
+        {% endblock heading %}
+
+        {% block labels scoped %}
+          {% with labels = function.labels %}
+            {% include "labels"|get_template with context %}
+          {% endwith %}
+        {% endblock labels %}
+
+      {% endfilter %}
+
+      {% block signature scoped %}
+        {#- Signature block.
+        
+        This block renders only the main signature and deliberately omits the overloads.
+        -#}
+        {% if config.separate_signature %}
+          {% filter format_signature(function, config.line_length, crossrefs=config.signature_crossrefs) %}
+            {{ function.name }}
+          {% endfilter %}
+        {% endif %}
+      {% endblock signature %}
+
+    {% else %}
+
+      {% if config.show_root_toc_entry %}
+        {% filter heading(
+            heading_level,
+            role="function",
+            id=html_id,
+            toc_label=(('<code class="doc-symbol doc-symbol-toc doc-symbol-' + symbol_type + '"></code>&nbsp;')|safe if config.show_symbol_type_toc else '') + (config.toc_label if config.toc_label and root else function.name),
+            hidden=True,
+          ) %}
+        {% endfilter %}
+      {% endif %}
+      {% set heading_level = heading_level - 1 %}
+    {% endif %}
+
+    <div class="doc doc-contents {% if root %}first{% endif %}">
+      {% block contents scoped %}
+        {#- Contents block.
+        
+        This block renders the function’s docstring and source.
+        -#}
+        {% block docstring scoped %}
+          {% with docstring_sections = function.docstring.parsed %}
+            {% include "docstring"|get_template with context %}
+          {% endwith %}
+        {% endblock docstring %}
+
+        {% block source scoped %}
+          {% if config.show_source and function.source %}
+            <details class="quote">
+              <summary>{{ lang.t("Source code in") }} <code>
+                {%- if function.relative_filepath.is_absolute() -%}
+                  {{ function.relative_package_filepath }}
+                {%- else -%}
+                  {{ function.relative_filepath }}
+                {%- endif -%}
+              </code></summary>
+              {{ function.source|highlight(language="python", linestart=function.lineno or 0, linenums=True) }}
+            </details>
+          {% endif %}
+        {% endblock source %}
+      {% endblock contents %}
+    </div>
+
+  {% endwith %}
+</div>

docs/tests/unit_tests/test_api_reference.py

@@ -14,11 +14,11 @@ from langgraph.types import interrupt
 """
 
 EXPECTED_MARKDOWN = """\
+API Reference: <a href="https://langchain-ai.github.io/langgraph/reference/types/#langgraph.types.interrupt">interrupt</a>
+
 ```python
 from langgraph.types import interrupt
 ```
-
-API Reference: <a href="https://langchain-ai.github.io/langgraph/reference/types/#langgraph.types.interrupt">interrupt</a>
 """
 
 

libs/checkpoint-postgres/langgraph/checkpoint/postgres/__init__.py

@@ -357,6 +357,29 @@ class PostgresSaver(BasePostgresSaver):
                 ),
             )
 
+    def delete_thread(self, thread_id: str) -> None:
+        """Delete all checkpoints and writes associated with a thread ID.
+
+        Args:
+            thread_id (str): The thread ID to delete.
+
+        Returns:
+            None
+        """
+        with self._cursor(pipeline=True) as cur:
+            cur.execute(
+                "DELETE FROM checkpoints WHERE thread_id = %s",
+                (str(thread_id),),
+            )
+            cur.execute(
+                "DELETE FROM checkpoint_blobs WHERE thread_id = %s",
+                (str(thread_id),),
+            )
+            cur.execute(
+                "DELETE FROM checkpoint_writes WHERE thread_id = %s",
+                (str(thread_id),),
+            )
+
     @contextmanager
     def _cursor(self, *, pipeline: bool = False) -> Iterator[Cursor[DictRow]]:
         """Create a database cursor as a context manager.

libs/checkpoint-postgres/langgraph/checkpoint/postgres/aio.py

@@ -314,6 +314,29 @@ class AsyncPostgresSaver(BasePostgresSaver):
         async with self._cursor(pipeline=True) as cur:
             await cur.executemany(query, params)
 
+    async def adelete_thread(self, thread_id: str) -> None:
+        """Delete all checkpoints and writes associated with a thread ID.
+
+        Args:
+            thread_id (str): The thread ID to delete.
+
+        Returns:
+            None
+        """
+        async with self._cursor(pipeline=True) as cur:
+            await cur.execute(
+                "DELETE FROM checkpoints WHERE thread_id = %s",
+                (str(thread_id),),
+            )
+            await cur.execute(
+                "DELETE FROM checkpoint_blobs WHERE thread_id = %s",
+                (str(thread_id),),
+            )
+            await cur.execute(
+                "DELETE FROM checkpoint_writes WHERE thread_id = %s",
+                (str(thread_id),),
+            )
+
     @asynccontextmanager
     async def _cursor(
         self, *, pipeline: bool = False
@@ -481,5 +504,30 @@ class AsyncPostgresSaver(BasePostgresSaver):
             self.aput_writes(config, writes, task_id, task_path), self.loop
         ).result()
 
+    def delete_thread(self, thread_id: str) -> None:
+        """Delete all checkpoints and writes associated with a thread ID.
+
+        Args:
+            thread_id (str): The thread ID to delete.
+
+        Returns:
+            None
+        """
+        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 AsyncPostgresSaver 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.adelete_thread(thread_id), self.loop
+        ).result()
+
 
 __all__ = ["AsyncPostgresSaver", "AsyncShallowPostgresSaver", "Conn"]

libs/checkpoint-sqlite/langgraph/checkpoint/sqlite/__init__.py

@@ -464,6 +464,25 @@ class SqliteSaver(BaseCheckpointSaver[str]):
                 ],
             )
 
+    def delete_thread(self, thread_id: str) -> None:
+        """Delete all checkpoints and writes associated with a thread ID.
+
+        Args:
+            thread_id (str): The thread ID to delete.
+
+        Returns:
+            None
+        """
+        with self.cursor() as cur:
+            cur.execute(
+                "DELETE FROM checkpoints WHERE thread_id = ?",
+                (str(thread_id),),
+            )
+            cur.execute(
+                "DELETE FROM writes WHERE thread_id = ?",
+                (str(thread_id),),
+            )
+
     async def aget_tuple(self, config: RunnableConfig) -> Optional[CheckpointTuple]:
         """Get a checkpoint tuple from the database asynchronously.
 

libs/checkpoint-sqlite/langgraph/checkpoint/sqlite/aio.py

@@ -244,6 +244,31 @@ class AsyncSqliteSaver(BaseCheckpointSaver[str]):
             self.aput_writes(config, writes, task_id, task_path), self.loop
         ).result()
 
+    def delete_thread(self, thread_id: str) -> None:
+        """Delete all checkpoints and writes associated with a thread ID.
+
+        Args:
+            thread_id (str): The thread ID to delete.
+
+        Returns:
+            None
+        """
+        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 AsyncSqliteSaver are only allowed from a "
+                    "different thread. From the main thread, use the async interface. "
+                    "For example, use `checkpointer.alist(...)` or `await "
+                    "graph.ainvoke(...)`."
+                )
+        except RuntimeError:
+            pass
+        return asyncio.run_coroutine_threadsafe(
+            self.adelete_thread(thread_id), self.loop
+        ).result()
+
     async def setup(self) -> None:
         """Set up the checkpoint database asynchronously.
 
@@ -535,6 +560,26 @@ class AsyncSqliteSaver(BaseCheckpointSaver[str]):
             )
             await self.conn.commit()
 
+    async def adelete_thread(self, thread_id: str) -> None:
+        """Delete all checkpoints and writes associated with a thread ID.
+
+        Args:
+            thread_id (str): The thread ID to delete.
+
+        Returns:
+            None
+        """
+        async with self.lock, self.conn.cursor() as cur:
+            await cur.execute(
+                "DELETE FROM checkpoints WHERE thread_id = ?",
+                (str(thread_id),),
+            )
+            await cur.execute(
+                "DELETE FROM writes WHERE thread_id = ?",
+                (str(thread_id),),
+            )
+            await self.conn.commit()
+
     def get_next_version(self, current: Optional[str], channel: ChannelProtocol) -> str:
         """Generate the next version ID for a channel.
 

libs/checkpoint/langgraph/checkpoint/base/__init__.py

@@ -321,6 +321,17 @@ class BaseCheckpointSaver(Generic[V]):
         """
         raise NotImplementedError
 
+    def delete_thread(
+        self,
+        thread_id: str,
+    ) -> None:
+        """Delete all checkpoints and writes associated with a specific thread ID.
+
+        Args:
+            thread_id (str): The thread ID whose checkpoints should be deleted.
+        """
+        raise NotImplementedError
+
     async def aget(self, config: RunnableConfig) -> Optional[Checkpoint]:
         """Asynchronously fetch a checkpoint using the given configuration.
 
@@ -415,6 +426,17 @@ class BaseCheckpointSaver(Generic[V]):
         """
         raise NotImplementedError
 
+    async def adelete_thread(
+        self,
+        thread_id: str,
+    ) -> None:
+        """Delete all checkpoints and writes associated with a specific thread ID.
+
+        Args:
+            thread_id (str): The thread ID whose checkpoints should be deleted.
+        """
+        raise NotImplementedError
+
     def get_next_version(self, current: Optional[V], channel: ChannelProtocol) -> V:
         """Generate the next version ID for a channel.
 

libs/checkpoint/langgraph/checkpoint/memory/__init__.py

@@ -69,7 +69,7 @@ class InMemorySaver(
         ],
     ]
     writes: defaultdict[
-        tuple[str, str, str],
+        tuple[str, str, str],  # thread ID, checkpoint NS, checkpoint ID
         dict[tuple[str, int], tuple[str, str, tuple[str, bytes], str]],
     ]
     blobs: dict[
@@ -451,6 +451,24 @@ class InMemorySaver(
                 task_path,
             )
 
+    def delete_thread(self, thread_id: str) -> None:
+        """Delete all checkpoints and writes associated with a thread ID.
+
+        Args:
+            thread_id (str): The thread ID to delete.
+
+        Returns:
+            None
+        """
+        if thread_id in self.storage:
+            del self.storage[thread_id]
+        for k in list(self.writes.keys()):
+            if k[0] == thread_id:
+                del self.writes[k]
+        for k in list(self.blobs.keys()):
+            if k[0] == thread_id:
+                del self.blobs[k]
+
     async def aget_tuple(self, config: RunnableConfig) -> Optional[CheckpointTuple]:
         """Asynchronous version of get_tuple.
 
@@ -530,6 +548,17 @@ class InMemorySaver(
         """
         return self.put_writes(config, writes, task_id, task_path)
 
+    async def adelete_thread(self, thread_id: str) -> None:
+        """Delete all checkpoints and writes associated with a thread ID.
+
+        Args:
+            thread_id (str): The thread ID to delete.
+
+        Returns:
+            None
+        """
+        return self.delete_thread(thread_id)
+
     def get_next_version(self, current: Optional[str], channel: ChannelProtocol) -> str:
         if current is None:
             current_v = 0
@@ -615,4 +644,4 @@ class PersistentDict(defaultdict):
                 except Exception:
                     logging.error(f"Failed to load file: {fileobj.name}")
                     raise
-            raise ValueError("File not in a supported f ormat")
+            raise ValueError("File not in a supported format")

libs/cli/langgraph_cli/config.py

@@ -326,6 +326,10 @@ class Config(TypedDict, total=False):
     Must be >= 20 if provided.
     """
 
+    _INTERNAL_docker_tag: Optional[str]
+    """Optional. Internal use only.
+    """
+
     pip_config_file: Optional[str]
     """Optional. Path to a pip config file (e.g., "/etc/pip.conf" or "pip.ini") for controlling
     package installation (custom indices, credentials, etc.).
@@ -443,16 +447,22 @@ def _parse_node_version(version_str: str) -> int:
         ) from None
 
 
-def _is_python_graph(spec: Union[str, dict]) -> bool:
-    """Check if a graph is a Python graph based on the file extension."""
-
-    # handle new style config
+def _is_node_graph(spec: Union[str, dict]) -> bool:
+    """Check if a graph is a Node.js graph based on the file extension."""
     if isinstance(spec, dict):
         spec = spec.get("path")
 
     file_path = spec.split(":")[0]
     file_ext = os.path.splitext(file_path)[1]
-    return file_ext in [".py", ".pyx", ".pyd", ".pyi"]
+
+    return file_ext in [
+        ".ts",
+        ".mts",
+        ".cts",
+        ".js",
+        ".mjs",
+        ".cjs",
+    ]
 
 
 def validate_config(config: Config) -> Config:
@@ -460,8 +470,8 @@ def validate_config(config: Config) -> Config:
 
     graphs = config.get("graphs", {})
 
-    some_python = any(_is_python_graph(spec) for spec in graphs.values())
-    some_node = any(not _is_python_graph(spec) for spec in graphs.values())
+    some_node = any(_is_node_graph(spec) for spec in graphs.values())
+    some_python = any(not _is_node_graph(spec) for spec in graphs.values())
 
     node_version = config.get(
         "node_version", DEFAULT_NODE_VERSION if some_node else None
@@ -474,6 +484,7 @@ def validate_config(config: Config) -> Config:
         "node_version": node_version,
         "python_version": python_version,
         "pip_config_file": config.get("pip_config_file"),
+        "_INTERNAL_docker_tag": config.get("_INTERNAL_docker_tag"),
         "dependencies": config.get("dependencies", []),
         "dockerfile_lines": config.get("dockerfile_lines", []),
         "graphs": config.get("graphs", {}),
@@ -1019,7 +1030,9 @@ def _get_node_pm_install_cmd(config_path: pathlib.Path, config: Config) -> str:
 
 
 def python_config_to_docker(
-    config_path: pathlib.Path, config: Config, base_image: str
+    config_path: pathlib.Path,
+    config: Config,
+    base_image: str,
 ) -> tuple[str, dict[str, str]]:
     """Generate a Dockerfile from the configuration."""
     # configure pip
@@ -1034,6 +1047,8 @@ def python_config_to_docker(
         else ""
     )
 
+    docker_tag = config.get("_INTERNAL_docker_tag") or config["python_version"]
+
     # collect dependencies
     pypi_deps = [dep for dep in config["dependencies"] if not dep.startswith(".")]
     local_deps = _assemble_local_deps(config_path, config)
@@ -1154,7 +1169,7 @@ ADD {relpath} /deps/{name}
         )
 
     docker_file_contents = [
-        f"FROM {base_image}:{config['python_version']}",
+        f"FROM {base_image}:{docker_tag}",
         "",
         os.linesep.join(config["dockerfile_lines"]),
         "",
@@ -1186,10 +1201,13 @@ ADD {relpath} /deps/{name}
 
 
 def node_config_to_docker(
-    config_path: pathlib.Path, config: Config, base_image: str
+    config_path: pathlib.Path,
+    config: Config,
+    base_image: str,
 ) -> tuple[str, dict[str, str]]:
     faux_path = f"/deps/{config_path.parent.name}"
     install_cmd = _get_node_pm_install_cmd(config_path, config)
+    docker_tag = config.get("_INTERNAL_docker_tag") or config["node_version"]
 
     env_vars: list[str] = []
 
@@ -1216,7 +1234,7 @@ def node_config_to_docker(
     env_vars.append(f"ENV LANGSERVE_GRAPHS='{json.dumps(config['graphs'])}'")
 
     docker_file_contents = [
-        f"FROM {base_image}:{config['node_version']}",
+        f"FROM {base_image}:{docker_tag}",
         "",
         os.linesep.join(config["dockerfile_lines"]),
         "",
@@ -1240,8 +1258,13 @@ def default_base_image(config: Config) -> str:
     return "langchain/langgraph-api"
 
 
-def docker_tag(config: Config, base_image: Optional[str] = None) -> str:
+def docker_tag(
+    config: Config,
+    base_image: Optional[str] = None,
+) -> str:
     base_image = base_image or default_base_image(config)
+    if config.get("_INTERNAL_docker_tag"):
+        return f"{base_image}:{config['_INTERNAL_docker_tag']}"
 
     if config.get("node_version") and not config.get("python_version"):
         return f"{base_image}:{config['node_version']}"
@@ -1249,7 +1272,9 @@ def docker_tag(config: Config, base_image: Optional[str] = None) -> str:
 
 
 def config_to_docker(
-    config_path: pathlib.Path, config: Config, base_image: Optional[str] = None
+    config_path: pathlib.Path,
+    config: Config,
+    base_image: Optional[str] = None,
 ) -> tuple[str, dict[str, str]]:
     base_image = base_image or default_base_image(config)
 

libs/cli/langgraph_cli/progress.py

@@ -44,21 +44,30 @@ class Progress:
             sys.stdout.flush()
 
     def __enter__(self) -> Callable[[str], None]:
-        self.thread = threading.Thread(target=self.spinner_task)
-        self.thread.start()
+        if sys.stdout.isatty():
+            self.thread = threading.Thread(target=self.spinner_task)
+            self.thread.start()
 
-        def set_message(message):
-            self.message = message
-            if not message:
-                self.thread.join()
+            def set_message(message):
+                self.message = message
+                if not message:
+                    self.thread.join()
 
-        return set_message
+            return set_message
+        else:
+
+            def set_message(message):
+                sys.stderr.write(message + "\n")
+                sys.stderr.flush()
+
+            return set_message
 
     def __exit__(self, exception, value, tb):
-        self.message = ""
-        try:
-            self.thread.join()
-        finally:
-            del self.thread
-        if exception is not None:
-            return False
+        if sys.stdout.isatty():
+            self.message = ""
+            try:
+                self.thread.join()
+            finally:
+                del self.thread
+            if exception is not None:
+                return False

libs/cli/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langgraph-cli"
-version = "0.2.2"
+version = "0.2.5"
 description = "CLI for interacting with LangGraph API"
 authors = []
 license = "MIT"

libs/cli/schemas/schema.json

@@ -29,6 +29,17 @@
               ],
               "description": "Optional. Path to a pip config file (e.g., \"/etc/pip.conf\" or \"pip.ini\") for controlling\npackage installation (custom indices, credentials, etc.).\n\nOnly relevant if Python dependencies are installed via pip. If omitted, default pip settings are used.\n"
             },
+            "_INTERNAL_docker_tag": {
+              "anyOf": [
+                {
+                  "type": "string"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Internal use only.\n"
+            },
             "auth": {
               "anyOf": [
                 {
@@ -145,6 +156,17 @@
               ],
               "description": "Optional. Node.js version as a major version (e.g. '20'), if your deployment needs Node.\nMust be >= 20 if provided.\n"
             },
+            "_INTERNAL_docker_tag": {
+              "anyOf": [
+                {
+                  "type": "string"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Internal use only.\n"
+            },
             "auth": {
               "anyOf": [
                 {

libs/cli/schemas/schema.v0.json

@@ -29,6 +29,17 @@
               ],
               "description": "Optional. Path to a pip config file (e.g., \"/etc/pip.conf\" or \"pip.ini\") for controlling\npackage installation (custom indices, credentials, etc.).\n\nOnly relevant if Python dependencies are installed via pip. If omitted, default pip settings are used.\n"
             },
+            "_INTERNAL_docker_tag": {
+              "anyOf": [
+                {
+                  "type": "string"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Internal use only.\n"
+            },
             "auth": {
               "anyOf": [
                 {
@@ -145,6 +156,17 @@
               ],
               "description": "Optional. Node.js version as a major version (e.g. '20'), if your deployment needs Node.\nMust be >= 20 if provided.\n"
             },
+            "_INTERNAL_docker_tag": {
+              "anyOf": [
+                {
+                  "type": "string"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Internal use only.\n"
+            },
             "auth": {
               "anyOf": [
                 {

libs/cli/tests/unit_tests/test_config.py

@@ -29,6 +29,7 @@ def test_validate_config():
     }
     actual_config = validate_config(expected_config)
     expected_config = {
+        "_INTERNAL_docker_tag": None,
         "python_version": "3.11",
         "node_version": None,
         "pip_config_file": None,
@@ -47,6 +48,7 @@ def test_validate_config():
     # full config
     env = ".env"
     expected_config = {
+        "_INTERNAL_docker_tag": None,
         "python_version": "3.12",
         "node_version": None,
         "pip_config_file": "pipconfig.txt",
@@ -234,6 +236,17 @@ def test_validate_config_multiplatform():
     assert config["node_version"] == "20"
     assert config["python_version"] == "3.12"
 
+    # no known extension (assumes python)
+    config = validate_config(
+        {
+            "dependencies": ["./local", "./shared_utils"],
+            "graphs": {"agent": "local.workflow:graph"},
+            "env": ".env",
+        }
+    )
+    assert config["node_version"] is None
+    assert config["python_version"] == "3.11"
+
 
 # config_to_docker
 def test_config_to_docker_simple():
@@ -556,6 +569,39 @@ RUN (test ! -f /api/langgraph_api/js/build.mts && echo "Prebuild script not foun
     assert additional_contexts == {}
 
 
+def test_config_to_docker_nodejs_internal_docker_tag():
+    graphs = {"agent": "./graphs/agent.js:graph"}
+    actual_docker_stdin, additional_contexts = config_to_docker(
+        PATH_TO_CONFIG,
+        validate_config(
+            {
+                "node_version": "20",
+                "graphs": graphs,
+                "dockerfile_lines": ["ARG meow", "ARG foo"],
+                "auth": {"path": "./graphs/auth.mts:auth"},
+                "ui": {"agent": "./graphs/agent.ui.jsx"},
+                "ui_config": {"shared": ["nuqs"]},
+                "_INTERNAL_docker_tag": "my-tag",
+            }
+        ),
+        "langchain/langgraphjs-api",
+    )
+    expected_docker_stdin = """FROM langchain/langgraphjs-api:my-tag
+ARG meow
+ARG foo
+ADD . /deps/unit_tests
+RUN cd /deps/unit_tests && npm i
+ENV LANGGRAPH_AUTH='{"path": "./graphs/auth.mts:auth"}'
+ENV LANGGRAPH_UI='{"agent": "./graphs/agent.ui.jsx"}'
+ENV LANGGRAPH_UI_CONFIG='{"shared": ["nuqs"]}'
+ENV LANGSERVE_GRAPHS='{"agent": "./graphs/agent.js:graph"}'
+WORKDIR /deps/unit_tests
+RUN (test ! -f /api/langgraph_api/js/build.mts && echo "Prebuild script not found, skipping") || tsx /api/langgraph_api/js/build.mts"""
+
+    assert clean_empty_lines(actual_docker_stdin) == expected_docker_stdin
+    assert additional_contexts == {}
+
+
 def test_config_to_docker_gen_ui_python():
     graphs = {"agent": "./agent.py:graph"}
     actual_docker_stdin, additional_contexts = config_to_docker(

libs/langgraph/Makefile

@@ -62,7 +62,7 @@ MAXFAIL_ARGS := $(if $(MAXFAIL),--maxfail $(MAXFAIL),)
 XDIST_ARGS := $(if $(WORKERS),-x $(XDIST_ARGS),)
 
 test_watch:
-	make start-postgres && poetry run ptw . -- --ff -vv $(XDIST_ARGS) $(MAXFAIL_ARGS) --snapshot-update --tb short $(TEST); \
+	make start-postgres && poetry run ptw . -- --ff -vv $(XDIST_ARGS) $(MAXFAIL_ARGS) $(TEST); \
 	EXIT_CODE=$$?; \
 	make stop-postgres; \
 	exit $$EXIT_CODE

libs/langgraph/README.md

@@ -40,6 +40,9 @@ agent.invoke(
 )
 ```
 
+> [!TIP]
+> Check out [this guide](https://langchain-ai.github.io/langgraph/tutorials/workflows/) that walks through implementing common patterns (workflows and agents) in LangGraph. 
+
 ## Why use LangGraph?
 
 LangGraph is built for developers who want to build powerful, adaptable AI agents. Developers choose LangGraph for:

libs/langgraph/bench/fanout_to_subgraph.py

@@ -106,6 +106,7 @@ def fanout_to_subgraph_sync() -> StateGraph:
 if __name__ == "__main__":
     import asyncio
     import random
+    import time
 
     import uvloop
 
@@ -123,4 +124,7 @@ if __name__ == "__main__":
         len([c async for c in graph.astream(input, config=config)])
 
     uvloop.install()
+    start = time.time()
     asyncio.run(run())
+    end = time.time()
+    print(f"Time taken: {end - start:.4f} seconds")

libs/langgraph/langgraph/constants.py

@@ -69,6 +69,8 @@ CONFIG_KEY_ENSURE_LATEST = sys.intern("__pregel_ensure_latest")
 # (for distributed mode)
 CONFIG_KEY_DELEGATE = sys.intern("__pregel_delegate")
 # holds a boolean indicating whether to delegate subgraphs (for distributed mode)
+CONFIG_KEY_THREAD_ID = sys.intern("thread_id")
+# holds the thread ID for the current invocation
 CONFIG_KEY_CHECKPOINT_MAP = sys.intern("checkpoint_map")
 # holds a mapping of checkpoint_ns -> checkpoint_id for parent graphs
 CONFIG_KEY_CHECKPOINT_ID = sys.intern("checkpoint_id")

libs/langgraph/langgraph/func/__init__.py

@@ -9,6 +9,7 @@ from typing import (
     Callable,
     Generic,
     Optional,
+    Sequence,
     TypeVar,
     Union,
     get_args,
@@ -19,7 +20,7 @@ from typing import (
 from langgraph.channels.ephemeral_value import EphemeralValue
 from langgraph.channels.last_value import LastValue
 from langgraph.checkpoint.base import BaseCheckpointSaver
-from langgraph.constants import END, PREVIOUS, START, TAG_HIDDEN
+from langgraph.constants import END, PREVIOUS, START
 from langgraph.pregel import Pregel
 from langgraph.pregel.call import (
     P,
@@ -38,7 +39,7 @@ from langgraph.types import _DC_KWARGS, RetryPolicy, StreamMode
 def task(
     *,
     name: Optional[str] = None,
-    retry: Optional[RetryPolicy] = None,
+    retry: Optional[Union[RetryPolicy, Sequence[RetryPolicy]]] = None,
 ) -> Callable[
     [Union[Callable[P, Awaitable[T]], Callable[P, T]]],
     Callable[P, SyncAsyncFuture[T]],
@@ -55,7 +56,7 @@ def task(
     __func_or_none__: Optional[Union[Callable[P, Awaitable[T]], Callable[P, T]]] = None,
     *,
     name: Optional[str] = None,
-    retry: Optional[RetryPolicy] = None,
+    retry: Optional[Union[RetryPolicy, Sequence[RetryPolicy]]] = None,
 ) -> Union[
     Callable[
         [Union[Callable[P, Awaitable[T]], Callable[P, T]]],
@@ -119,6 +120,10 @@ def task(
         await add_one.ainvoke([1, 2, 3])  # Returns [2, 3, 4]
         ```
     """
+    if isinstance(retry, RetryPolicy):
+        retry_policies: Optional[Sequence[RetryPolicy]] = (retry,)
+    else:
+        retry_policies = retry
 
     def decorator(
         func: Union[Callable[P, Awaitable[T]], Callable[P, T]],
@@ -137,7 +142,7 @@ def task(
                 # handle regular functions / partials / callable classes, etc.
                 func.__name__ = name
 
-        call_func = functools.partial(call, func, retry=retry)
+        call_func = functools.partial(call, func, retry=retry_policies)
         object.__setattr__(call_func, "_is_pregel_task", True)
         return functools.update_wrapper(call_func, func)
 
@@ -429,8 +434,7 @@ class entrypoint:
                             [
                                 ChannelWriteEntry(END, mapper=_pluck_return_value),
                                 ChannelWriteEntry(PREVIOUS, mapper=_pluck_save_value),
-                            ],
-                            tags=[TAG_HIDDEN],
+                            ]
                         )
                     ],
                 )

libs/langgraph/langgraph/graph/branch.py

@@ -1,4 +1,3 @@
-import asyncio
 from inspect import (
     isfunction,
     ismethod,
@@ -178,7 +177,7 @@ class Branch(NamedTuple):
         ],
     ) -> Runnable:
         if reader:
-            value = await asyncio.to_thread(reader, config)
+            value = reader(config)
             # passthrough additional keys from node to branch
             # only doable when using dict states
             if (

libs/langgraph/langgraph/graph/graph.py

@@ -366,16 +366,14 @@ class CompiledGraph(Pregel):
         self.nodes[key] = (
             PregelNode(channels=[], triggers=[], metadata=node.metadata)
             | node.runnable
-            | ChannelWrite([ChannelWriteEntry(key)], tags=[TAG_HIDDEN])
+            | ChannelWrite([ChannelWriteEntry(key)])
         )
         cast(list[str], self.stream_channels).append(key)
 
     def attach_edge(self, start: str, end: str) -> None:
         if end == END:
             # publish to end channel
-            self.nodes[start].writers.append(
-                ChannelWrite([ChannelWriteEntry(END)], tags=[TAG_HIDDEN])
-            )
+            self.nodes[start].writers.append(ChannelWrite([ChannelWriteEntry(END)]))
         else:
             # subscribe to start channel
             self.nodes[end].triggers.append(start)
@@ -393,10 +391,7 @@ class CompiledGraph(Pregel):
                 )
                 for p in packets
             ]
-            return ChannelWrite(
-                cast(Sequence[Union[ChannelWriteEntry, Send]], writes),
-                tags=[TAG_HIDDEN],
-            )
+            return ChannelWrite(cast(Sequence[Union[ChannelWriteEntry, Send]], writes))
 
         # add hidden start node
         if start == START and start not in self.nodes:

libs/langgraph/langgraph/graph/state.py

@@ -43,8 +43,8 @@ from langgraph.constants import (
     MISSING,
     NS_END,
     NS_SEP,
-    SELF,
     TAG_HIDDEN,
+    TASKS,
 )
 from langgraph.errors import (
     ErrorCode,
@@ -77,9 +77,9 @@ from langgraph.pregel.write import (
 )
 from langgraph.store.base import BaseStore
 from langgraph.types import All, Checkpointer, Command, RetryPolicy
-from langgraph.utils.fields import get_field_default
+from langgraph.utils.fields import get_field_default, get_update_as_tuples
 from langgraph.utils.pydantic import create_model
-from langgraph.utils.runnable import RunnableCallable, RunnableLike, coerce_to_runnable
+from langgraph.utils.runnable import RunnableLike, coerce_to_runnable
 
 logger = logging.getLogger(__name__)
 
@@ -109,7 +109,7 @@ class StateNodeSpec(NamedTuple):
     runnable: Runnable
     metadata: Optional[dict[str, Any]]
     input: Type[Any]
-    retry_policy: Optional[RetryPolicy]
+    retry_policy: Optional[Union[RetryPolicy, Sequence[RetryPolicy]]]
     ends: Optional[Union[tuple[str, ...], dict[str, str]]] = EMPTY_SEQ
 
 
@@ -251,7 +251,7 @@ class StateGraph(Graph):
         *,
         metadata: Optional[dict[str, Any]] = None,
         input: Optional[Type[Any]] = None,
-        retry: Optional[RetryPolicy] = None,
+        retry: Optional[Union[RetryPolicy, Sequence[RetryPolicy]]] = None,
         destinations: Optional[Union[dict[str, str], tuple[str, ...]]] = None,
     ) -> Self:
         """Adds a new node to the state graph.
@@ -276,7 +276,7 @@ class StateGraph(Graph):
         *,
         metadata: Optional[dict[str, Any]] = None,
         input: Optional[Type[Any]] = None,
-        retry: Optional[RetryPolicy] = None,
+        retry: Optional[Union[RetryPolicy, Sequence[RetryPolicy]]] = None,
         destinations: Optional[Union[dict[str, str], tuple[str, ...]]] = None,
     ) -> Self:
         """Adds a new node to the state graph.
@@ -300,7 +300,7 @@ class StateGraph(Graph):
         *,
         metadata: Optional[dict[str, Any]] = None,
         input: Optional[Type[Any]] = None,
-        retry: Optional[RetryPolicy] = None,
+        retry: Optional[Union[RetryPolicy, Sequence[RetryPolicy]]] = None,
         destinations: Optional[Union[dict[str, str], tuple[str, ...]]] = None,
     ) -> Self:
         """Adds a new node to the state graph.
@@ -312,7 +312,8 @@ class StateGraph(Graph):
             action (Optional[RunnableLike]): The action associated with the node. (default: None)
             metadata (Optional[dict[str, Any]]): The metadata associated with the node. (default: None)
             input (Optional[Type[Any]]): The input schema for the node. (default: the graph's input schema)
-            retry (Optional[RetryPolicy]): The policy for retrying the node. (default: None)
+            retry (Optional[Union[RetryPolicy, Sequence[RetryPolicy]]]): The policy for retrying the node. (default: None)
+                If a sequence is provided, the first matching policy will be applied.
             destinations (Optional[Union[dict[str, str], tuple[str, ...]]]): Destinations that indicate where a node can route to.
                 This is useful for edgeless graphs with nodes that return `Command` objects.
                 If a dict is provided, the keys will be used as the target node names and the values will be used as the labels for the edges.
@@ -638,6 +639,7 @@ class StateGraph(Graph):
 
         compiled = CompiledStateGraph(
             builder=self,
+            schema_to_mapper={},
             config_type=self.config_schema,
             input_model=(
                 self.input
@@ -669,10 +671,6 @@ class StateGraph(Graph):
         for key, node in self.nodes.items():
             compiled.attach_node(key, node)
 
-        compiled.attach_branch(START, SELF, CONTROL_BRANCH, with_reader=False)
-        for key, node in self.nodes.items():
-            compiled.attach_branch(key, SELF, CONTROL_BRANCH, with_reader=False)
-
         for start, end in self.edges:
             compiled.attach_edge(start, end)
 
@@ -688,6 +686,16 @@ class StateGraph(Graph):
 
 class CompiledStateGraph(CompiledGraph):
     builder: StateGraph
+    schema_to_mapper: dict[Type[Any], Optional[Callable[[Any], Any]]]
+
+    def __init__(
+        self,
+        *,
+        schema_to_mapper: dict[Type[Any], Optional[Callable[[Any], Any]]],
+        **kwargs: Any,
+    ) -> None:
+        super().__init__(**kwargs)
+        self.schema_to_mapper = schema_to_mapper
 
     def get_input_schema(
         self, config: Optional[RunnableConfig] = None
@@ -723,28 +731,6 @@ class CompiledStateGraph(CompiledGraph):
                 if is_writable_managed_value(v)
             ]
 
-        def _get_root(input: Any) -> Optional[Sequence[tuple[str, Any]]]:
-            if isinstance(input, Command):
-                if input.graph == Command.PARENT:
-                    return ()
-                return input._update_as_tuples()
-            elif (
-                isinstance(input, (list, tuple))
-                and input
-                and any(isinstance(i, Command) for i in input)
-            ):
-                updates: list[tuple[str, Any]] = []
-                for i in input:
-                    if isinstance(i, Command):
-                        if i.graph == Command.PARENT:
-                            continue
-                        updates.extend(i._update_as_tuples())
-                    else:
-                        updates.append(("__root__", i))
-                return updates
-            elif input is not None:
-                return [("__root__", input)]
-
         def _get_updates(
             input: Union[None, dict, Any],
         ) -> Optional[Sequence[tuple[str, Any]]]:
@@ -775,32 +761,7 @@ class CompiledStateGraph(CompiledGraph):
                         updates.extend(_get_updates(i) or ())
                 return updates
             elif (t := type(input)) and get_type_hints(t):
-                # Pydantic v2
-                if isinstance(input, BaseModelV1):
-                    keep: Optional[set[str]] = input.__fields_set__
-                    defaults = {k: v.default for k, v in t.__fields__.items()}
-                elif isinstance(input, BaseModel):
-                    keep = input.model_fields_set
-                    defaults = {k: v.default for k, v in input.model_fields.items()}
-                # Pydantic v1
-                else:
-                    keep = None
-                    defaults = {}
-
-                # NOTE: This behavior for Pydantic is somewhat inelegant,
-                # but we keep around for backwards compatibility
-                # if input is a Pydantic model, only update values
-                # that are different from the default values or in the keep set
-                return [
-                    (k, value)
-                    for k in output_keys
-                    if (value := getattr(input, k, MISSING)) is not MISSING
-                    and (
-                        value is not None
-                        or defaults.get(k, MISSING) is not None
-                        or (keep is not None and k in keep)
-                    )
-                ]
+                return get_update_as_tuples(input, output_keys)
             else:
                 msg = create_error_message(
                     message=f"Expected dict, got {input}",
@@ -813,6 +774,7 @@ class CompiledStateGraph(CompiledGraph):
             ChannelWriteTupleEntry(
                 mapper=_get_root if output_keys == ["__root__"] else _get_updates
             ),
+            ChannelWriteTupleEntry(mapper=_control_branch),
         )
 
         # add node and output channel
@@ -821,12 +783,21 @@ class CompiledStateGraph(CompiledGraph):
                 tags=[TAG_HIDDEN],
                 triggers=[START],
                 channels=[START],
-                writers=[ChannelWrite(write_entries, tags=[TAG_HIDDEN])],
+                writers=[ChannelWrite(write_entries)],
             )
         elif node is not None:
             input_schema = node.input if node else self.builder.schema
             input_values = {k: k for k in self.builder.schemas[input_schema]}
             is_single_input = len(input_values) == 1 and "__root__" in input_values
+            if input_schema in self.schema_to_mapper:
+                mapper = self.schema_to_mapper[input_schema]
+            else:
+                mapper = _pick_mapper(
+                    list(input_values),
+                    input_schema,
+                    self.builder.type_hints[input_schema],
+                )
+                self.schema_to_mapper[input_schema] = mapper
 
             branch_channel = CHANNEL_BRANCH_TO.format(key)
             self.channels[branch_channel] = EphemeralValue(Any, guard=False)
@@ -835,13 +806,9 @@ class CompiledStateGraph(CompiledGraph):
                 # read state keys and managed values
                 channels=(list(input_values) if is_single_input else input_values),
                 # coerce state dict to schema class (eg. pydantic model)
-                mapper=_pick_mapper(
-                    list(input_values),
-                    input_schema,
-                    self.builder.type_hints[input_schema],
-                ),
+                mapper=mapper,
                 # publish to state keys
-                writers=[ChannelWrite(write_entries, tags=[TAG_HIDDEN])],
+                writers=[ChannelWrite(write_entries)],
                 metadata=node.metadata,
                 retry_policy=node.retry_policy,
                 bound=node.runnable,
@@ -867,9 +834,7 @@ class CompiledStateGraph(CompiledGraph):
             # publish to channel
             for start in starts:
                 self.nodes[start].writers.append(
-                    ChannelWrite(
-                        (ChannelWriteEntry(channel_name, start),), tags=[TAG_HIDDEN]
-                    )
+                    ChannelWrite((ChannelWriteEntry(channel_name, start),))
                 )
 
     def attach_branch(
@@ -900,19 +865,33 @@ class CompiledStateGraph(CompiledGraph):
                     config, cast(Sequence[Union[Send, ChannelWriteEntry]], writes)
                 )
 
-        schema = branch.input_schema or (
-            self.builder.nodes[start].input
-            if start in self.builder.nodes
-            else self.builder.schema
-        )
+        if with_reader:
+            # get schema
+            schema = branch.input_schema or (
+                self.builder.nodes[start].input
+                if start in self.builder.nodes
+                else self.builder.schema
+            )
+            channels = list(self.builder.schemas[schema])
+            # get mapper
+            if schema in self.schema_to_mapper:
+                mapper = self.schema_to_mapper[schema]
+            else:
+                mapper = _pick_mapper(channels, schema, self.builder.type_hints[schema])
+                self.schema_to_mapper[schema] = mapper
+            # create reader
+            reader: Optional[Callable[[RunnableConfig], Any]] = partial(
+                ChannelRead.do_read,
+                select=channels[0] if channels == ["__root__"] else channels,
+                fresh=True,
+                # coerce state dict to schema class (eg. pydantic model)
+                mapper=mapper,
+            )
+        else:
+            reader = None
 
         # attach branch publisher
-        self.nodes[start].writers.append(
-            branch.run(
-                branch_writer,
-                _get_state_reader(self.builder, schema) if with_reader else None,
-            )
-        )
+        self.nodes[start].writers.append(branch.run(branch_writer, reader))
 
         # attach then subscriber
         if branch.then and branch.then != END:
@@ -927,9 +906,7 @@ class CompiledStateGraph(CompiledGraph):
             for end in ends:
                 if end != END:
                     self.nodes[end].writers.append(
-                        ChannelWrite(
-                            [ChannelWriteEntry(channel_name, end)], tags=[TAG_HIDDEN]
-                        )
+                        ChannelWrite((ChannelWriteEntry(channel_name, end),))
                     )
 
     def _migrate_checkpoint(self, checkpoint: Checkpoint) -> None:
@@ -1037,20 +1014,6 @@ class CompiledStateGraph(CompiledGraph):
                         seen[INTERRUPT].pop(k, MISSING)
 
 
-def _get_state_reader(
-    builder: StateGraph, schema: Type[Any]
-) -> Callable[[RunnableConfig], Any]:
-    state_keys = list(builder.channels)
-    select = list(builder.schemas[schema])
-    return partial(
-        ChannelRead.do_read,
-        select=select[0] if select == ["__root__"] else select,
-        fresh=True,
-        # coerce state dict to schema class (eg. pydantic model)
-        mapper=_pick_mapper(state_keys, schema, builder.type_hints[schema]),
-    )
-
-
 def _pick_mapper(
     state_keys: Sequence[str], schema: Type[Any], type_hints: Optional[dict[str, Any]]
 ) -> Optional[Callable[[Any], Any]]:
@@ -1068,9 +1031,9 @@ def _coerce_state(schema: Type[Any], input: dict[str, Any]) -> dict[str, Any]:
     return schema(**input)
 
 
-def _control_branch(value: Any) -> Sequence[Union[str, Send]]:
+def _control_branch(value: Any) -> Sequence[tuple[str, Any]]:
     if isinstance(value, Send):
-        return [value]
+        return ((TASKS, value),)
     commands: list[Command] = []
     if isinstance(value, Command):
         commands.append(value)
@@ -1078,51 +1041,45 @@ def _control_branch(value: Any) -> Sequence[Union[str, Send]]:
         for cmd in value:
             if isinstance(cmd, Command):
                 commands.append(cmd)
-    rtn: list[Union[str, Send]] = []
+    rtn: list[tuple[str, Any]] = []
     for command in commands:
         if command.graph == Command.PARENT:
             raise ParentCommand(command)
         if isinstance(command.goto, Send):
-            rtn.append(command.goto)
+            rtn.append((TASKS, command.goto))
         elif isinstance(command.goto, str):
-            rtn.append(command.goto)
+            rtn.append((CHANNEL_BRANCH_TO.format(command.goto), None))
         else:
-            rtn.extend(command.goto)
+            rtn.extend(
+                (TASKS, go)
+                if isinstance(go, Send)
+                else (CHANNEL_BRANCH_TO.format(go), None)
+                for go in command.goto
+            )
     return rtn
 
 
-async def _acontrol_branch(value: Any) -> Sequence[Union[str, Send]]:
-    if isinstance(value, Send):
-        return [value]
-    commands: list[Command] = []
-    if isinstance(value, Command):
-        commands.append(value)
-    elif isinstance(value, (list, tuple)):
-        for cmd in value:
-            if isinstance(cmd, Command):
-                commands.append(cmd)
-    rtn: list[Union[str, Send]] = []
-    for command in commands:
-        if command.graph == Command.PARENT:
-            raise ParentCommand(command)
-        if isinstance(command.goto, Send):
-            rtn.append(command.goto)
-        elif isinstance(command.goto, str):
-            rtn.append(command.goto)
-        else:
-            rtn.extend(command.goto)
-    return rtn
-
-
-CONTROL_BRANCH_PATH = RunnableCallable(
-    _control_branch,
-    _acontrol_branch,
-    tags=[TAG_HIDDEN],
-    trace=False,
-    recurse=False,
-    func_accepts_config=False,
-)
-CONTROL_BRANCH = Branch(CONTROL_BRANCH_PATH, None)
+def _get_root(input: Any) -> Optional[Sequence[tuple[str, Any]]]:
+    if isinstance(input, Command):
+        if input.graph == Command.PARENT:
+            return ()
+        return input._update_as_tuples()
+    elif (
+        isinstance(input, (list, tuple))
+        and input
+        and any(isinstance(i, Command) for i in input)
+    ):
+        updates: list[tuple[str, Any]] = []
+        for i in input:
+            if isinstance(i, Command):
+                if i.graph == Command.PARENT:
+                    continue
+                updates.extend(i._update_as_tuples())
+            else:
+                updates.append(("__root__", i))
+        return updates
+    elif input is not None:
+        return [("__root__", input)]
 
 
 def _get_channels(

libs/langgraph/langgraph/pregel/__init__.py

@@ -66,6 +66,7 @@ from langgraph.constants import (
     CONFIG_KEY_STREAM,
     CONFIG_KEY_STREAM_WRITER,
     CONFIG_KEY_TASK_ID,
+    CONFIG_KEY_THREAD_ID,
     END,
     ERROR,
     INPUT,
@@ -498,8 +499,8 @@ class Pregel(PregelProtocol):
     store: Optional[BaseStore] = None
     """Memory store to use for SharedValues. Defaults to None."""
 
-    retry_policy: Optional[RetryPolicy] = None
-    """Retry policy to use when running tasks. Set to None to disable."""
+    retry_policy: Optional[Sequence[RetryPolicy]] = None
+    """Retry policies to use when running tasks. Set to None to disable."""
 
     config_type: Optional[Type[Any]] = None
 
@@ -528,7 +529,7 @@ class Pregel(PregelProtocol):
         debug: Optional[bool] = None,
         checkpointer: Optional[BaseCheckpointSaver] = None,
         store: Optional[BaseStore] = None,
-        retry_policy: Optional[RetryPolicy] = None,
+        retry_policy: Optional[Union[RetryPolicy, Sequence[RetryPolicy]]] = None,
         config_type: Optional[Type[Any]] = None,
         input_model: Optional[Type[BaseModel]] = None,
         config: Optional[RunnableConfig] = None,
@@ -548,7 +549,10 @@ class Pregel(PregelProtocol):
         self.debug = debug if debug is not None else get_debug()
         self.checkpointer = checkpointer
         self.store = store
-        self.retry_policy = retry_policy
+        if isinstance(retry_policy, RetryPolicy):
+            self.retry_policy: Sequence[RetryPolicy] = (retry_policy,)
+        else:
+            self.retry_policy = retry_policy
         self.config_type = config_type
         self.input_model = input_model
         self.config = config
@@ -1041,6 +1045,9 @@ class Pregel(PregelProtocol):
             config = merge_configs(
                 config, {CONF: {CONFIG_KEY_CHECKPOINT_NS: recast_checkpoint_ns(ns)}}
             )
+        thread_id = config[CONF][CONFIG_KEY_THREAD_ID]
+        if not isinstance(thread_id, str):
+            config[CONF][CONFIG_KEY_THREAD_ID] = str(thread_id)
 
         saved = checkpointer.get_tuple(config)
         return self._prepare_state_snapshot(
@@ -1080,6 +1087,9 @@ class Pregel(PregelProtocol):
             config = merge_configs(
                 config, {CONF: {CONFIG_KEY_CHECKPOINT_NS: recast_checkpoint_ns(ns)}}
             )
+        thread_id = config[CONF][CONFIG_KEY_THREAD_ID]
+        if not isinstance(thread_id, str):
+            config[CONF][CONFIG_KEY_THREAD_ID] = str(thread_id)
 
         saved = await checkpointer.aget_tuple(config)
         return await self._aprepare_state_snapshot(
@@ -1125,7 +1135,12 @@ class Pregel(PregelProtocol):
         config = merge_configs(
             self.config,
             config,
-            {CONF: {CONFIG_KEY_CHECKPOINT_NS: checkpoint_ns}},
+            {
+                CONF: {
+                    CONFIG_KEY_CHECKPOINT_NS: checkpoint_ns,
+                    CONFIG_KEY_THREAD_ID: str(config[CONF][CONFIG_KEY_THREAD_ID]),
+                }
+            },
         )
         # eagerly consume list() to avoid holding up the db cursor
         for checkpoint_tuple in list(
@@ -1172,7 +1187,12 @@ class Pregel(PregelProtocol):
         config = merge_configs(
             self.config,
             config,
-            {CONF: {CONFIG_KEY_CHECKPOINT_NS: checkpoint_ns}},
+            {
+                CONF: {
+                    CONFIG_KEY_CHECKPOINT_NS: checkpoint_ns,
+                    CONFIG_KEY_THREAD_ID: str(config[CONF][CONFIG_KEY_THREAD_ID]),
+                }
+            },
         )
         # eagerly consume list() to avoid holding up the db cursor
         for checkpoint_tuple in [
@@ -1592,7 +1612,9 @@ class Pregel(PregelProtocol):
 
             return patch_checkpoint_map(next_config, saved.metadata if saved else None)
 
-        current_config = config
+        current_config = patch_configurable(
+            config, {CONFIG_KEY_THREAD_ID: str(config[CONF][CONFIG_KEY_THREAD_ID])}
+        )
         for superstep in supersteps:
             current_config = perform_superstep(current_config, superstep)
         return current_config
@@ -2002,7 +2024,9 @@ class Pregel(PregelProtocol):
                     await checkpointer.aput_writes(next_config, push_writes, task_id)
             return patch_checkpoint_map(next_config, saved.metadata if saved else None)
 
-        current_config = config
+        current_config = patch_configurable(
+            config, {CONFIG_KEY_THREAD_ID: str(config[CONF][CONFIG_KEY_THREAD_ID])}
+        )
         for superstep in supersteps:
             current_config = await aperform_superstep(current_config, superstep)
         return current_config
@@ -2547,11 +2571,12 @@ class Pregel(PregelProtocol):
         do_stream = (
             next(
                 (
-                    cast(_StreamingCallbackHandler, h)
+                    True
                     for h in run_manager.handlers
                     if isinstance(h, _StreamingCallbackHandler)
+                    and not isinstance(h, StreamMessagesHandler)
                 ),
-                None,
+                False,
             )
             if _StreamingCallbackHandler is not None
             else False
@@ -2621,7 +2646,7 @@ class Pregel(PregelProtocol):
                     ),
                     put_writes=weakref.WeakMethod(loop.put_writes),
                     schedule_task=weakref.WeakMethod(loop.accept_push),
-                    use_astream=do_stream is not None,
+                    use_astream=do_stream,
                     node_finished=config[CONF].get(CONFIG_KEY_NODE_FINISHED),
                 )
                 # enable subgraph streaming

libs/langgraph/langgraph/pregel/algo.py

@@ -3,13 +3,13 @@ import itertools
 import sys
 import threading
 from collections import defaultdict, deque
+from copy import copy
 from functools import partial
 from hashlib import sha1
 from typing import (
     Any,
     Callable,
     Iterable,
-    Iterator,
     Literal,
     Mapping,
     NamedTuple,
@@ -49,6 +49,7 @@ from langgraph.constants import (
     EMPTY_SEQ,
     ERROR,
     INTERRUPT,
+    MISSING,
     NO_WRITES,
     NS_END,
     NS_SEP,
@@ -63,12 +64,12 @@ from langgraph.constants import (
     TASKS,
     Send,
 )
-from langgraph.errors import EmptyChannelError, InvalidUpdateError
+from langgraph.errors import InvalidUpdateError
 from langgraph.managed.base import ManagedValueMapping
 from langgraph.pregel.call import get_runnable_for_task
-from langgraph.pregel.io import read_channel, read_channels
+from langgraph.pregel.io import read_channels
 from langgraph.pregel.log import logger
-from langgraph.pregel.read import PregelNode
+from langgraph.pregel.read import INPUT_CACHE_KEY_TYPE, PregelNode
 from langgraph.store.base import BaseStore
 from langgraph.types import (
     All,
@@ -115,7 +116,7 @@ class Call:
 
     func: Callable
     input: Any
-    retry: Optional[RetryPolicy]
+    retry: Optional[Sequence[RetryPolicy]]
     callbacks: Callbacks
 
     def __init__(
@@ -123,7 +124,7 @@ class Call:
         func: Callable,
         input: Any,
         *,
-        retry: Optional[RetryPolicy],
+        retry: Optional[Sequence[RetryPolicy]],
         callbacks: Callbacks,
     ) -> None:
         self.func = func
@@ -423,6 +424,7 @@ def prepare_next_tasks(
         are the tasks themselves. This is the union of all PUSH tasks (Sends)
         and PULL tasks (nodes triggered by edges).
     """
+    input_cache: dict[INPUT_CACHE_KEY_TYPE, Any] = {}
     checkpoint_id_bytes = binascii.unhexlify(checkpoint["id"].replace("-", ""))
     null_version = checkpoint_null_version(checkpoint)
     tasks: list[Union[PregelTask, PregelExecutableTask]] = []
@@ -444,6 +446,7 @@ def prepare_next_tasks(
             store=store,
             checkpointer=checkpointer,
             manager=manager,
+            input_cache=input_cache,
         ):
             tasks.append(task)
 
@@ -486,6 +489,7 @@ def prepare_next_tasks(
             store=store,
             checkpointer=checkpointer,
             manager=manager,
+            input_cache=input_cache,
         ):
             tasks.append(task)
     return {t.id: t for t in tasks}
@@ -511,6 +515,7 @@ def prepare_single_task(
     store: Optional[BaseStore] = None,
     checkpointer: Optional[BaseCheckpointSaver] = None,
     manager: Union[None, ParentRunManager, AsyncParentRunManager] = None,
+    input_cache: Optional[dict[INPUT_CACHE_KEY_TYPE, Any]] = None,
 ) -> Union[None, PregelTask, PregelExecutableTask]:
     """Prepares a single task for the next Pregel step, given a task path, which
     uniquely identifies a PUSH or PULL task within the graph."""
@@ -729,11 +734,15 @@ def prepare_single_task(
         ):
             triggers = tuple(sorted(proc.triggers))
             try:
-                val = next(
-                    _proc_input(proc, managed, channels, for_execution=for_execution)
+                val = _proc_input(
+                    proc,
+                    managed,
+                    channels,
+                    for_execution=for_execution,
+                    input_cache=input_cache,
                 )
-            except StopIteration:
-                return
+                if val is MISSING:
+                    return
             except Exception as exc:
                 if SUPPORTS_EXC_NOTES:
                     exc.add_note(
@@ -926,34 +935,32 @@ def _proc_input(
     channels: Mapping[str, BaseChannel],
     *,
     for_execution: bool,
-) -> Iterator[Any]:
+    input_cache: Optional[dict[INPUT_CACHE_KEY_TYPE, Any]],
+) -> Any:
     """Prepare input for a PULL task, based on the process's channels and triggers."""
+    # if in cache return shallow copy
+    if input_cache is not None and proc.input_cache_key in input_cache:
+        return copy(input_cache[proc.input_cache_key])
     # If all trigger channels subscribed by this process are not empty
     # then invoke the process with the values of all non-empty channels
     if isinstance(proc.channels, dict):
-        try:
-            val: dict[str, Any] = {}
-            for k, chan in proc.channels.items():
-                if chan in proc.triggers:
-                    val[k] = read_channel(channels, chan, catch=False)
-                elif chan in channels:
-                    try:
-                        val[k] = read_channel(channels, chan, catch=False)
-                    except EmptyChannelError:
-                        continue
-                else:
-                    val[k] = managed[k]()
-        except EmptyChannelError:
-            return
+        val: dict[str, Any] = {}
+        for k, chan in proc.channels.items():
+            if chan in channels:
+                if channels[chan].is_available():
+                    val[k] = channels[chan].get()
+            else:
+                val[k] = managed[k]()
     elif isinstance(proc.channels, list):
         for chan in proc.channels:
-            try:
-                val = read_channel(channels, chan, catch=False)
-                break
-            except EmptyChannelError:
-                pass
+            if chan in channels:
+                if channels[chan].is_available():
+                    val = channels[chan].get()
+                    break
+            else:
+                val[k] = managed[k]()
         else:
-            return
+            return MISSING
     else:
         raise RuntimeError(
             "Invalid channels type, expected list or dict, got {proc.channels}"
@@ -963,7 +970,11 @@ def _proc_input(
     if for_execution and proc.mapper is not None:
         val = proc.mapper(val)
 
-    yield val
+    # Cache the input value
+    if input_cache is not None:
+        input_cache[proc.input_cache_key] = val
+
+    return val
 
 
 def _uuid5_str(namespace: bytes, *parts: str) -> str:

libs/langgraph/langgraph/pregel/call.py

@@ -5,12 +5,12 @@ import functools
 import inspect
 import sys
 import types
-from typing import Any, Callable, Generator, Generic, Optional, TypeVar, cast
+from typing import Any, Callable, Generator, Generic, Optional, Sequence, TypeVar, cast
 
 from langchain_core.runnables import Runnable
 from typing_extensions import ParamSpec
 
-from langgraph.constants import CONF, CONFIG_KEY_CALL, RETURN, TAG_HIDDEN
+from langgraph.constants import CONF, CONFIG_KEY_CALL, RETURN
 from langgraph.pregel.write import ChannelWrite, ChannelWriteEntry
 from langgraph.types import RetryPolicy
 from langgraph.utils.config import get_config
@@ -197,7 +197,7 @@ def get_runnable_for_task(func: Callable[..., Any]) -> RunnableSeq:
             )
         seq = RunnableSeq(
             run,
-            ChannelWrite([ChannelWriteEntry(RETURN)], tags=[TAG_HIDDEN]),
+            ChannelWrite([ChannelWriteEntry(RETURN)]),
             name=name,
             trace_inputs=functools.partial(
                 _explode_args_trace_inputs, inspect.signature(func)
@@ -224,7 +224,7 @@ class SyncAsyncFuture(Generic[T], concurrent.futures.Future[T]):
 def call(
     func: Callable[P, T],
     *args: Any,
-    retry: Optional[RetryPolicy] = None,
+    retry: Optional[Sequence[RetryPolicy]] = None,
     **kwargs: Any,
 ) -> SyncAsyncFuture[T]:
     config = get_config()

libs/langgraph/langgraph/pregel/loop.py

@@ -52,6 +52,7 @@ from langgraph.constants import (
     CONFIG_KEY_SCRATCHPAD,
     CONFIG_KEY_STREAM,
     CONFIG_KEY_TASK_ID,
+    CONFIG_KEY_THREAD_ID,
     EMPTY_SEQ,
     ERROR,
     INPUT,
@@ -285,6 +286,12 @@ class PregelLoop(LoopProtocol):
             )
         else:
             self.checkpoint_config = self.config
+        if thread_id := self.checkpoint_config[CONF].get(CONFIG_KEY_THREAD_ID):
+            if not isinstance(thread_id, str):
+                self.checkpoint_config = patch_configurable(
+                    self.checkpoint_config,
+                    {CONFIG_KEY_THREAD_ID: str(thread_id)},
+                )
         self.checkpoint_ns = (
             tuple(cast(str, self.config[CONF][CONFIG_KEY_CHECKPOINT_NS]).split(NS_SEP))
             if self.config[CONF].get(CONFIG_KEY_CHECKPOINT_NS)
@@ -1043,16 +1050,16 @@ class SyncPregelLoop(PregelLoop, ContextManager):
             saved = None
         if saved is None:
             saved = CheckpointTuple(
-                self.config, empty_checkpoint(), {"step": -2}, None, []
+                self.checkpoint_config, empty_checkpoint(), {"step": -2}, None, []
             )
         elif self._migrate_checkpoint is not None:
             self._migrate_checkpoint(saved.checkpoint)
         self.checkpoint_config = {
-            **self.config,
+            **self.checkpoint_config,
             **saved.config,
             CONF: {
                 CONFIG_KEY_CHECKPOINT_NS: "",
-                **self.config.get(CONF, {}),
+                **self.checkpoint_config.get(CONF, {}),
                 **saved.config.get(CONF, {}),
             },
         }
@@ -1193,16 +1200,16 @@ class AsyncPregelLoop(PregelLoop, AsyncContextManager):
             saved = None
         if saved is None:
             saved = CheckpointTuple(
-                self.config, empty_checkpoint(), {"step": -2}, None, []
+                self.checkpoint_config, empty_checkpoint(), {"step": -2}, None, []
             )
         elif self._migrate_checkpoint is not None:
             self._migrate_checkpoint(saved.checkpoint)
         self.checkpoint_config = {
-            **self.config,
+            **self.checkpoint_config,
             **saved.config,
             CONF: {
                 CONFIG_KEY_CHECKPOINT_NS: "",
-                **self.config.get(CONF, {}),
+                **self.checkpoint_config.get(CONF, {}),
                 **saved.config.get(CONF, {}),
             },
         }

libs/langgraph/langgraph/pregel/messages.py

@@ -18,7 +18,7 @@ from langchain_core.messages import BaseMessage
 from langchain_core.outputs import ChatGenerationChunk, LLMResult
 
 from langgraph.constants import NS_SEP, TAG_HIDDEN, TAG_NOSTREAM
-from langgraph.types import StreamChunk
+from langgraph.types import Command, StreamChunk
 
 try:
     from langchain_core.tracers._streaming import _StreamingCallbackHandler
@@ -153,6 +153,17 @@ class StreamMessagesHandler(BaseCallbackHandler, _StreamingCallbackHandler):
         **kwargs: Any,
     ) -> Any:
         if meta := self.metadata.pop(run_id, None):
+            if isinstance(response, Command):
+                response = response.update
+
+            if isinstance(response, Sequence) and any(
+                isinstance(value, Command) for value in response
+            ):
+                response = [
+                    value.update if isinstance(value, Command) else value
+                    for value in response
+                ]
+
             if isinstance(response, BaseMessage):
                 self._emit(meta, response, dedupe=True)
             elif isinstance(response, Sequence):

libs/langgraph/langgraph/pregel/read.py

@@ -30,6 +30,7 @@ from langgraph.utils.config import merge_configs
 from langgraph.utils.runnable import RunnableCallable, RunnableSeq
 
 READ_TYPE = Callable[[Union[str, Sequence[str]], bool], Union[Any, dict[str, Any]]]
+INPUT_CACHE_KEY_TYPE = tuple[Callable[..., Any], tuple[str, ...]]
 
 
 class ChannelRead(RunnableCallable):
@@ -67,6 +68,7 @@ class ChannelRead(RunnableCallable):
             afunc=self._aread,
             tags=tags,
             name=None,
+            trace=False,
             func_accepts_config=True,
         )
         self.fresh = fresh
@@ -144,8 +146,8 @@ class PregelNode(Runnable):
     """The main logic of the node. This will be invoked with the input from 
     `channels`."""
 
-    retry_policy: Optional[RetryPolicy]
-    """The retry policy to use when invoking the node."""
+    retry_policy: Optional[Sequence[RetryPolicy]]
+    """The retry policies to use when invoking the node."""
 
     tags: Optional[Sequence[str]]
     """Tags to attach to the node for tracing."""
@@ -166,7 +168,7 @@ class PregelNode(Runnable):
         tags: Optional[list[str]] = None,
         metadata: Optional[Mapping[str, Any]] = None,
         bound: Optional[Runnable[Any, Any]] = None,
-        retry_policy: Optional[RetryPolicy] = None,
+        retry_policy: Optional[Union[RetryPolicy, Sequence[RetryPolicy]]] = None,
         subgraphs: Optional[Sequence[PregelProtocol]] = None,
     ) -> None:
         self.channels = channels
@@ -174,7 +176,10 @@ class PregelNode(Runnable):
         self.mapper = mapper
         self.writers = writers or []
         self.bound = bound if bound is not None else DEFAULT_BOUND
-        self.retry_policy = retry_policy
+        if isinstance(retry_policy, RetryPolicy):
+            self.retry_policy: Sequence[RetryPolicy] = (retry_policy,)
+        else:
+            self.retry_policy = retry_policy
         self.tags = tags
         self.metadata = metadata
         if subgraphs is not None:
@@ -228,6 +233,17 @@ class PregelNode(Runnable):
         else:
             return self.bound
 
+    @cached_property
+    def input_cache_key(self) -> INPUT_CACHE_KEY_TYPE:
+        """Get a cache key for the input to the node.
+        This is used to avoid calculating the same input multiple times."""
+        return (
+            self.mapper,
+            tuple(f"{key}:{value}" for key, value in self.channels.items())
+            if isinstance(self.channels, dict)
+            else tuple(self.channels),
+        )
+
     def join(self, channels: Sequence[str]) -> PregelNode:
         assert isinstance(channels, list) or isinstance(
             channels, tuple

libs/langgraph/langgraph/pregel/retry.py

@@ -22,12 +22,11 @@ SUPPORTS_EXC_NOTES = sys.version_info >= (3, 11)
 
 def run_with_retry(
     task: PregelExecutableTask,
-    retry_policy: Optional[RetryPolicy],
+    retry_policy: Optional[Sequence[RetryPolicy]],
     configurable: Optional[dict[str, Any]] = None,
 ) -> None:
     """Run a task with retries."""
     retry_policy = task.retry_policy or retry_policy
-    interval = retry_policy.initial_interval if retry_policy else 0
     attempts = 0
     config = task.config
     if configurable is not None:
@@ -63,38 +62,39 @@ def run_with_retry(
                 exc.add_note(f"During task with name '{task.name}' and id '{task.id}'")
             if retry_policy is None:
                 raise
+
+            # Check which retry policy applies to this exception
+            matching_policy = None
+            for policy in retry_policy:
+                if _should_retry_on(policy, exc):
+                    matching_policy = policy
+                    break
+
+            if not matching_policy:
+                raise
+
             # increment attempts
             attempts += 1
-            # check if we should retry
-            if isinstance(retry_policy.retry_on, Sequence):
-                if not isinstance(exc, tuple(retry_policy.retry_on)):
-                    raise
-            elif isinstance(retry_policy.retry_on, type) and issubclass(
-                retry_policy.retry_on, Exception
-            ):
-                if not isinstance(exc, retry_policy.retry_on):
-                    raise
-            elif callable(retry_policy.retry_on):
-                if not retry_policy.retry_on(exc):  # type: ignore[call-arg]
-                    raise
-            else:
-                raise TypeError(
-                    "retry_on must be an Exception class, a list or tuple of Exception classes, or a callable"
-                )
             # check if we should give up
-            if attempts >= retry_policy.max_attempts:
+            if attempts >= matching_policy.max_attempts:
                 raise
             # sleep before retrying
+            interval = matching_policy.initial_interval
+            # Apply backoff factor based on attempt count
             interval = min(
-                retry_policy.max_interval,
-                interval * retry_policy.backoff_factor,
+                matching_policy.max_interval,
+                interval * (matching_policy.backoff_factor ** (attempts - 1)),
             )
-            time.sleep(
-                interval + random.uniform(0, 1) if retry_policy.jitter else interval
+
+            # Apply jitter if configured
+            sleep_time = (
+                interval + random.uniform(0, 1) if matching_policy.jitter else interval
             )
+            time.sleep(sleep_time)
+
             # log the retry
             logger.info(
-                f"Retrying task {task.name} after {interval:.2f} seconds (attempt {attempts}) after {exc.__class__.__name__} {exc}",
+                f"Retrying task {task.name} after {sleep_time:.2f} seconds (attempt {attempts}) after {exc.__class__.__name__} {exc}",
                 exc_info=exc,
             )
             # signal subgraphs to resume (if available)
@@ -103,13 +103,12 @@ def run_with_retry(
 
 async def arun_with_retry(
     task: PregelExecutableTask,
-    retry_policy: Optional[RetryPolicy],
+    retry_policies: Optional[Sequence[RetryPolicy]],
     stream: bool = False,
     configurable: Optional[dict[str, Any]] = None,
 ) -> None:
     """Run a task asynchronously with retries."""
-    retry_policy = task.retry_policy or retry_policy
-    interval = retry_policy.initial_interval if retry_policy else 0
+    retry_policies = task.retry_policy or retry_policies
     attempts = 0
     config = task.config
     if configurable is not None:
@@ -149,41 +148,58 @@ async def arun_with_retry(
         except Exception as exc:
             if SUPPORTS_EXC_NOTES:
                 exc.add_note(f"During task with name '{task.name}' and id '{task.id}'")
-            if retry_policy is None:
+            if retry_policies is None:
                 raise
+
+            # Check which retry policy applies to this exception
+            matching_policy = None
+            for policy in retry_policies:
+                if _should_retry_on(policy, exc):
+                    matching_policy = policy
+                    break
+
+            if not matching_policy:
+                raise
+
             # increment attempts
             attempts += 1
-            # check if we should retry
-            if isinstance(retry_policy.retry_on, Sequence):
-                if not isinstance(exc, tuple(retry_policy.retry_on)):
-                    raise
-            elif isinstance(retry_policy.retry_on, type) and issubclass(
-                retry_policy.retry_on, Exception
-            ):
-                if not isinstance(exc, retry_policy.retry_on):
-                    raise
-            elif callable(retry_policy.retry_on):
-                if not retry_policy.retry_on(exc):  # type: ignore[call-arg]
-                    raise
-            else:
-                raise TypeError(
-                    "retry_on must be an Exception class, a list or tuple of Exception classes, or a callable"
-                )
             # check if we should give up
-            if attempts >= retry_policy.max_attempts:
+            if attempts >= matching_policy.max_attempts:
                 raise
             # sleep before retrying
+            interval = matching_policy.initial_interval
+            # Apply backoff factor based on attempt count
             interval = min(
-                retry_policy.max_interval,
-                interval * retry_policy.backoff_factor,
+                matching_policy.max_interval,
+                interval * (matching_policy.backoff_factor ** (attempts - 1)),
             )
-            await asyncio.sleep(
-                interval + random.uniform(0, 1) if retry_policy.jitter else interval
+
+            # Apply jitter if configured
+            sleep_time = (
+                interval + random.uniform(0, 1) if matching_policy.jitter else interval
             )
+            await asyncio.sleep(sleep_time)
+
             # log the retry
             logger.info(
-                f"Retrying task {task.name} after {interval:.2f} seconds (attempt {attempts}) after {exc.__class__.__name__} {exc}",
+                f"Retrying task {task.name} after {sleep_time:.2f} seconds (attempt {attempts}) after {exc.__class__.__name__} {exc}",
                 exc_info=exc,
             )
             # signal subgraphs to resume (if available)
             config = patch_configurable(config, {CONFIG_KEY_RESUMING: True})
+
+
+def _should_retry_on(retry_policy: RetryPolicy, exc: Exception) -> bool:
+    """Check if the given exception should be retried based on the retry policy."""
+    if isinstance(retry_policy.retry_on, Sequence):
+        return isinstance(exc, tuple(retry_policy.retry_on))
+    elif isinstance(retry_policy.retry_on, type) and issubclass(
+        retry_policy.retry_on, Exception
+    ):
+        return isinstance(exc, retry_policy.retry_on)
+    elif callable(retry_policy.retry_on):
+        return retry_policy.retry_on(exc)  # type: ignore[call-arg]
+    else:
+        raise TypeError(
+            "retry_on must be an Exception class, a list or tuple of Exception classes, or a callable"
+        )

libs/langgraph/langgraph/pregel/runner.py

@@ -140,7 +140,7 @@ class PregelRunner:
         *,
         reraise: bool = True,
         timeout: Optional[float] = None,
-        retry_policy: Optional[RetryPolicy] = None,
+        retry_policy: Optional[Sequence[RetryPolicy]] = None,
         get_waiter: Optional[Callable[[], concurrent.futures.Future[None]]] = None,
     ) -> Iterator[None]:
         tasks = tuple(tasks)
@@ -269,7 +269,7 @@ class PregelRunner:
         *,
         reraise: bool = True,
         timeout: Optional[float] = None,
-        retry_policy: Optional[RetryPolicy] = None,
+        retry_policy: Optional[Sequence[RetryPolicy]] = None,
         get_waiter: Optional[Callable[[], asyncio.Future[None]]] = None,
     ) -> AsyncIterator[None]:
         loop = asyncio.get_event_loop()
@@ -519,7 +519,7 @@ def _call(
     func: Callable[[Any], Union[Awaitable[Any], Any]],
     input: Any,
     *,
-    retry: Optional[RetryPolicy] = None,
+    retry: Optional[Sequence[RetryPolicy]] = None,
     callbacks: Callbacks = None,
     futures: weakref.ref[FuturesDict],
     schedule_task: weakref.ref[
@@ -600,7 +600,7 @@ def _acall(
     func: Callable[[Any], Union[Awaitable[Any], Any]],
     input: Any,
     *,
-    retry: Optional[RetryPolicy] = None,
+    retry: Optional[Sequence[RetryPolicy]] = None,
     callbacks: Callbacks = None,
     # injected dependencies
     futures: weakref.ref[FuturesDict],

libs/langgraph/langgraph/pregel/write.py

@@ -54,14 +54,14 @@ class ChannelWrite(RunnableCallable):
         self,
         writes: Sequence[Union[ChannelWriteEntry, ChannelWriteTupleEntry, Send]],
         *,
-        tags: Optional[Sequence[str]] = None,
+        tags: Optional[Sequence[str]] = None,  # ignored
         require_at_least_one_of: Optional[Sequence[str]] = None,  # ignored
     ):
         super().__init__(
             func=self._write,
             afunc=self._awrite,
             name=None,
-            tags=tags,
+            trace=False,
             func_accepts_config=True,
         )
         self.writes = cast(
@@ -152,6 +152,8 @@ class ChannelWrite(RunnableCallable):
                 tuples.append((w.channel, value))
             else:
                 raise ValueError(f"Invalid write entry: {w}")
+        # if we want to persist writes found before hitting a ParentCommand
+        # can move this to a finally block
         write: TYPE_SEND = config[CONF][CONFIG_KEY_SEND]
         write(tuples)
 

libs/langgraph/langgraph/types.py

@@ -23,6 +23,7 @@ from langchain_core.runnables import Runnable, RunnableConfig
 from typing_extensions import Self
 
 from langgraph.checkpoint.base import BaseCheckpointSaver, CheckpointMetadata
+from langgraph.utils.fields import get_update_as_tuples
 
 if TYPE_CHECKING:
     from langgraph.pregel.protocol import PregelProtocol
@@ -75,6 +76,10 @@ def default_retry_on(exc: Exception) -> bool:
 
     if isinstance(exc, ConnectionError):
         return True
+    if isinstance(exc, httpx.HTTPStatusError):
+        return 500 <= exc.response.status_code < 600
+    if isinstance(exc, requests.HTTPError):
+        return 500 <= exc.response.status_code < 600 if exc.response else True
     if isinstance(
         exc,
         (
@@ -93,10 +98,6 @@ def default_retry_on(exc: Exception) -> bool:
         ),
     ):
         return False
-    if isinstance(exc, httpx.HTTPStatusError):
-        return 500 <= exc.response.status_code < 600
-    if isinstance(exc, requests.HTTPError):
-        return 500 <= exc.response.status_code < 600 if exc.response else True
     return True
 
 
@@ -172,7 +173,7 @@ class PregelExecutableTask:
     writes: deque[tuple[str, Any]]
     config: RunnableConfig
     triggers: Sequence[str]
-    retry_policy: Optional[RetryPolicy]
+    retry_policy: Optional[Sequence[RetryPolicy]]
     cache_policy: Optional[CachePolicy]
     id: str
     path: tuple[Union[str, int, tuple], ...]
@@ -318,7 +319,7 @@ class Command(Generic[N], ToolOutputMixin):
         ):
             return self.update
         elif hints := get_type_hints(type(self.update)):
-            return [(k, getattr(self.update, k)) for k in hints]
+            return get_update_as_tuples(self.update, tuple(hints.keys()))
         elif self.update is not None:
             return [("__root__", self.update)]
         else:

libs/langgraph/langgraph/utils/fields.py

@@ -1,8 +1,14 @@
 import dataclasses
-from typing import Any, Generator, Optional, Type, Union, get_type_hints
+from typing import Any, Generator, Optional, Sequence, Type, Union, get_type_hints
 
+from pydantic import BaseModel
+from pydantic.v1 import BaseModel as BaseModelV1
 from typing_extensions import Annotated, NotRequired, ReadOnly, Required, get_origin
 
+# NOTE: this is redefined here separately from langgraph.constants
+# to avoid a circular import
+MISSING = object()
+
 
 def _is_optional_type(type_: Any) -> bool:
     """Check if a type is Optional."""
@@ -147,3 +153,33 @@ def get_enhanced_type_hints(
             pass
 
         yield name, typ, default, description
+
+
+def get_update_as_tuples(input: Any, keys: Sequence[str]) -> list[tuple[str, Any]]:
+    """Get Pydantic state update as a list of (key, value) tuples."""
+    # Pydantic v1
+    if isinstance(input, BaseModelV1):
+        keep: Optional[set[str]] = input.__fields_set__
+        defaults = {k: v.default for k, v in input.__fields__.items()}
+    # Pydantic v2
+    elif isinstance(input, BaseModel):
+        keep = input.model_fields_set
+        defaults = {k: v.default for k, v in input.model_fields.items()}
+    else:
+        keep = None
+        defaults = {}
+
+    # NOTE: This behavior for Pydantic is somewhat inelegant,
+    # but we keep around for backwards compatibility
+    # if input is a Pydantic model, only update values
+    # that are different from the default values or in the keep set
+    return [
+        (k, value)
+        for k in keys
+        if (value := getattr(input, k, MISSING)) is not MISSING
+        and (
+            value is not None
+            or defaults.get(k, MISSING) is not None
+            or (keep is not None and k in keep)
+        )
+    ]

libs/langgraph/langgraph/utils/runnable.py

@@ -36,6 +36,7 @@ from langchain_core.runnables.config import (
     var_child_runnable_config,
 )
 from langchain_core.runnables.utils import Input, Output
+from langchain_core.tracers.langchain import LangChainTracer
 from typing_extensions import TypeGuard
 
 from langgraph.constants import (
@@ -60,58 +61,34 @@ except ImportError:
 
 
 def _set_config_context(
-    config: RunnableConfig,
-) -> tuple[Token[Optional[RunnableConfig]], Optional[dict[str, Any]]]:
+    config: RunnableConfig, run: Any = None
+) -> Token[Optional[RunnableConfig]]:
     """Set the child Runnable config + tracing context.
 
     Args:
         config (RunnableConfig): The config to set.
     """
-    from langchain_core.tracers.langchain import LangChainTracer
-
     config_token = var_child_runnable_config.set(config)
-    current_context = None
-    if (
-        (callbacks := config.get("callbacks"))
-        and (
-            parent_run_id := getattr(callbacks, "parent_run_id", None)
-        )  # Is callback manager
-        and (
-            tracer := next(
-                (
-                    handler
-                    for handler in getattr(callbacks, "handlers", [])
-                    if isinstance(handler, LangChainTracer)
-                ),
-                None,
-            )
-        )
-        and (run := tracer.run_map.get(str(parent_run_id)))
-    ):
-        from langsmith.run_helpers import _set_tracing_context, get_tracing_context
+    if run is not None:
+        from langsmith.run_helpers import _set_tracing_context
 
-        current_context = get_tracing_context()
         _set_tracing_context({"parent": run})
-    return config_token, current_context
+    return config_token
 
 
-@contextmanager
-def set_config_context(config: RunnableConfig) -> Generator[Context, None, None]:
+def _unset_config_context(
+    token: Token[Optional[RunnableConfig]], run: Any = None
+) -> None:
     """Set the child Runnable config + tracing context.
 
     Args:
         config (RunnableConfig): The config to set.
     """
-    from langsmith.run_helpers import _set_tracing_context
+    var_child_runnable_config.reset(token)
+    if run is not None:
+        from langsmith.run_helpers import _set_tracing_context
 
-    ctx = copy_context()
-    config_token, _ = ctx.run(_set_config_context, config)
-    try:
-        yield ctx
-    finally:
-        ctx.run(var_child_runnable_config.reset, config_token)
-        ctx.run(
-            _set_tracing_context,
+        _set_tracing_context(
             {
                 "parent": None,
                 "project_name": None,
@@ -119,10 +96,27 @@ def set_config_context(config: RunnableConfig) -> Generator[Context, None, None]
                 "metadata": None,
                 "enabled": None,
                 "client": None,
-            },
+            }
         )
 
 
+@contextmanager
+def set_config_context(
+    config: RunnableConfig, run: Any = None
+) -> Generator[Context, None, None]:
+    """Set the child Runnable config + tracing context.
+
+    Args:
+        config (RunnableConfig): The config to set.
+    """
+    ctx = copy_context()
+    config_token = ctx.run(_set_config_context, config, run)
+    try:
+        yield ctx
+    finally:
+        ctx.run(_unset_config_context, config_token, run)
+
+
 # Before Python 3.11 native StrEnum is not available
 class StrEnum(str, enum.Enum):
     """A string enum."""
@@ -363,7 +357,15 @@ class RunnableCallable(Runnable):
             )
             try:
                 child_config = patch_config(config, callbacks=run_manager.get_child())
-                with set_config_context(child_config) as context:
+                # get the run
+                for h in run_manager.handlers:
+                    if isinstance(h, LangChainTracer):
+                        run = h.run_map.get(str(run_manager.run_id))
+                        break
+                else:
+                    run = None
+                # run in context
+                with set_config_context(child_config, run) as context:
                     ret = context.run(self.func, *args, **kwargs)
             except BaseException as e:
                 run_manager.on_chain_error(e)
@@ -371,9 +373,8 @@ class RunnableCallable(Runnable):
             else:
                 run_manager.on_chain_end(ret)
         else:
-            with set_config_context(config) as context:
-                ret = context.run(self.func, *args, **kwargs)
-        if isinstance(ret, Runnable) and self.recurse:
+            ret = self.func(*args, **kwargs)
+        if self.recurse and isinstance(ret, Runnable):
             return ret.invoke(input, config)
         return ret
 
@@ -417,25 +418,26 @@ class RunnableCallable(Runnable):
             )
             try:
                 child_config = patch_config(config, callbacks=run_manager.get_child())
-                with set_config_context(child_config) as context:
-                    coro = cast(Coroutine[None, None, Any], self.afunc(*args, **kwargs))
-                    if ASYNCIO_ACCEPTS_CONTEXT:
-                        ret = await asyncio.create_task(coro, context=context)
+                coro = cast(Coroutine[None, None, Any], self.afunc(*args, **kwargs))
+                if ASYNCIO_ACCEPTS_CONTEXT:
+                    for h in run_manager.handlers:
+                        if isinstance(h, LangChainTracer):
+                            run = h.run_map.get(str(run_manager.run_id))
+                            break
                     else:
-                        ret = await coro
+                        run = None
+                    with set_config_context(child_config, run) as context:
+                        ret = await asyncio.create_task(coro, context=context)
+                else:
+                    ret = await coro
             except BaseException as e:
                 await run_manager.on_chain_error(e)
                 raise
             else:
                 await run_manager.on_chain_end(ret)
         else:
-            with set_config_context(config) as context:
-                if ASYNCIO_ACCEPTS_CONTEXT:
-                    coro = cast(Coroutine[None, None, Any], self.afunc(*args, **kwargs))
-                    ret = await asyncio.create_task(coro, context=context)
-                else:
-                    ret = await self.afunc(*args, **kwargs)
-        if isinstance(ret, Runnable) and self.recurse:
+            ret = await self.afunc(*args, **kwargs)
+        if self.recurse and isinstance(ret, Runnable):
             return await ret.ainvoke(input, config)
         return ret
 
@@ -598,7 +600,6 @@ class RunnableSeq(Runnable):
             name=config.get("run_name") or self.get_name(),
             run_id=config.pop("run_id", None),
         )
-
         # invoke all steps in sequence
         try:
             for i, step in enumerate(self.steps):
@@ -606,8 +607,19 @@ class RunnableSeq(Runnable):
                 config = patch_config(
                     config, callbacks=run_manager.get_child(f"seq:step:{i + 1}")
                 )
+                # 1st step is the actual node,
+                # others are writers which don't need to be run in context
                 if i == 0:
-                    input = step.invoke(input, config, **kwargs)
+                    # get the run object
+                    for h in run_manager.handlers:
+                        if isinstance(h, LangChainTracer):
+                            run = h.run_map.get(str(run_manager.run_id))
+                            break
+                    else:
+                        run = None
+                    # run in context
+                    with set_config_context(config, run) as context:
+                        input = context.run(step.invoke, input, config, **kwargs)
                 else:
                     input = step.invoke(input, config)
         # finish the root run
@@ -643,8 +655,24 @@ class RunnableSeq(Runnable):
                 config = patch_config(
                     config, callbacks=run_manager.get_child(f"seq:step:{i + 1}")
                 )
+                # 1st step is the actual node,
+                # others are writers which don't need to be run in context
                 if i == 0:
-                    input = await step.ainvoke(input, config, **kwargs)
+                    if ASYNCIO_ACCEPTS_CONTEXT:
+                        # get the run object
+                        for h in run_manager.handlers:
+                            if isinstance(h, LangChainTracer):
+                                run = h.run_map.get(str(run_manager.run_id))
+                                break
+                        else:
+                            run = None
+                        # run in context
+                        with set_config_context(config, run) as context:
+                            input = await asyncio.create_task(
+                                step.ainvoke(input, config, **kwargs), context=context
+                            )
+                    else:
+                        input = await step.ainvoke(input, config, **kwargs)
                 else:
                     input = await step.ainvoke(input, config)
         # finish the root run
@@ -672,53 +700,48 @@ class RunnableSeq(Runnable):
             name=config.get("run_name") or self.get_name(),
             run_id=config.pop("run_id", None),
         )
-
-        try:
-            # stream the last steps
-            # transform the input stream of each step with the next
-            # steps that don't natively support transforming an input stream will
-            # buffer input in memory until all available, and then start emitting output
-            for idx, step in enumerate(self.steps):
-                config = patch_config(
-                    config,
-                    callbacks=run_manager.get_child(f"seq:step:{idx + 1}"),
-                )
-                if idx == 0:
-                    iterator = step.stream(input, config, **kwargs)
-                else:
-                    iterator = step.transform(iterator, config)
-            if _StreamingCallbackHandler is not None and (
-                stream_handler := next(
-                    (
-                        cast(_StreamingCallbackHandler, h)
-                        for h in run_manager.handlers
-                        if isinstance(h, _StreamingCallbackHandler)
-                    ),
-                    None,
-                )
-            ):
-                # populates streamed_output in astream_log() output if needed
-                iterator = stream_handler.tap_output_iter(run_manager.run_id, iterator)
-            output: Any = None
-            add_supported = False
-            for chunk in iterator:
-                yield chunk
-                # collect final output
-                if output is None:
-                    output = chunk
-                elif add_supported:
-                    try:
-                        output = output + chunk
-                    except TypeError:
-                        output = chunk
-                        add_supported = False
-                else:
-                    output = chunk
-        except BaseException as e:
-            run_manager.on_chain_error(e)
-            raise
+        # get the run object
+        for h in run_manager.handlers:
+            if isinstance(h, LangChainTracer):
+                run = h.run_map.get(str(run_manager.run_id))
+                break
         else:
-            run_manager.on_chain_end(output)
+            run = None
+        # create first step config
+        config = patch_config(
+            config,
+            callbacks=run_manager.get_child(f"seq:step:{1}"),
+        )
+        # run all in context
+        with set_config_context(config, run) as context:
+            try:
+                # stream the last steps
+                # transform the input stream of each step with the next
+                # steps that don't natively support transforming an input stream will
+                # buffer input in memory until all available, and then start emitting output
+                for idx, step in enumerate(self.steps):
+                    if idx == 0:
+                        iterator = step.stream(input, config, **kwargs)
+                    else:
+                        config = patch_config(
+                            config,
+                            callbacks=run_manager.get_child(f"seq:step:{idx + 1}"),
+                        )
+                        iterator = step.transform(iterator, config)
+                # populates streamed_output in astream_log() output if needed
+                if _StreamingCallbackHandler is not None:
+                    for h in run_manager.handlers:
+                        if isinstance(h, _StreamingCallbackHandler):
+                            iterator = h.tap_output_iter(run_manager.run_id, iterator)
+                # consume into final output
+                output = context.run(_consume_iter, iterator)
+                # sequence doesn't emit output, yield to mark as generator
+                yield
+            except BaseException as e:
+                run_manager.on_chain_error(e)
+                raise
+            else:
+                run_manager.on_chain_end(output)
 
     async def astream(
         self,
@@ -737,53 +760,121 @@ class RunnableSeq(Runnable):
             name=config.get("run_name") or self.get_name(),
             run_id=config.pop("run_id", None),
         )
-
-        try:
-            async with AsyncExitStack() as stack:
-                # stream the last steps
-                # transform the input stream of each step with the next
-                # steps that don't natively support transforming an input stream will
-                # buffer input in memory until all available, and then start emitting output
-                for idx, step in enumerate(self.steps):
-                    config = patch_config(
-                        config,
-                        callbacks=run_manager.get_child(f"seq:step:{idx + 1}"),
-                    )
-                    if idx == 0:
-                        aiterator = step.astream(input, config, **kwargs)
-                    else:
-                        aiterator = step.atransform(aiterator, config)
-                    if hasattr(aiterator, "aclose"):
-                        stack.push_async_callback(aiterator.aclose)
-                if _StreamingCallbackHandler is not None and (
-                    stream_handler := next(
-                        (
-                            cast(_StreamingCallbackHandler, h)
-                            for h in run_manager.handlers
-                            if isinstance(h, _StreamingCallbackHandler)
-                        ),
-                        None,
-                    )
-                ):
-                    # populates streamed_output in astream_log() output if needed
-                    aiterator = stream_handler.tap_output_aiter(
-                        run_manager.run_id, aiterator
-                    )
-                output: Any = None
-                add_supported = False
-                async for chunk in aiterator:
-                    yield chunk
-                    # collect final output
-                    if add_supported:
-                        try:
-                            output = output + chunk
-                        except TypeError:
-                            output = chunk
-                            add_supported = False
-                    else:
-                        output = chunk
-        except BaseException as e:
-            await run_manager.on_chain_error(e)
-            raise
+        # stream the last steps
+        # transform the input stream of each step with the next
+        # steps that don't natively support transforming an input stream will
+        # buffer input in memory until all available, and then start emitting output
+        if ASYNCIO_ACCEPTS_CONTEXT:
+            # get the run object
+            for h in run_manager.handlers:
+                if isinstance(h, LangChainTracer):
+                    run = h.run_map.get(str(run_manager.run_id))
+                    break
+            else:
+                run = None
+            # create first step config
+            config = patch_config(
+                config,
+                callbacks=run_manager.get_child(f"seq:step:{1}"),
+            )
+            # run all in context
+            with set_config_context(config, run) as context:
+                try:
+                    async with AsyncExitStack() as stack:
+                        for idx, step in enumerate(self.steps):
+                            if idx == 0:
+                                aiterator = step.astream(input, config, **kwargs)
+                            else:
+                                config = patch_config(
+                                    config,
+                                    callbacks=run_manager.get_child(
+                                        f"seq:step:{idx + 1}"
+                                    ),
+                                )
+                                aiterator = step.atransform(aiterator, config)
+                            if hasattr(aiterator, "aclose"):
+                                stack.push_async_callback(aiterator.aclose)
+                        # populates streamed_output in astream_log() output if needed
+                        if _StreamingCallbackHandler is not None:
+                            for h in run_manager.handlers:
+                                if isinstance(h, _StreamingCallbackHandler):
+                                    aiterator = h.tap_output_aiter(
+                                        run_manager.run_id, aiterator
+                                    )
+                        # consume into final output
+                        output = await asyncio.create_task(
+                            _consume_aiter(aiterator), context=context
+                        )
+                        # sequence doesn't emit output, yield to mark as generator
+                        yield
+                except BaseException as e:
+                    await run_manager.on_chain_error(e)
+                    raise
+                else:
+                    await run_manager.on_chain_end(output)
         else:
-            await run_manager.on_chain_end(output)
+            try:
+                async with AsyncExitStack() as stack:
+                    for idx, step in enumerate(self.steps):
+                        config = patch_config(
+                            config,
+                            callbacks=run_manager.get_child(f"seq:step:{idx + 1}"),
+                        )
+                        if idx == 0:
+                            aiterator = step.astream(input, config, **kwargs)
+                        else:
+                            aiterator = step.atransform(aiterator, config)
+                        if hasattr(aiterator, "aclose"):
+                            stack.push_async_callback(aiterator.aclose)
+                    # populates streamed_output in astream_log() output if needed
+                    if _StreamingCallbackHandler is not None:
+                        for h in run_manager.handlers:
+                            if isinstance(h, _StreamingCallbackHandler):
+                                aiterator = h.tap_output_aiter(
+                                    run_manager.run_id, aiterator
+                                )
+                    # consume into final output
+                    output = await _consume_aiter(aiterator)
+                    # sequence doesn't emit output, yield to mark as generator
+                    yield
+            except BaseException as e:
+                await run_manager.on_chain_error(e)
+                raise
+            else:
+                await run_manager.on_chain_end(output)
+
+
+def _consume_iter(it: Iterator[Any]) -> Any:
+    """Consume an iterator."""
+    output: Any = None
+    add_supported = False
+    for chunk in it:
+        # collect final output
+        if output is None:
+            output = chunk
+        elif add_supported:
+            try:
+                output = output + chunk
+            except TypeError:
+                output = chunk
+                add_supported = False
+        else:
+            output = chunk
+    return output
+
+
+async def _consume_aiter(it: AsyncIterator[Any]) -> Any:
+    """Consume an async iterator."""
+    output: Any = None
+    add_supported = False
+    async for chunk in it:
+        # collect final output
+        if add_supported:
+            try:
+                output = output + chunk
+            except TypeError:
+                output = chunk
+                add_supported = False
+        else:
+            output = chunk
+    return output

libs/langgraph/poetry.lock

@@ -1,4 +1,4 @@
-# This file is automatically @generated by Poetry 2.0.0 and should not be changed by hand.
+# This file is automatically @generated by Poetry 2.0.1 and should not be changed by hand.
 
 [[package]]
 name = "aiosqlite"
@@ -1348,7 +1348,7 @@ typing-extensions = ">=4.7"
 
 [[package]]
 name = "langgraph-checkpoint"
-version = "2.0.21"
+version = "2.0.24"
 description = "Library with base interfaces for LangGraph checkpoint savers."
 optional = false
 python-versions = "^3.9.0,<4.0"
@@ -1404,7 +1404,7 @@ url = "../checkpoint-sqlite"
 
 [[package]]
 name = "langgraph-prebuilt"
-version = "0.1.4"
+version = "0.1.8"
 description = "Library with high-level APIs for creating and executing LangGraph agents and tools."
 optional = false
 python-versions = "^3.9.0,<4.0"
@@ -1422,7 +1422,7 @@ url = "../prebuilt"
 
 [[package]]
 name = "langgraph-sdk"
-version = "0.1.58"
+version = "0.1.61"
 description = "SDK for interacting with LangGraph API"
 optional = false
 python-versions = "^3.9.0,<4.0"
@@ -3646,4 +3646,4 @@ type = ["pytest-mypy"]
 [metadata]
 lock-version = "2.1"
 python-versions = ">=3.9.0,<4.0"
-content-hash = "b03760d1062e13e4df0b4052a194bedb8abb3baf80da0c036b39d2ebe26b0b5c"
+content-hash = "1a6454eb63ce88ddfddd0d49530a6fef0f02ac6b0b52be5d6710af8e47b1de24"

libs/langgraph/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langgraph"
-version = "0.3.27"
+version = "0.3.31"
 description = "Building stateful, multi-actor applications with LLMs"
 authors = []
 license = "MIT"
@@ -12,7 +12,7 @@ python = ">=3.9.0,<4.0"
 langchain-core = ">=0.1,<0.4"
 langgraph-checkpoint = "^2.0.10"
 langgraph-sdk = "^0.1.42"
-langgraph-prebuilt = ">=0.1.1,<0.2"
+langgraph-prebuilt = ">=0.1.8,<0.2"
 xxhash = "^3.5.0"
 
 [tool.poetry.group.dev.dependencies]

libs/langgraph/tests/test_large_cases.py

@@ -4660,7 +4660,7 @@ def test_root_graph(
                     content="result for query",
                     name="search_api",
                     tool_call_id="tool_call123",
-                    id="00000000-0000-4000-8000-000000000040",
+                    id="00000000-0000-4000-8000-000000000024",
                 )
             ]
         },
@@ -4683,7 +4683,7 @@ def test_root_graph(
                     content="result for another",
                     name="search_api",
                     tool_call_id="tool_call456",
-                    id="00000000-0000-4000-8000-000000000049",
+                    id="00000000-0000-4000-8000-000000000030",
                 )
             ]
         },
@@ -5387,7 +5387,7 @@ def test_root_graph(
         "__root__": [
             HumanMessage(
                 content="what is weather in sf",
-                id="00000000-0000-4000-8000-000000000083",
+                id="00000000-0000-4000-8000-000000000051",
             ),
             AIMessage(
                 content="",
@@ -5407,7 +5407,7 @@ def test_root_graph(
             ),
             AIMessage(content="answer", id="ai2"),
             AIMessage(
-                content="an extra message", id="00000000-0000-4000-8000-000000000107"
+                content="an extra message", id="00000000-0000-4000-8000-000000000066"
             ),
             HumanMessage(content="what is weather in la"),
         ],

libs/langgraph/tests/test_pregel.py

@@ -4601,7 +4601,7 @@ def test_checkpoint_metadata() -> None:
         assert chkpnt_tuple.metadata["test_config_4"] == "bar"
 
 
-@pytest.mark.parametrize("checkpointer_name", ALL_CHECKPOINTERS_SYNC)
+@pytest.mark.parametrize("checkpointer_name", REGULAR_CHECKPOINTERS_SYNC)
 def test_remove_message_via_state_update(
     request: pytest.FixtureRequest, checkpointer_name: str
 ) -> None:
@@ -4631,6 +4631,12 @@ def test_remove_message_via_state_update(
     assert len(updated_state.values) == 1
     assert updated_state.values[-1].content == "Hi"
 
+    app.checkpointer.delete_thread(config["configurable"]["thread_id"])
+
+    # Verify that the message was removed from the checkpointer
+    assert app.checkpointer.get_tuple(config) is None
+    assert [*app.get_state_history(config)] == []
+
 
 def test_remove_message_from_node():
     from langchain_core.messages import AIMessage, HumanMessage, RemoveMessage
@@ -6034,7 +6040,7 @@ def test_concurrent_execution_thread_safety():
         assert result["counter"] == 1
 
 
-@pytest.mark.parametrize("checkpointer_name", ALL_CHECKPOINTERS_SYNC)
+@pytest.mark.parametrize("checkpointer_name", REGULAR_CHECKPOINTERS_SYNC)
 def test_checkpoint_recovery(request: pytest.FixtureRequest, checkpointer_name: str):
     """Test recovery from checkpoints after failures."""
     checkpointer = request.getfixturevalue(f"checkpointer_{checkpointer_name}")
@@ -6087,6 +6093,11 @@ def test_checkpoint_recovery(request: pytest.FixtureRequest, checkpointer_name:
     failed_checkpoint = next(c for c in history if c.tasks and c.tasks[0].error)
     assert "RuntimeError('Simulated failure')" in failed_checkpoint.tasks[0].error
 
+    # Verify delete leaves it empty
+    graph.checkpointer.delete_thread(config["configurable"]["thread_id"])
+    assert graph.checkpointer.get_tuple(config) is None
+    assert [*graph.get_state_history(config)] == []
+
 
 def test_multiple_updates_root() -> None:
     def node_a(state):
@@ -6156,7 +6167,7 @@ def test_falsy_return_from_task(
         falsy_task().result()
         interrupt("test")
 
-    configurable = {"configurable": {"thread_id": str(uuid.uuid4())}}
+    configurable = {"configurable": {"thread_id": uuid.uuid4()}}
     assert [
         chunk for chunk in graph.stream({"a": 5}, configurable, stream_mode="debug")
     ] == [
@@ -6169,7 +6180,7 @@ def test_falsy_return_from_task(
                         "checkpoint_ns": "",
                         "thread_id": AnyStr(),
                     },
-                    "metadata": configurable["configurable"],
+                    "metadata": {},
                     "recursion_limit": 25,
                     "tags": [],
                 },
@@ -6177,7 +6188,6 @@ def test_falsy_return_from_task(
                     "parents": {},
                     "source": "input",
                     "step": -1,
-                    "thread_id": AnyStr(),
                     "writes": {
                         "__start__": {
                             "a": 5,
@@ -6268,6 +6278,7 @@ def test_falsy_return_from_task(
             "type": "task_result",
         },
     ]
+    print(type(configurable["configurable"]["thread_id"]))
     assert [
         c
         for c in graph.stream(Command(resume="123"), configurable, stream_mode="debug")
@@ -6281,7 +6292,7 @@ def test_falsy_return_from_task(
                         "checkpoint_ns": "",
                         "thread_id": AnyStr(),
                     },
-                    "metadata": configurable["configurable"],
+                    "metadata": {},
                     "recursion_limit": 25,
                     "tags": [],
                 },
@@ -6376,7 +6387,7 @@ def test_falsy_return_from_task(
                         "checkpoint_ns": "",
                         "thread_id": AnyStr(),
                     },
-                    "metadata": configurable["configurable"],
+                    "metadata": {},
                     "recursion_limit": 25,
                     "tags": [],
                 },
@@ -6384,7 +6395,6 @@ def test_falsy_return_from_task(
                     "parents": {},
                     "source": "loop",
                     "step": 0,
-                    "thread_id": AnyStr(),
                     "writes": {
                         "falsy_task": False,
                         "graph": None,
@@ -6398,7 +6408,7 @@ def test_falsy_return_from_task(
                         "checkpoint_ns": "",
                         "thread_id": AnyStr(),
                     },
-                    "metadata": configurable["configurable"],
+                    "metadata": {},
                     "recursion_limit": 25,
                     "tags": [],
                 },
@@ -7168,6 +7178,52 @@ def test_tags_stream_mode_messages() -> None:
     ]
 
 
+def test_stream_mode_messages_command() -> None:
+    from langchain_core.messages import HumanMessage
+
+    def my_node(state):
+        return {"messages": HumanMessage(content="foo")}
+
+    def my_other_node(state):
+        return Command(update={"messages": HumanMessage(content="bar")})
+
+    graph = (
+        StateGraph(MessagesState)
+        .add_sequence([my_node, my_other_node])
+        .add_edge(START, "my_node")
+        .compile()
+    )
+    assert list(
+        graph.stream(
+            {
+                "messages": [],
+            },
+            stream_mode="messages",
+        )
+    ) == [
+        (
+            _AnyIdHumanMessage(content="foo"),
+            {
+                "langgraph_step": 1,
+                "langgraph_node": "my_node",
+                "langgraph_triggers": ("branch:to:my_node",),
+                "langgraph_path": ("__pregel_pull", "my_node"),
+                "langgraph_checkpoint_ns": AnyStr("my_node:"),
+            },
+        ),
+        (
+            _AnyIdHumanMessage(content="bar"),
+            {
+                "langgraph_step": 2,
+                "langgraph_node": "my_other_node",
+                "langgraph_triggers": ("branch:to:my_other_node",),
+                "langgraph_path": ("__pregel_pull", "my_other_node"),
+                "langgraph_checkpoint_ns": AnyStr("my_other_node:"),
+            },
+        ),
+    ]
+
+
 def test_node_destinations() -> None:
     class State(TypedDict):
         foo: Annotated[str, operator.add]
@@ -7246,6 +7302,39 @@ def test_pydantic_none_state_update() -> None:
     assert graph.invoke({"foo": ""}) == {"foo": None}
 
 
+def test_pydantic_state_update_command() -> None:
+    from pydantic import BaseModel
+
+    class State(BaseModel):
+        foo: Optional[str]
+
+    def node_a(state: State) -> State:
+        return Command(update=State(foo=None))
+
+    graph = StateGraph(State).add_node(node_a).add_edge(START, "node_a").compile()
+    assert graph.invoke({"foo": ""}) == {"foo": None}
+
+    class State(BaseModel):
+        foo: Optional[str] = None
+        bar: Optional[str] = None
+
+    def node_a(state: State):
+        return State(foo="foo")
+
+    def node_b(state: State):
+        return Command(update=State(bar="bar"))
+
+    builder = StateGraph(State)
+    builder.add_node(node_a)
+    builder.add_node(node_b)
+    builder.add_edge(START, "node_a")
+    builder.add_edge("node_a", "node_b")
+    builder.add_edge("node_b", END)
+    graph = builder.compile()
+
+    assert graph.invoke(State()) == {"foo": "foo", "bar": "bar"}
+
+
 def test_pydantic_state_mutation() -> None:
     from pydantic import BaseModel, Field
 
@@ -7280,6 +7369,40 @@ def test_pydantic_state_mutation() -> None:
     assert graph.invoke({"outer": 1}) == {"outer": 10, "inner": Inner(a=5)}
 
 
+def test_pydantic_state_mutation_command() -> None:
+    from pydantic import BaseModel, Field
+
+    class Inner(BaseModel):
+        a: int = 0
+
+    class State(BaseModel):
+        inner: Inner = Inner()
+        outer: int = 0
+
+    def my_node(state: State) -> State:
+        state.inner.a = 5
+        state.outer = 10
+        return Command(update=state)
+
+    graph = StateGraph(State).add_node(my_node).add_edge(START, "my_node").compile()
+
+    assert graph.invoke({"outer": 1}) == {"outer": 10, "inner": Inner(a=5)}
+
+    # test w/ default_factory
+    class State(BaseModel):
+        inner: Inner = Field(default_factory=Inner)
+        outer: int = 0
+
+    def my_node(state: State) -> State:
+        state.inner.a = 5
+        state.outer = 10
+        return Command(update=state)
+
+    graph = StateGraph(State).add_node(my_node).add_edge(START, "my_node").compile()
+
+    assert graph.invoke({"outer": 1}) == {"outer": 10, "inner": Inner(a=5)}
+
+
 def test_get_stream_writer() -> None:
     class State(TypedDict):
         foo: str

libs/langgraph/tests/test_pregel_async.py

@@ -7894,6 +7894,53 @@ async def test_tags_stream_mode_messages() -> None:
     ]
 
 
+async def test_stream_mode_messages_command() -> None:
+    from langchain_core.messages import HumanMessage
+
+    async def my_node(state):
+        return {"messages": HumanMessage(content="foo")}
+
+    async def my_other_node(state):
+        return Command(update={"messages": HumanMessage(content="bar")})
+
+    graph = (
+        StateGraph(MessagesState)
+        .add_sequence([my_node, my_other_node])
+        .add_edge(START, "my_node")
+        .compile()
+    )
+    assert [
+        c
+        async for c in graph.astream(
+            {
+                "messages": [],
+            },
+            stream_mode="messages",
+        )
+    ] == [
+        (
+            _AnyIdHumanMessage(content="foo"),
+            {
+                "langgraph_step": 1,
+                "langgraph_node": "my_node",
+                "langgraph_triggers": ("branch:to:my_node",),
+                "langgraph_path": ("__pregel_pull", "my_node"),
+                "langgraph_checkpoint_ns": AnyStr("my_node:"),
+            },
+        ),
+        (
+            _AnyIdHumanMessage(content="bar"),
+            {
+                "langgraph_step": 2,
+                "langgraph_node": "my_other_node",
+                "langgraph_triggers": ("branch:to:my_other_node",),
+                "langgraph_path": ("__pregel_pull", "my_other_node"),
+                "langgraph_checkpoint_ns": AnyStr("my_other_node:"),
+            },
+        ),
+    ]
+
+
 async def test_stream_messages_dedupe_inputs() -> None:
     from langchain_core.messages import AIMessage
 

libs/langgraph/tests/test_retry.py

@@ -0,0 +1,342 @@
+from unittest.mock import Mock, patch
+
+import pytest
+from typing_extensions import TypedDict
+
+from langgraph.graph import START, StateGraph
+from langgraph.pregel.retry import _should_retry_on
+from langgraph.types import RetryPolicy
+
+
+def test_should_retry_on_single_exception():
+    """Test retry with a single exception type."""
+    policy = RetryPolicy(retry_on=ValueError)
+
+    # Should retry on ValueError
+    assert _should_retry_on(policy, ValueError("test error")) is True
+
+    # Should not retry on other exceptions
+    assert _should_retry_on(policy, TypeError("test error")) is False
+    assert _should_retry_on(policy, Exception("test error")) is False
+
+
+def test_should_retry_on_sequence_of_exceptions():
+    """Test retry with a sequence of exception types."""
+    policy = RetryPolicy(retry_on=(ValueError, KeyError))
+
+    # Should retry on listed exceptions
+    assert _should_retry_on(policy, ValueError("test error")) is True
+    assert _should_retry_on(policy, KeyError("test error")) is True
+
+    # Should not retry on other exceptions
+    assert _should_retry_on(policy, TypeError("test error")) is False
+    assert _should_retry_on(policy, Exception("test error")) is False
+
+
+def test_should_retry_on_subclass_of_exception():
+    """Test retry on subclass of specified exception."""
+
+    class CustomError(ValueError):
+        pass
+
+    policy = RetryPolicy(retry_on=ValueError)
+
+    # Should retry on subclass of specified exception
+    assert _should_retry_on(policy, CustomError("test error")) is True
+
+
+def test_should_retry_on_callable():
+    """Test retry with a callable predicate."""
+
+    # Only retry on ValueError with message containing 'retry'
+    def should_retry(exc: Exception) -> bool:
+        return isinstance(exc, ValueError) and "retry" in str(exc)
+
+    policy = RetryPolicy(retry_on=should_retry)
+
+    # Should retry when predicate returns True
+    assert _should_retry_on(policy, ValueError("please retry this")) is True
+
+    # Should not retry when predicate returns False
+    assert _should_retry_on(policy, ValueError("other error")) is False
+    assert _should_retry_on(policy, TypeError("please retry this")) is False
+
+
+def test_should_retry_on_invalid_type():
+    """Test retry with an invalid retry_on type."""
+    policy = RetryPolicy(retry_on=123)  # type: ignore
+
+    with pytest.raises(TypeError, match="retry_on must be an Exception class"):
+        _should_retry_on(policy, ValueError("test error"))
+
+
+def test_should_retry_on_empty_sequence():
+    """Test retry with an empty sequence."""
+    policy = RetryPolicy(retry_on=())
+
+    # Should not retry when sequence is empty
+    assert _should_retry_on(policy, ValueError("test error")) is False
+
+
+def test_should_retry_default_retry_on():
+    """Test the default retry_on function."""
+    import httpx
+    import requests
+
+    # Create a RetryPolicy with default_retry_on
+    policy = RetryPolicy()
+
+    # Should retry on ConnectionError
+    assert _should_retry_on(policy, ConnectionError("connection refused")) is True
+
+    # Should not retry on common programming errors
+    assert _should_retry_on(policy, ValueError("invalid value")) is False
+    assert _should_retry_on(policy, TypeError("invalid type")) is False
+    assert _should_retry_on(policy, ArithmeticError("division by zero")) is False
+    assert _should_retry_on(policy, ImportError("module not found")) is False
+    assert _should_retry_on(policy, LookupError("key not found")) is False
+    assert _should_retry_on(policy, NameError("name not defined")) is False
+    assert _should_retry_on(policy, SyntaxError("invalid syntax")) is False
+    assert _should_retry_on(policy, RuntimeError("runtime error")) is False
+    assert _should_retry_on(policy, ReferenceError("weak reference")) is False
+    assert _should_retry_on(policy, StopIteration()) is False
+    assert _should_retry_on(policy, StopAsyncIteration()) is False
+    assert _should_retry_on(policy, OSError("file not found")) is False
+
+    # Should retry on httpx.HTTPStatusError with 5xx status code
+    response_5xx = Mock()
+    response_5xx.status_code = 503
+    http_error_5xx = httpx.HTTPStatusError(
+        "server error", request=Mock(), response=response_5xx
+    )
+    assert _should_retry_on(policy, http_error_5xx) is True
+
+    # Should not retry on httpx.HTTPStatusError with 4xx status code
+    response_4xx = Mock()
+    response_4xx.status_code = 404
+    http_error_4xx = httpx.HTTPStatusError(
+        "not found", request=Mock(), response=response_4xx
+    )
+    assert _should_retry_on(policy, http_error_4xx) is False
+
+    # Should retry on requests.HTTPError with 5xx status code
+    response_req_5xx = Mock()
+    response_req_5xx.status_code = 502
+    req_error_5xx = requests.HTTPError("bad gateway")
+    req_error_5xx.response = response_req_5xx
+    assert _should_retry_on(policy, req_error_5xx) is True
+
+    # Should not retry on requests.HTTPError with 4xx status code
+    response_req_4xx = Mock()
+    response_req_4xx.status_code = 400
+    req_error_4xx = requests.HTTPError("bad request")
+    req_error_4xx.response = response_req_4xx
+    assert _should_retry_on(policy, req_error_4xx) is False
+
+    # Should retry on requests.HTTPError with no response
+    req_error_no_resp = requests.HTTPError("connection error")
+    req_error_no_resp.response = None
+    assert _should_retry_on(policy, req_error_no_resp) is True
+
+    # Should retry on other exceptions by default
+    class CustomException(Exception):
+        pass
+
+    assert _should_retry_on(policy, CustomException("custom error")) is True
+
+
+def test_graph_with_single_retry_policy():
+    """Test a simple graph with a single RetryPolicy for a node."""
+
+    class State(TypedDict):
+        foo: str
+
+    attempt_count = 0
+
+    def failing_node(state: State):
+        nonlocal attempt_count
+        attempt_count += 1
+        if attempt_count < 3:  # Fail the first two attempts
+            raise ValueError("Intentional failure")
+        return {"foo": "success"}
+
+    def other_node(state: State):
+        return {"foo": "other_node"}
+
+    # Create a retry policy with specific parameters
+    retry_policy = RetryPolicy(
+        max_attempts=3,
+        initial_interval=0.01,  # Short interval for tests
+        backoff_factor=2.0,
+        jitter=False,  # Disable jitter for predictable timing
+        retry_on=ValueError,
+    )
+
+    # Create and compile the graph
+    graph = (
+        StateGraph(State)
+        .add_node("failing_node", failing_node, retry=retry_policy)
+        .add_node("other_node", other_node)
+        .add_edge(START, "failing_node")
+        .add_edge("failing_node", "other_node")
+        .compile()
+    )
+
+    with patch("time.sleep") as mock_sleep:
+        result = graph.invoke({"foo": ""})
+
+    # Verify retry behavior
+    assert attempt_count == 3  # The node should have been tried 3 times
+    assert result["foo"] == "other_node"  # Final result should be from other_node
+
+    # Verify the sleep intervals
+    call_args_list = [args[0][0] for args in mock_sleep.call_args_list]
+    assert call_args_list == [0.01, 0.02]
+
+
+def test_graph_with_jitter_retry_policy():
+    """Test a graph with a RetryPolicy that uses jitter."""
+
+    class State(TypedDict):
+        foo: str
+
+    attempt_count = 0
+
+    def failing_node(state):
+        nonlocal attempt_count
+        attempt_count += 1
+        if attempt_count < 2:  # Fail the first attempt
+            raise ValueError("Intentional failure")
+        return {"foo": "success"}
+
+    # Create a retry policy with jitter enabled
+    retry_policy = RetryPolicy(
+        max_attempts=3,
+        initial_interval=0.01,
+        jitter=True,  # Enable jitter for randomized backoff
+        retry_on=ValueError,
+    )
+
+    # Create and compile the graph
+    graph = (
+        StateGraph(State)
+        .add_node("failing_node", failing_node, retry=retry_policy)
+        .add_edge(START, "failing_node")
+        .compile()
+    )
+
+    # Test graph execution with mocked random and sleep
+    with patch("random.uniform", return_value=0.05) as mock_random, patch(
+        "time.sleep"
+    ) as mock_sleep:
+        result = graph.invoke({"foo": ""})
+
+    # Verify retry behavior
+    assert attempt_count == 2  # The node should have been tried twice
+    assert result["foo"] == "success"
+
+    # Verify jitter was applied
+    mock_random.assert_called_with(0, 1)  # Jitter should use random.uniform(0, 1)
+    mock_sleep.assert_called_with(0.01 + 0.05)  # Sleep should include jitter
+
+
+def test_graph_with_multiple_retry_policies():
+    """Test a graph with multiple retry policies for a node."""
+
+    class State(TypedDict):
+        foo: str
+        error_type: str
+
+    attempt_counts = {"value_error": 0, "key_error": 0}
+
+    def failing_node(state):
+        error_type = state["error_type"]
+
+        if error_type == "value_error":
+            attempt_counts["value_error"] += 1
+            if attempt_counts["value_error"] < 2:
+                raise ValueError("Value error")
+        elif error_type == "key_error":
+            attempt_counts["key_error"] += 1
+            if attempt_counts["key_error"] < 3:
+                raise KeyError("Key error")
+
+        return {"foo": f"recovered_from_{error_type}"}
+
+    # Create multiple retry policies
+    value_error_policy = RetryPolicy(
+        max_attempts=2,
+        initial_interval=0.01,
+        jitter=False,
+        retry_on=ValueError,
+    )
+
+    key_error_policy = RetryPolicy(
+        max_attempts=3,
+        initial_interval=0.02,
+        jitter=False,
+        retry_on=KeyError,
+    )
+
+    # Create and compile the graph with a list of retry policies
+    graph = (
+        StateGraph(State)
+        .add_node(
+            "failing_node",
+            failing_node,
+            retry=(value_error_policy, key_error_policy),
+        )
+        .add_edge(START, "failing_node")
+        .compile()
+    )
+
+    # Test ValueError scenario
+    with patch("time.sleep"):
+        result_value_error = graph.invoke({"foo": "", "error_type": "value_error"})
+
+    assert attempt_counts["value_error"] == 2
+    assert result_value_error["foo"] == "recovered_from_value_error"
+
+    # Reset attempt counts
+    attempt_counts = {"value_error": 0, "key_error": 0}
+
+    # Test KeyError scenario
+    with patch("time.sleep"):
+        result_key_error = graph.invoke({"foo": "", "error_type": "key_error"})
+
+    assert attempt_counts["key_error"] == 3
+    assert result_key_error["foo"] == "recovered_from_key_error"
+
+
+def test_graph_with_max_attempts_exceeded():
+    """Test a graph where max_attempts is exceeded."""
+
+    class State(TypedDict):
+        foo: str
+
+    def always_failing_node(state):
+        raise ValueError("Always fails")
+
+    # Create a retry policy with limited attempts
+    retry_policy = RetryPolicy(
+        max_attempts=2,
+        initial_interval=0.01,
+        jitter=False,
+        retry_on=ValueError,
+    )
+
+    # Create and compile the graph
+    graph = (
+        StateGraph(State)
+        .add_node("always_failing", always_failing_node, retry=retry_policy)
+        .add_edge(START, "always_failing")
+        .compile()
+    )
+
+    # Test graph execution
+    with patch("time.sleep") as mock_sleep, pytest.raises(
+        ValueError, match="Always fails"
+    ):
+        graph.invoke({"foo": ""})
+
+    mock_sleep.assert_called_with(0.01)

libs/sdk-js/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@langchain/langgraph-sdk",
-  "version": "0.0.66",
+  "version": "0.0.67",
   "description": "Client library for interacting with the LangGraph API",
   "type": "module",
   "packageManager": "yarn@1.22.19",

libs/sdk-js/src/client.ts

@@ -84,18 +84,37 @@ class BaseClient {
   protected defaultHeaders: Record<string, string | null | undefined>;
 
   constructor(config?: ClientConfig) {
-    this.asyncCaller = new AsyncCaller({
+    const callerOptions = {
       maxRetries: 4,
       maxConcurrency: 4,
       ...config?.callerOptions,
-    });
+    };
 
+    let defaultApiUrl = "http://localhost:8123";
+    if (
+      !config?.apiUrl &&
+      typeof globalThis === "object" &&
+      globalThis != null
+    ) {
+      const fetchSmb = Symbol.for("langgraph_api:fetch");
+      const urlSmb = Symbol.for("langgraph_api:url");
+
+      const global = globalThis as unknown as {
+        [fetchSmb]?: typeof fetch;
+        [urlSmb]?: string;
+      };
+
+      if (global[fetchSmb]) callerOptions.fetch ??= global[fetchSmb];
+      if (global[urlSmb]) defaultApiUrl = global[urlSmb];
+    }
+
+    this.asyncCaller = new AsyncCaller(callerOptions);
     this.timeoutMs = config?.timeoutMs;
 
     // default limit being capped by Chrome
     // https://github.com/nodejs/undici/issues/1373
     // Regex to remove trailing slash, if present
-    this.apiUrl = config?.apiUrl?.replace(/\/$/, "") || "http://localhost:8123";
+    this.apiUrl = config?.apiUrl?.replace(/\/$/, "") || defaultApiUrl;
     this.defaultHeaders = config?.defaultHeaders || {};
     const apiKey = getApiKey(config?.apiKey);
     if (apiKey) {

libs/sdk-js/src/react/index.ts

@@ -1 +1,6 @@
-export { useStream, type MessageMetadata } from "./stream.js";
+export {
+  useStream,
+  type MessageMetadata,
+  type UseStream,
+  type UseStreamOptions,
+} from "./stream.js";

libs/sdk-js/src/react/stream.tsx

@@ -405,7 +405,7 @@ type GetCustomEventType<Bag extends BagTemplate> = Bag extends {
   ? Bag["CustomEventType"]
   : unknown;
 
-interface UseStreamOptions<
+export interface UseStreamOptions<
   StateType extends Record<string, unknown> = Record<string, unknown>,
   Bag extends BagTemplate = BagTemplate,
 > {

libs/sdk-py/langgraph_sdk/schema.py

@@ -125,7 +125,7 @@ class Checkpoint(TypedDict):
     thread_id: str
     """Unique identifier for the thread associated with this checkpoint."""
     checkpoint_ns: str
-    """Namespace for the checkpoint, used for organization and retrieval."""
+    """Namespace for the checkpoint; used internally to manage subgraph state."""
     checkpoint_id: Optional[str]
     """Optional unique identifier for the checkpoint itself."""
     checkpoint_map: Optional[dict[str, Any]]