langgraph-checkpoint-postgres 2.0.19 (#3945)

Leverage information about which channels were updated in the previous
step to determine which tasks should be triggered. This can result in
significant speed up in prepare_next_tasks in some situations.

.github/workflows/_lint.yml

@@ -54,7 +54,7 @@ jobs:
         if: steps.changed-files.outputs.all
         shell: bash
         working-directory: ${{ inputs.working-directory }}
-        run: poetry lock --check
+        run: poetry check --lock
 
       - name: Install dependencies
         if: steps.changed-files.outputs.all

.github/workflows/_test.yml

@@ -39,6 +39,12 @@ jobs:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_RO_TOKEN }}
 
+      - name: Check Lock
+        shell: bash
+        working-directory: ${{ inputs.working-directory }}
+        run: |
+          poetry check --lock
+
       - name: Install dependencies
         shell: bash
         working-directory: ${{ inputs.working-directory }}

.github/workflows/ci.yml

@@ -114,6 +114,42 @@ jobs:
       - name: Run check_sdk_methods script
         run: python .github/scripts/check_sdk_methods.py
 
+  check-schema:
+    needs: changes
+    if: needs.changes.outputs.python == 'true'
+    name: "Check CLI schema hasn't changed #${{ matrix.python-version }}"
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version:
+          - "3.11"
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set up Python ${{ matrix.python-version }} + Poetry ${{ env.POETRY_VERSION }}
+        uses: "./.github/actions/poetry_setup"
+        with:
+          python-version: "3.11"
+          poetry-version: ${{ env.POETRY_VERSION }}
+          cache-key: schema-check-cli
+      - name: Install CLI dependencies
+        run: |
+          cd libs/cli
+          poetry install
+      - name: Generate schema and check for changes
+        run: |
+          cd libs/cli
+          # Create a temporary copy of the current schema
+          cp schemas/schema.json schemas/schema.current.json
+          # Generate new schema
+          poetry run python generate_schema.py
+          # Compare the new schema with the original
+          if ! diff -q schemas/schema.json schemas/schema.current.json > /dev/null; then
+            echo "Error: Langgraph.json configuration schema has changed. Please run 'poetry run python generate_schema.py' in the libs/cli directory and commit the changes."
+            diff schemas/schema.json schemas/schema.current.json
+            exit 1
+          fi
+          echo "Schema check passed - no changes detected"
+
   integration-test:
     needs: changes
     if: needs.changes.outputs.python == 'true'
@@ -180,6 +216,8 @@ jobs:
         test,
         test-langgraph,
         test-scheduler-kafka,
+        check-sdk-methods,
+        check-schema,
         integration-test,
         test-js,
       ]

.github/workflows/deploy_docs.yml

@@ -102,7 +102,14 @@ jobs:
       - name: Build llms-text
         run: make llms-text
       - name: Build site
-        run: make build-docs
+        run: |
+          # If this is main branch, then we want to download stats. we do this 
+          # with the env variable DOWNLOAD_STATS=true
+          if [ "${{ github.ref }}" == "refs/heads/main" ]; then
+            DOWNLOAD_STATS=true make build-docs
+          else
+            make build-docs
+          fi
         env:
           MKDOCS_GIT_COMMITTERS_APIKEY: ${{ secrets.MKDOCS_GIT_COMMITTERS_APIKEY }}
           OPENAI_API_KEY: sf-proj-1234567890 # fake placeholder, shouldn't actually be used
@@ -127,6 +134,7 @@ jobs:
               --check-links-ignore "https://openai\.com/.*" \
               --check-links-ignore "https://www\.uber\.com/.*" \
               --check-links-ignore "https://pepy\.tech/.*" \
+              --check-links-ignore "docs/docs/static/wordmark_*" \
               --check-links $(find site -name "index.html" | grep -v 'storm/index.html')
               
           else
@@ -147,6 +155,7 @@ jobs:
                 --check-links-ignore "https://twitter.com/.*" \
                 --check-links-ignore "https://github\.com/.*" \
                 --check-links-ignore "/.*\.(ipynb|html)$" \
+                --check-links-ignore "docs/docs/static/wordmark_*" \
                 --check-links ${CHANGED_FILES} \
                 || ([ $? = 5 ] && exit 0 || exit $?)
             else

.github/workflows/run_notebooks.yml

@@ -57,13 +57,13 @@ jobs:
         env:
           # these won't actually be used because of the VCR cassettes
           # but need to set them to avoid triggering getpass()
-          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
-          TAVILY_API_KEY: ${{ secrets.TAVILY_API_KEY }}
-          LANGSMITH_API_KEY: ${{ secrets.LANGSMITH_API_KEY }}
-          NOMIC_API_KEY: ${{ secrets.NOMIC_API_KEY }}
-          COHERE_API_KEY: ${{ secrets.COHERE_API_KEY }}
-          FIREWORKS_API_KEY: ${{ secrets.FIREWORKS_API_KEY }}
+          OPENAI_API_KEY: "very-secret-key"
+          ANTHROPIC_API_KEY: "very-secret-key"
+          TAVILY_API_KEY: "very-secret-key"
+          LANGSMITH_API_KEY: "very-secret-key"
+          NOMIC_API_KEY: "very-secret-key"
+          COHERE_API_KEY: "very-secret-key"
+          FIREWORKS_API_KEY: "very-secret-key"
         run: |
           if [ "${{ github.event_name }}" = "workflow_dispatch" ] || [ "${{ github.event_name }}" = "schedule" ]; then
             echo "Running all notebooks"

.github/workflows/size.yml

@@ -1,29 +0,0 @@
-name: Check File Size
-
-on:
-  push:
-    branches:
-      - main
-  pull_request:
-    branches:
-      - main
-  workflow_dispatch:
-
-jobs:
-  file-size-check:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-      - name: Get changed files
-        id: changed-files
-        uses: tj-actions/changed-files@v44
-      - name: Filter by size
-        # TODO: roll back the web voyager hack
-        run: |
-          large_added_files=$(find ${{ steps.changed-files.outputs.added_files }} -maxdepth 0 -size +1M | grep -v "web_voyager" || true)
-          if [ -n "$large_added_files" ]; then
-            echo "Large files added: $large_added_files"
-            echo "# Large files added:" >> $GITHUB_STEP_SUMMARY
-            echo "$large_added_files" >> $GITHUB_STEP_SUMMARY
-            exit 1
-          fi

README.md

@@ -1,339 +1,87 @@
-# 🦜🕸️LangGraph
+<picture class="github-only">
+  <source media="(prefers-color-scheme: light)" srcset="docs/docs/static/wordmark_dark.svg">
+  <source media="(prefers-color-scheme: dark)" srcset="docs/docs/static/wordmark_light.svg">
+  <img alt="LangGraph Logo" src="docs/docs/static/wordmark_dark.svg" width="80%">
+</picture>
 
-![Version](https://img.shields.io/pypi/v/langgraph)
+<div>
+<br>
+</div>
+
+[![Version](https://img.shields.io/pypi/v/langgraph.svg)](https://pypi.org/project/langgraph/)
 [![Downloads](https://static.pepy.tech/badge/langgraph/month)](https://pepy.tech/project/langgraph)
 [![Open Issues](https://img.shields.io/github/issues-raw/langchain-ai/langgraph)](https://github.com/langchain-ai/langgraph/issues)
 [![Docs](https://img.shields.io/badge/docs-latest-blue)](https://langchain-ai.github.io/langgraph/)
 
-⚡ Building language agents as graphs ⚡
-
 > [!NOTE]
 > Looking for the JS version? See the [JS repo](https://github.com/langchain-ai/langgraphjs) and the [JS docs](https://langchain-ai.github.io/langgraphjs/).
 
-## Overview
+LangGraph — used by Replit, Uber, LinkedIn, GitLab and more — is a low-level orchestration framework for building controllable agents. While langchain provides integrations and composable components to streamline LLM application development, the LangGraph library enables agent orchestration — offering customizable architectures, long-term memory, and human-in-the-loop to reliably handle complex tasks.
 
-[LangGraph](https://langchain-ai.github.io/langgraph/) is a library for building
-stateful, multi-actor applications with LLMs, used to create agent and multi-agent
-workflows. Check out an introductory tutorial [here](https://langchain-ai.github.io/langgraph/tutorials/introduction/).
-
-
-LangGraph is inspired by [Pregel](https://research.google/pubs/pub37252/) and [Apache Beam](https://beam.apache.org/). The public interface draws inspiration from [NetworkX](https://networkx.org/documentation/latest/). LangGraph is built by LangChain Inc, the creators of LangChain, but can be used without LangChain.
-
-### Why use LangGraph?
-
-LangGraph powers [production-grade agents](https://www.langchain.com/built-with-langgraph), trusted by Linkedin, Uber, Klarna, GitLab, and many more. LangGraph provides fine-grained control over both the flow and state of your agent applications. It implements a central [persistence layer](https://langchain-ai.github.io/langgraph/concepts/persistence/), enabling features that are common to most agent architectures:
-
-- **Memory**: LangGraph persists arbitrary aspects of your application's state,
-supporting memory of conversations and other updates within and across user
-interactions;
-- **Human-in-the-loop**: Because state is checkpointed, execution can be interrupted
-and resumed, allowing for decisions, validation, and corrections at key stages via
-human input.
-
-Standardizing these components allows individuals and teams to focus on the behavior
-of their agent, instead of its supporting infrastructure.
-
-Through [LangGraph Platform](#langgraph-platform), LangGraph also provides tooling for
-the development, deployment, debugging, and monitoring of your applications.
-
-LangGraph integrates seamlessly with
-[LangChain](https://python.langchain.com/docs/introduction/) and
-[LangSmith](https://docs.smith.langchain.com/) (but does not require them).
-
-To learn more about LangGraph, check out our first LangChain Academy
-course, *Introduction to LangGraph*, available for free
-[here](https://academy.langchain.com/courses/intro-to-langgraph).
-
-### LangGraph Platform
-
-[LangGraph Platform](https://langchain-ai.github.io/langgraph/concepts/langgraph_platform) is infrastructure for deploying LangGraph agents. It is a commercial solution for deploying agentic applications to production, built on the open-source LangGraph framework. The LangGraph Platform consists of several components that work together to support the development, deployment, debugging, and monitoring of LangGraph applications: [LangGraph Server](https://langchain-ai.github.io/langgraph/concepts/langgraph_server) (APIs), [LangGraph SDKs](https://langchain-ai.github.io/langgraph/concepts/sdk) (clients for the APIs), [LangGraph CLI](https://langchain-ai.github.io/langgraph/concepts/langgraph_cli) (command line tool for building the server), and [LangGraph Studio](https://langchain-ai.github.io/langgraph/concepts/langgraph_studio) (UI/debugger).
-
-See deployment options [here](https://langchain-ai.github.io/langgraph/concepts/deployment_options/)
-(includes a free tier).
-
-Here are some common issues that arise in complex deployments, which LangGraph Platform addresses:
-
-- **Streaming support**: LangGraph Server provides [multiple streaming modes](https://langchain-ai.github.io/langgraph/concepts/streaming) optimized for various application needs
-- **Background runs**: Runs agents asynchronously in the background
-- **Support for long running agents**: Infrastructure that can handle long running processes
-- **[Double texting](https://langchain-ai.github.io/langgraph/concepts/double_texting)**: Handle the case where you get two messages from the user before the agent can respond
-- **Handle burstiness**: Task queue for ensuring requests are handled consistently without loss, even under heavy loads
-
-## Installation
-
-```shell
+```bash
 pip install -U langgraph
 ```
 
-## Example
-
-Let's build a tool-calling [ReAct-style](https://langchain-ai.github.io/langgraph/concepts/agentic_concepts/#react-implementation) agent that uses a search tool!
-
-```shell
-pip install langchain-anthropic
-```
-
-```shell
-export ANTHROPIC_API_KEY=sk-...
-```
-
-Optionally, we can set up [LangSmith](https://docs.smith.langchain.com/) for best-in-class observability.
-
-```shell
-export LANGSMITH_TRACING=true
-export LANGSMITH_API_KEY=lsv2_sk_...
-```
-
-The simplest way to create a tool-calling agent in LangGraph is to use `create_react_agent`:
-
-<details open>
-  <summary>High-level implementation</summary>
+To learn more about how to use LangGraph, check out [the docs](https://langchain-ai.github.io/langgraph/). We show a simple example below of how to create a ReAct agent.
 
 ```python
+# This code depends on pip install langchain[anthropic]
 from langgraph.prebuilt import create_react_agent
-from langgraph.checkpoint.memory import MemorySaver
-from langchain_anthropic import ChatAnthropic
-from langchain_core.tools import tool
 
-# Define the tools for the agent to use
-@tool
 def search(query: str):
     """Call to surf the web."""
-    # This is a placeholder, but don't tell the LLM that...
     if "sf" in query.lower() or "san francisco" in query.lower():
         return "It's 60 degrees and foggy."
     return "It's 90 degrees and sunny."
 
-
-tools = [search]
-model = ChatAnthropic(model="claude-3-5-sonnet-latest", temperature=0)
-
-# Initialize memory to persist state between graph runs
-checkpointer = MemorySaver()
-
-app = create_react_agent(model, tools, checkpointer=checkpointer)
-
-# Use the agent
-final_state = app.invoke(
-    {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-    config={"configurable": {"thread_id": 42}}
+agent = create_react_agent("anthropic:claude-3-7-sonnet-latest", tools=[search])
+agent.invoke(
+    {"messages": [{"role": "user", "content": "what is the weather in sf"}]}
 )
-final_state["messages"][-1].content
-```
-```
-"Based on the search results, I can tell you that the current weather in San Francisco is:\n\nTemperature: 60 degrees Fahrenheit\nConditions: Foggy\n\nSan Francisco is known for its microclimates and frequent fog, especially during the summer months. The temperature of 60°F (about 15.5°C) is quite typical for the city, which tends to have mild temperatures year-round. The fog, often referred to as "Karl the Fog" by locals, is a characteristic feature of San Francisco\'s weather, particularly in the mornings and evenings.\n\nIs there anything else you\'d like to know about the weather in San Francisco or any other location?"
 ```
 
-Now when we pass the same <code>"thread_id"</code>, the conversation context is retained via the saved state (i.e. stored list of messages)
+## Why use LangGraph?
 
-```python
-final_state = app.invoke(
-    {"messages": [{"role": "user", "content": "what about ny"}]},
-    config={"configurable": {"thread_id": 42}}
-)
-final_state["messages"][-1].content
-```
+LangGraph is built for developers who want to build powerful, adaptable AI agents. Developers choose LangGraph for:
 
-```
-"Based on the search results, I can tell you that the current weather in New York City is:\n\nTemperature: 90 degrees Fahrenheit (approximately 32.2 degrees Celsius)\nConditions: Sunny\n\nThis weather is quite different from what we just saw in San Francisco. New York is experiencing much warmer temperatures right now. Here are a few points to note:\n\n1. The temperature of 90°F is quite hot, typical of summer weather in New York City.\n2. The sunny conditions suggest clear skies, which is great for outdoor activities but also means it might feel even hotter due to direct sunlight.\n3. This kind of weather in New York often comes with high humidity, which can make it feel even warmer than the actual temperature suggests.\n\nIt's interesting to see the stark contrast between San Francisco's mild, foggy weather and New York's hot, sunny conditions. This difference illustrates how varied weather can be across different parts of the United States, even on the same day.\n\nIs there anything else you'd like to know about the weather in New York or any other location?"
-```
-</details>
+- **Reliability and controllability.** Steer agent actions with moderation checks and human-in-the-loop approvals. LangGraph persists context for long-running workflows, keeping your agents on course.
+- **Low-level and extensible.** Build custom agents with fully descriptive, low-level primitives – free from rigid abstractions that limit customization. Design scalable multi-agent systems, with each agent serving a specific role tailored to your use case.
+- **First-class streaming support.** With token-by-token streaming and streaming of intermediate steps, LangGraph gives users clear visibility into agent reasoning and actions as they unfold in real time.
 
-> [!TIP]
-> LangGraph is a **low-level** framework that allows you to implement any custom agent
-architectures. Click on the low-level implementation below to see how to implement a
-tool-calling agent from scratch.
+LangGraph is trusted in production and powering agents for companies like:
 
-<details>
-<summary>Low-level implementation</summary>
+- [Klarna](https://blog.langchain.dev/customers-klarna/): Customer support bot for 85 million active users
+- [Elastic](https://www.elastic.co/blog/elastic-security-generative-ai-features): Security AI assistant for threat detection
+- [Uber](https://dpe.org/sessions/ty-smith-adam-huda/this-year-in-ubers-ai-driven-developer-productivity-revolution/): Automated unit test generation
+- [Replit](https://www.langchain.com/breakoutagents/replit): Code generation
+- And many more ([see list here](https://www.langchain.com/built-with-langgraph))
 
-```python
-from typing import Literal
+## LangGraph’s ecosystem
 
-from langchain_anthropic import ChatAnthropic
-from langchain_core.tools import tool
-from langgraph.checkpoint.memory import MemorySaver
-from langgraph.graph import END, START, StateGraph, MessagesState
-from langgraph.prebuilt import ToolNode
+While LangGraph can be used standalone, it also integrates seamlessly with any LangChain product, giving developers a full suite of tools for building agents. To improve your LLM application development, pair LangGraph with:
 
+- [LangSmith](http://www.langchain.com/langsmith) — Helpful for agent evals and observability. Debug poor-performing LLM app runs, evaluate agent trajectories, gain visibility in production, and improve performance over time.
+- [LangGraph Platform](https://langchain-ai.github.io/langgraph/concepts/#langgraph-platform) — Deploy and scale agents effortlessly with a purpose-built deployment platform for long running, stateful workflows. Discover, reuse, configure, and share agents across teams — and iterate quickly with visual prototyping in [LangGraph Studio](https://langchain-ai.github.io/langgraph/concepts/langgraph_studio/).
 
-# Define the tools for the agent to use
-@tool
-def search(query: str):
-    """Call to surf the web."""
-    # This is a placeholder, but don't tell the LLM that...
-    if "sf" in query.lower() or "san francisco" in query.lower():
-        return "It's 60 degrees and foggy."
-    return "It's 90 degrees and sunny."
+## Pairing with LangGraph Platform
 
+While LangGraph is our open-source agent orchestration framework, enterprises that need scalable agent deployment can benefit from [LangGraph Platform](https://langchain-ai.github.io/langgraph/concepts/langgraph_platform/).
 
-tools = [search]
+LangGraph Platform can help engineering teams:
 
-tool_node = ToolNode(tools)
+- **Accelerate agent development**: Quickly create agent UXs with configurable templates and [LangGraph Studio](https://langchain-ai.github.io/langgraph/concepts/langgraph_studio/) for visualizing and debugging agent interactions.
+- **Deploy seamlessly**: We handle the complexity of deploying your agent. LangGraph Platform includes robust APIs for memory, threads, and cron jobs plus auto-scaling task queues & servers.
+- **Centralize agent management & reusability**: Discover, reuse, and manage agents across the organization. Business users can also modify agents without coding.
 
-model = ChatAnthropic(model="claude-3-5-sonnet-latest", temperature=0).bind_tools(tools)
+## Additional resources
 
-# Define the function that determines whether to continue or not
-def should_continue(state: MessagesState) -> Literal["tools", END]:
-    messages = state['messages']
-    last_message = messages[-1]
-    # If the LLM makes a tool call, then we route to the "tools" node
-    if last_message.tool_calls:
-        return "tools"
-    # Otherwise, we stop (reply to the user)
-    return END
+- [LangChain Academy](https://academy.langchain.com/courses/intro-to-langgraph): Learn the basics of LangGraph in our free, structured course.
+- [Tutorials](https://langchain-ai.github.io/langgraph/tutorials/): Simple walkthroughs with guided examples on getting started with LangGraph.
+- [Templates](https://langchain-ai.github.io/langgraph/concepts/template_applications/): Pre-built reference apps for common agentic workflows (e.g. ReAct agent, memory, retrieval etc.) that can be cloned and adapted.
+- [How-to Guides](https://langchain-ai.github.io/langgraph/how-tos/): Quick, actionable code snippets for topics such as streaming, adding memory & persistence, and design patterns (e.g. branching, subgraphs, etc.).
+- [API Reference](https://langchain-ai.github.io/langgraph/reference/graphs/): Detailed reference on core classes, methods, how to use the graph and checkpointing APIs, and higher-level prebuilt components.
+- [Built with LangGraph](https://www.langchain.com/built-with-langgraph): Hear how industry leaders use LangGraph to ship powerful, production-ready AI applications.
 
+## Acknowledgements
 
-# Define the function that calls the model
-def call_model(state: MessagesState):
-    messages = state['messages']
-    response = model.invoke(messages)
-    # We return a list, because this will get added to the existing list
-    return {"messages": [response]}
-
-
-# Define a new graph
-workflow = StateGraph(MessagesState)
-
-# Define the two nodes we will cycle between
-workflow.add_node("agent", call_model)
-workflow.add_node("tools", tool_node)
-
-# Set the entrypoint as `agent`
-# This means that this node is the first one called
-workflow.add_edge(START, "agent")
-
-# We now add a conditional edge
-workflow.add_conditional_edges(
-    # First, we define the start node. We use `agent`.
-    # This means these are the edges taken after the `agent` node is called.
-    "agent",
-    # Next, we pass in the function that will determine which node is called next.
-    should_continue,
-)
-
-# We now add a normal edge from `tools` to `agent`.
-# This means that after `tools` is called, `agent` node is called next.
-workflow.add_edge("tools", 'agent')
-
-# Initialize memory to persist state between graph runs
-checkpointer = MemorySaver()
-
-# Finally, we compile it!
-# This compiles it into a LangChain Runnable,
-# meaning you can use it as you would any other runnable.
-# Note that we're (optionally) passing the memory when compiling the graph
-app = workflow.compile(checkpointer=checkpointer)
-
-# Use the agent
-final_state = app.invoke(
-    {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-    config={"configurable": {"thread_id": 42}}
-)
-final_state["messages"][-1].content
-```
-
-<b>Step-by-step Breakdown</b>:
-
-<details>
-<summary>Initialize the model and tools.</summary>
-<ul>
-  <li>
-    We use <code>ChatAnthropic</code> as our LLM. <strong>NOTE:</strong> we need to make sure the model knows that it has these tools available to call. We can do this by converting the LangChain tools into the format for OpenAI tool calling using the <code>.bind_tools()</code> method.
-  </li>
-  <li>
-    We define the tools we want to use - a search tool in our case. It is really easy to create your own tools - see documentation here on how to do that <a href="https://python.langchain.com/docs/how_to/custom_tools/">here</a>.
-  </li>
-</ul>
-</details>
-
-<details>
-<summary>Initialize graph with state.</summary>
-
-<ul>
-    <li>We initialize graph (<code>StateGraph</code>) by passing state schema (in our case <code>MessagesState</code>)</li>
-    <li><code>MessagesState</code> is a prebuilt state schema that has one attribute -- a list of LangChain <code>Message</code> objects, as well as logic for merging the updates from each node into the state.</li>
-</ul>
-</details>
-
-<details>
-<summary>Define graph nodes.</summary>
-
-There are two main nodes we need:
-
-<ul>
-    <li>The <code>agent</code> node: responsible for deciding what (if any) actions to take.</li>
-    <li>The <code>tools</code> node that invokes tools: if the agent decides to take an action, this node will then execute that action.</li>
-</ul>
-</details>
-
-<details>
-<summary>Define entry point and graph edges.</summary>
-
-First, we need to set the entry point for graph execution - <code>agent</code> node.
-
-Then we define one normal and one conditional edge. Conditional edge means that the destination depends on the contents of the graph's state (<code>MessagesState</code>). In our case, the destination is not known until the agent (LLM) decides.
-
-<ul>
-  <li>Conditional edge: after the agent is called, we should either:
-    <ul>
-      <li>a. Run tools if the agent said to take an action, OR</li>
-      <li>b. Finish (respond to the user) if the agent did not ask to run tools</li>
-    </ul>
-  </li>
-  <li>Normal edge: after the tools are invoked, the graph should always return to the agent to decide what to do next</li>
-</ul>
-</details>
-
-<details>
-<summary>Compile the graph.</summary>
-
-<ul>
-  <li>
-    When we compile the graph, we turn it into a LangChain 
-    <a href="https://python.langchain.com/docs/concepts/runnables/">Runnable</a>, 
-    which automatically enables calling <code>.invoke()</code>, <code>.stream()</code> and <code>.batch()</code> 
-    with your inputs
-  </li>
-  <li>
-    We can also optionally pass checkpointer object for persisting state between graph runs, and enabling memory, 
-    human-in-the-loop workflows, time travel and more. In our case we use <code>MemorySaver</code> - 
-    a simple in-memory checkpointer
-  </li>
-</ul>
-</details>
-
-<details>
-<summary>Execute the graph.</summary>
-
-<ol>
-  <li>LangGraph adds the input message to the internal state, then passes the state to the entrypoint node, <code>"agent"</code>.</li>
-  <li>The <code>"agent"</code> node executes, invoking the chat model.</li>
-  <li>The chat model returns an <code>AIMessage</code>. LangGraph adds this to the state.</li>
-  <li>Graph cycles the following steps until there are no more <code>tool_calls</code> on <code>AIMessage</code>:
-    <ul>
-      <li>If <code>AIMessage</code> has <code>tool_calls</code>, <code>"tools"</code> node executes</li>
-      <li>The <code>"agent"</code> node executes again and returns <code>AIMessage</code></li>
-    </ul>
-  </li>
-  <li>Execution progresses to the special <code>END</code> value and outputs the final state. And as a result, we get a list of all our chat messages as output.</li>
-</ol>
-</details>
-
-</details>
-
-## Documentation
-
-* [Tutorials](https://langchain-ai.github.io/langgraph/tutorials/): Learn to build with LangGraph through guided examples.
-* [How-to Guides](https://langchain-ai.github.io/langgraph/how-tos/): Accomplish specific things within LangGraph, from streaming, to adding memory & persistence, to common design patterns (branching, subgraphs, etc.), these are the place to go if you want to copy and run a specific code snippet.
-* [Conceptual Guides](https://langchain-ai.github.io/langgraph/concepts/high_level/): In-depth explanations of the key concepts and principles behind LangGraph, such as nodes, edges, state and more.
-* [API Reference](https://langchain-ai.github.io/langgraph/reference/graphs/): Review important classes and methods, simple examples of how to use the graph and checkpointing APIs, higher-level prebuilt components and more.
-* [LangGraph Platform](https://langchain-ai.github.io/langgraph/concepts/#langgraph-platform): LangGraph Platform is a commercial solution for deploying agentic applications in production, built on the open-source LangGraph framework.
-
-## Resources
-
-* [Built with LangGraph](https://www.langchain.com/built-with-langgraph): Hear how industry leaders use LangGraph to ship powerful, production-ready AI applications.
-
-## Contributing
-
-For more information on how to contribute, see [here](https://github.com/langchain-ai/langgraph/blob/main/CONTRIBUTING.md).
+LangGraph is inspired by [Pregel](https://research.google/pubs/pub37252/) and [Apache Beam](https://beam.apache.org/). The public interface draws inspiration from [NetworkX](https://networkx.org/documentation/latest/). LangGraph is built by LangChain Inc, the creators of LangChain, but can be used without LangChain.

docs/Makefile

@@ -10,7 +10,15 @@ build-prebuilt:
 	# Use to create an update to date prebuilt page.
 	# Looks up download stats for each of the prebuilt packages and
 	# generates the final prebuilt page.
-	poetry run python -m _scripts.third_party_page.get_download_stats stats.yml
+	@if [ "$(DOWNLOAD_STATS)" = "true" ]; then \
+		set -x; \
+		poetry run python -m _scripts.third_party_page.get_download_stats stats.yml; \
+		set +x; \
+	else \
+		set -x; \
+		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
 
 build-docs: build-typedoc build-prebuilt

docs/README.md

@@ -14,6 +14,8 @@ To run the documentation server locally you can run:
 make serve-docs
 ```
 
+This will start the documentation server on [http://127.0.0.1:8000/langgraph/](http://127.0.0.1:8000/langgraph/).
+
 ## Execute notebooks
 
 If you would like to automatically execute all of the notebooks, to mimic the "Run notebooks" GHA, you can run:

docs/_scripts/notebook_hooks.py

@@ -186,7 +186,7 @@ def _on_page_markdown_with_config(
 
     if remove_base64_images:
         # Remove base64 encoded images from markdown
-        markdown = re.sub(r"!\[.*?\]\(data:image/+;base64,[^\)]+\)", "", markdown)
+        markdown = re.sub(r"!\[.*?\]\(data:image/[^;]+;base64,[^)]+\)", "", markdown)
 
     return markdown
 

docs/_scripts/third_party_page/get_download_stats.py

@@ -30,10 +30,23 @@ PACKAGES_FILE = HERE / "packages.yml"
 PACKAGES = yaml.safe_load(PACKAGES_FILE.read_text())['packages']
 
 
-def _get_weekly_downloads(packages: list[Package]) -> list[ResolvedPackage]:
+def _get_weekly_downloads(packages: list[Package], fake: bool) -> list[ResolvedPackage]:
     """Retrieve the monthly download count for a list of packages from PyPIStats."""
     resolved_packages: list[ResolvedPackage] = []
 
+    if fake:
+        # To avoid making network requests during testing, return fake download counts
+        for package in packages:
+            resolved_packages.append(
+                {
+                    "name": package["name"],
+                    "repo": package["repo"],
+                    "weekly_downloads": -12345,
+                    "description": package["description"],
+                }
+            )
+        return resolved_packages
+
     for package in packages:
         # First check if package exists on PyPI
         pypi_url = f"https://pypi.org/pypi/{package['name']}/json"
@@ -88,13 +101,13 @@ def _get_weekly_downloads(packages: list[Package]) -> list[ResolvedPackage]:
 
 
 
-def main(output_file: str) -> None:
+def main(output_file: str, fake: bool) -> None:
     """Main function to generate package download information.
 
     Args:
         output_file: Path to the output YAML file.
     """
-    resolved_packages: list[ResolvedPackage] = _get_weekly_downloads(PACKAGES)
+    resolved_packages: list[ResolvedPackage] = _get_weekly_downloads(PACKAGES, fake)
 
     if not output_file.endswith(".yml"):
         raise ValueError("Output file must have a .yml extension")
@@ -115,6 +128,15 @@ if __name__ == "__main__":
             "downloads.yml"
         ),
     )
+    parser.add_argument(
+        "--fake",
+        default=False,
+        action="store_true",
+        help=(
+            "Generate fake download counts for testing purposes. "
+            "This option will not make any network requests."
+        ),
+    )
     args = parser.parse_args()
 
-    main(args.output_file)
+    main(args.output_file, args.fake)

docs/_scripts/third_party_page/packages.yml

@@ -23,4 +23,16 @@ packages:
     description: "Build swarm-style multi-agent systems using LangGraph."
   - name: "delve-taxonomy-generator"
     repo: "andrestorres123/delve"
-    description: "A taxonomy generator for unstructured data"
+    description: "A taxonomy generator for unstructured data"
+  - name: "nodeology"
+    repo: "xyin-anl/Nodeology"
+    description: "Enable researcher to build scientific workflows easily with simplified interface."
+  - name: "langgraph-bigtool"
+    repo: "langchain-ai/langgraph-bigtool"
+    description: "Build LangGraph agents with large numbers of tools."
+  - name: "ai-data-science-team"
+    repo: "business-science/ai-data-science-team"
+    description: "An AI-powered data science team of agents to help you perform common data science tasks 10X faster."
+  - name: "langgraph-reflection"
+    repo: "langchain-ai/langgraph-reflection"
+    description: "LangGraph agent that runs a reflection step."

docs/docs/adopters.md

@@ -1,4 +1,4 @@
-# 🦜🕸️ LangGraph Adopters
+# 🦜🕸️ Companies using LangGraph
 
 This list of companies using LangGraph and their success stories is compiled from public sources. If your company uses LangGraph, we'd love for you to share your story and add it to the list. You’re also welcome to contribute updates based on publicly available information from other companies, such as blog posts or press releases.
 

docs/docs/cloud/deployment/semantic_search.md

@@ -17,7 +17,7 @@ This guide explains how to add semantic search to your LangGraph deployment's cr
     ...
     "store": {
         "index": {
-            "embed": "openai:text-embeddings-3-small",
+            "embed": "openai:text-embedding-3-small",
             "dims": 1536,
             "fields": ["$"]
         }
@@ -27,7 +27,7 @@ This guide explains how to add semantic search to your LangGraph deployment's cr
 
 This configuration:
 
-- Uses OpenAI's text-embeddings-3-small model for generating embeddings
+- Uses OpenAI's text-embedding-3-small model for generating embeddings
 - Sets the embedding dimension to 1536 (matching the model's output)
 - Indexes all fields in your stored data (`["$"]` means index everything, or specify specific fields like `["text", "metadata.title"]`)
 

docs/docs/cloud/deployment/setup.md

@@ -36,21 +36,20 @@ Dependencies can optionally be specified in one of the following files: `pyproje
 The dependencies below will be included in the image, you can also use them in your code, as long as with a compatible version range:
 
 ```
-langgraph>=0.2.56,<0.3.0
-langgraph-checkpoint>=2.0.5,<3.0
+langgraph>=0.2.56,<0.4.0
+langgraph-sdk>=0.1.53
+langgraph-checkpoint>=2.0.15,<3.0
 langchain-core>=0.2.38,<0.4.0
 langsmith>=0.1.63
 orjson>=3.9.7
 httpx>=0.25.0
 tenacity>=8.0.0
 uvicorn>=0.26.0
-sse-starlette>=2.1.0
+sse-starlette>=2.1.0,<2.2.0
 uvloop>=0.18.0
 httptools>=0.5.0
-jsonschema-rs>=0.16.3
-croniter>=1.0.1
+jsonschema-rs>=0.20.0
 structlog>=23.1.0
-redis>=5.0.0,<6.0.0
 ```
 
 Example `requirements.txt` file:

docs/docs/cloud/deployment/setup_pyproject.md

@@ -36,21 +36,20 @@ Dependencies can optionally be specified in one of the following files: `pyproje
 The dependencies below will be included in the image, you can also use them in your code, as long as with a compatible version range:
 
 ```
-langgraph>=0.2.56,<0.3.0
-langgraph-checkpoint>=2.0.5,<3.0
+langgraph>=0.2.56,<0.4.0
+langgraph-sdk>=0.1.53
+langgraph-checkpoint>=2.0.15,<3.0
 langchain-core>=0.2.38,<0.4.0
 langsmith>=0.1.63
 orjson>=3.9.7
 httpx>=0.25.0
 tenacity>=8.0.0
 uvicorn>=0.26.0
-sse-starlette>=2.1.0
+sse-starlette>=2.1.0,<2.2.0
 uvloop>=0.18.0
 httptools>=0.5.0
-jsonschema-rs>=0.16.3
-croniter>=1.0.1
+jsonschema-rs>=0.20.0
 structlog>=23.1.0
-redis>=5.0.0,<6.0.0
 ```
 
 Example `pyproject.toml` file:
@@ -65,7 +64,7 @@ license = "MIT"
 readme = "README.md"
 
 [tool.poetry.dependencies]
-python = ">=3.9.0,<3.13"
+python = ">=3.9"
 langgraph = "^0.2.0"
 langchain-fireworks = "^0.1.3"
 

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

@@ -0,0 +1,312 @@
+# How to implement Generative User Interfaces with LangGraph
+
+!!! info "Prerequisites"
+
+    - [LangGraph Platform](../../concepts/langgraph_platform.md)
+    - [LangGraph Server](../../concepts/langgraph_server.md)
+    - [`useStream()` React Hook](./use_stream_react.md)
+
+Generative user interfaces (Generative UI) allows agents to go beyond text and generate rich user interfaces. This enables creating more interactive and context-aware applications where the UI adapts based on the conversation flow and AI responses.
+
+![Generative UI Sample](./img/generative_ui_sample.jpg)
+
+LangGraph Platform supports colocating your React components with your graph code. This allows you to focus on building specific UI components for your graph while easily plugging into existing chat interfaces such as [Agent Chat](https://agentchat.vercel.app) and loading the code only when actually needed.
+
+!!! warning "LangGraph.js only"
+
+    Currently only LangGraph.js supports Generative UI. Support for Python is coming soon.
+
+## Tutorial
+
+### 1. Define and configure UI components
+
+First, create your first UI component. For each component you need to provide an unique identifier that will be used to reference the component in your graph code.
+
+```tsx title="src/agent/ui.tsx"
+const WeatherComponent = (props: { city: string }) => {
+  return <div>Weather for {props.city}</div>;
+};
+
+export default {
+  weather: WeatherComponent,
+};
+```
+
+Next, define your UI components in your `langgraph.json` configuration:
+
+```json
+{
+  "node_version": "20",
+  "graphs": {
+    "agent": "./src/agent/index.ts:graph"
+  },
+  "ui": {
+    "agent": "./src/agent/ui.tsx"
+  }
+}
+```
+
+The `ui` section points to the UI components that will be used by graphs. By default, we recommend using the same key as the graph name, but you can split out the components however you like, see [Customise the namespace of UI components](#customise-the-namespace-of-ui-components) for more details.
+
+LangGraph Platform will automatically bundle your UI components code and styles and serve them as external assets that can be loaded by the `LoadExternalComponent` component. Some dependencies such as `react` and `react-dom` will be automatically excluded from the bundle.
+
+CSS and Tailwind 4.x is also supported out of the box, so you can freely use Tailwind classes as well as `shadcn/ui` in your UI components.
+
+=== "`src/agent/ui.tsx`"
+
+    ```tsx
+    import "./styles.css";
+
+    const WeatherComponent = (props: { city: string }) => {
+      return <div className="bg-red-500">Weather for {props.city}</div>;
+    };
+
+    export default {
+      weather: WeatherComponent,
+    };
+    ```
+
+=== "`src/agent/styles.css`"
+
+    ```css
+    @import "tailwindcss";
+    ```
+
+### 2. Send the UI components in your graph
+
+Use the `typedUi` utility to emit UI elements from your agent nodes:
+
+```typescript title="src/agent/index.ts"
+import {
+  typedUi,
+  uiMessageReducer,
+} from "@langchain/langgraph-sdk/react-ui/server";
+
+import { ChatOpenAI } from "@langchain/openai";
+import { v4 as uuidv4 } from "uuid";
+import { z } from "zod";
+
+import type ComponentMap from "./ui.js";
+
+import {
+  Annotation,
+  MessagesAnnotation,
+  StateGraph,
+  type LangGraphRunnableConfig,
+} from "@langchain/langgraph";
+
+const AgentState = Annotation.Root({
+  ...MessagesAnnotation.spec,
+  ui: Annotation({ reducer: uiMessageReducer, default: () => [] }),
+});
+
+export const graph = new StateGraph(AgentState)
+  .addNode("weather", async (state, config) => {
+    // Provide the type of the component map to ensure
+    // type safety of `ui.push()` calls as well as
+    // pushing the messages to the `ui` and sending a custom event as well.
+    const ui = typedUi<typeof ComponentMap>(config);
+
+    const weather = await new ChatOpenAI({ model: "gpt-4o-mini" })
+      .withStructuredOutput(z.object({ city: z.string() }))
+      .withConfig({ tags: ["langsmith:nostream"] })
+      .invoke(state.messages);
+
+    const response = {
+      id: uuidv4(),
+      type: "ai",
+      content: `Here's the weather for ${weather.city}`,
+    };
+
+    // Emit UI elements with associated AI message
+    ui.push({ name: "weather", props: weather }, { message: response });
+
+    return { messages: [response] };
+  })
+  .addEdge("__start__", "weather")
+  .compile();
+```
+
+### 3. Handle UI elements in your React application
+
+On the client side, you can use `useStream()` and `LoadExternalComponent` to display the UI elements.
+
+```tsx title="src/app/page.tsx"
+"use client";
+
+import { useStream } from "@langchain/langgraph-sdk/react";
+import { LoadExternalComponent } from "@langchain/langgraph-sdk/react-ui";
+
+export default function Page() {
+  const { thread, values } = useStream({
+    apiUrl: "http://localhost:2024",
+    assistantId: "agent",
+  });
+
+  return (
+    <div>
+      {thread.messages.map((message) => (
+        <div key={message.id}>
+          {message.content}
+          {values.ui
+            ?.filter((ui) => ui.metadata?.message_id === message.id)
+            .map((ui) => (
+              <LoadExternalComponent key={ui.id} stream={thread} message={ui} />
+            ))}
+        </div>
+      ))}
+    </div>
+  );
+}
+```
+
+Behind the scenes, `LoadExternalComponent` will fetch the JS and CSS for the UI components from LangGraph Platform and render them in a shadow DOM, thus ensuring style isolation from the rest of your application.
+
+## How-to guides
+
+### Show loading UI when components are loading
+
+You can provide a fallback UI to be rendered when the components are loading.
+
+```tsx
+<LoadExternalComponent
+  stream={thread}
+  message={ui}
+  fallback={<div>Loading...</div>}
+/>
+```
+
+### Provide custom components on the client side
+
+If you already have the components loaded in your client application, you can provide a map of such components to be rendered directly without fetching the UI code from LangGraph Platform.
+
+```tsx
+const clientComponents = {
+  weather: WeatherComponent,
+};
+
+<LoadExternalComponent
+  stream={thread}
+  message={ui}
+  components={clientComponents}
+/>;
+```
+
+### Customise the namespace of UI components.
+
+By default `LoadExternalComponent` will use the `assistantId` from `useStream()` hook to fetch the code for UI components. You can customise this by providing a `namespace` prop to the `LoadExternalComponent` component.
+
+=== "`src/app/page.tsx`"
+
+    ```tsx
+    <LoadExternalComponent
+      stream={thread}
+      message={ui}
+      namespace="custom-namespace"
+    />
+    ```
+
+=== "`langgraph.json`"
+
+    ```json
+    {
+      "ui": {
+        "custom-namespace": "./src/agent/ui.tsx"
+      }
+    }
+    ```
+
+### Access and interact with the thread state from the UI component
+
+You can access the thread state inside the UI component by using the `useStreamContext` hook.
+
+```tsx
+import { useStreamContext } from "@langchain/langgraph-sdk/react-ui";
+
+const WeatherComponent = (props: { city: string }) => {
+  const { thread, submit } = useStreamContext();
+  return (
+    <>
+      <div>Weather for {props.city}</div>
+
+      <button
+        onClick={() => {
+          const newMessage = {
+            type: "human",
+            content: `What's the weather in ${props.city}?`,
+          };
+
+          submit({ messages: [newMessage] });
+        }}
+      >
+        Retry
+      </button>
+    </>
+  );
+};
+```
+
+### Pass additional context to the client components
+
+You can pass additional context to the client components by providing a `meta` prop to the `LoadExternalComponent` component.
+
+```tsx
+<LoadExternalComponent stream={thread} message={ui} meta={{ userId: "123" }} />
+```
+
+Then, you can access the `meta` prop in the UI component by using the `useStreamContext` hook.
+
+```tsx
+import { useStreamContext } from "@langchain/langgraph-sdk/react-ui";
+
+const WeatherComponent = (props: { city: string }) => {
+  const { meta } = useStreamContext<
+    { city: string },
+    { MetaType: { userId?: string } }
+  >();
+
+  return (
+    <div>
+      Weather for {props.city} (user: {meta?.userId})
+    </div>
+  );
+};
+```
+
+### Streaming UI updates before the node execution is finished
+
+You can stream UI updates before the node execution is finished by using the `onCustomEvent` callback of the `useStream()` hook.
+
+```tsx
+import { uiMessageReducer } from "@langchain/langgraph-sdk/react-ui";
+
+const { thread, submit } = useStream({
+  apiUrl: "http://localhost:2024",
+  assistantId: "agent",
+  onCustomEvent: (event, options) => {
+    options.mutate((prev) => {
+      const ui = uiMessageReducer(prev.ui ?? [], event);
+      return { ...prev, ui };
+    });
+  },
+});
+```
+
+### Remove UI messages from state
+
+Similar to how messages can be removed from the state by appending a RemoveMessage you can remove an UI message from the state by calling `ui.delete` with the ID of the UI message.
+
+```tsx
+// pushed message
+const message = ui.push({ name: "weather", props: { city: "London" } });
+
+// remove said message
+ui.delete(message.id);
+
+// return new state to persist changes
+return { ui: ui.items };
+```
+
+## Learn more
+
+- [JS/TS SDK Reference](../reference/sdk/js_ts_sdk_ref.md)

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

@@ -63,7 +63,7 @@ Now, let's invoke our graph by interrupting before `ask_human` node:
         "messages": [
             {
                 "role": "user",
-                "content": "Use the search tool to ask the user where they are, then look up the weather there",
+                "content": "Ask the user where they are, then look up the weather there",
             }
         ]
     }
@@ -85,8 +85,7 @@ Now, let's invoke our graph by interrupting before `ask_human` node:
       messages: [
         {
           role: "human",
-          content: "Use the search tool to ask the user where they are, then look up the weather there"
-        }
+          content: "Ask the user where they are, then look up the weather there" }
       ]
     };
 
@@ -115,7 +114,7 @@ Now, let's invoke our graph by interrupting before `ask_human` node:
      --header 'Content-Type: application/json' \
      --data "{
        \"assistant_id\": \"agent\",
-       \"input\": {\"messages\": [{\"role\": \"human\", \"content\": \"Use the search tool to ask the user where they are, then look up the weather there\"}]},
+       \"input\": {\"messages\": [{\"role\": \"human\", \"content\": \"Ask the user where they are, then look up the weather there\"}]},
        \"interrupt_before\": [\"ask_human\"],
        \"stream_mode\": [
          \"updates\"

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

@@ -1,6 +1,133 @@
 # Prompt Engineering in LangGraph Studio
 
-In LangGraph Studio you can iterate on the prompts used within your graph by utilizing the LangSmith Playground. To do so:
+## Overview
+
+A central aspect of agent development is prompt engineering. LangGraph Studio makes it easy to iterate on the prompts used within your graph directly within the UI.
+
+## Setup
+
+The first step is to define your [configuration](https://langchain-ai.github.io/langgraph/how-tos/configuration/) such that LangGraph Studio is aware of the prompts you want to iterate on and which nodes they are associated with.
+
+### Reference
+
+When defining your configuration, you can use special metadata keys to instruct LangGraph Studio how to handle different fields. Here's a reference for the available configuration options:
+
+#### `langgraph_nodes`
+
+- **Description**: Specifies which graph nodes a configuration field is associated with.
+- **Value Type**: Array of strings, where each string is the name of a node in your graph.
+- **Usage Context**: Include in the `json_schema_extra` dictionary for Pydantic models or the `metadata["json_schema_extra"]` dictionary for dataclasses.
+- **Required**: No, but necessary if you want a field to be editable for specific nodes in the UI.
+- **Example**:
+  ```python
+  system_prompt: str = Field(
+      default="You are a helpful AI assistant.",
+      json_schema_extra={"langgraph_nodes": ["call_model", "other_node"]},
+  )
+  ```
+
+#### `langgraph_type`
+
+- **Description**: Specifies the type of configuration field, which determines how it's handled in the UI.
+- **Value Type**: String
+- **Supported Values**:
+  - `"prompt"`: Indicates the field contains prompt text that should be treated specially in the UI.
+- **Usage Context**: Include in the `json_schema_extra` dictionary for Pydantic models or the `metadata["json_schema_extra"]` dictionary for dataclasses.
+- **Required**: No, but helpful for prompt fields to enable special handling.
+- **Example**:
+  ```python
+  system_prompt: str = Field(
+      default="You are a helpful AI assistant.",
+      json_schema_extra={
+          "langgraph_nodes": ["call_model"],
+          "langgraph_type": "prompt",
+      },
+  )
+  ```
+
+### Example
+
+For example, if you have a node called `call_model` whose system prompt you want to iterate on, you can define a configuration like the following.
+
+```python
+## Using Pydantic
+from pydantic import BaseModel, Field
+from typing import Annotated, Literal
+
+class Configuration(BaseModel):
+    """The configuration for the agent."""
+
+    system_prompt: str = Field(
+        default="You are a helpful AI assistant.",
+        description="The system prompt to use for the agent's interactions. "
+        "This prompt sets the context and behavior for the agent.",
+        json_schema_extra={
+            "langgraph_nodes": ["call_model"],
+            "langgraph_type": "prompt",
+        },
+    )
+
+    model: Annotated[
+        Literal[
+            "anthropic/claude-3-7-sonnet-latest",
+            "anthropic/claude-3-5-haiku-latest",
+            "openai/o1",
+            "openai/gpt-4o-mini",
+            "openai/o1-mini",
+            "openai/o3-mini",
+        ],
+        {"__template_metadata__": {"kind": "llm"}},
+    ] = Field(
+        default="openai/gpt-4o-mini",
+        description="The name of the language model to use for the agent's main interactions. "
+        "Should be in the form: provider/model-name.",
+        json_schema_extra={"langgraph_nodes": ["call_model"]},
+    )
+
+## Using Dataclasses
+from dataclasses import dataclass, field
+
+@dataclass(kw_only=True)
+class Configuration:
+    """The configuration for the agent."""
+
+    system_prompt: str = field(
+        default="You are a helpful AI assistant.",
+        metadata={
+            "description": "The system prompt to use for the agent's interactions. "
+            "This prompt sets the context and behavior for the agent.",
+            "json_schema_extra": {"langgraph_nodes": ["call_model"]},
+        },
+    )
+
+    model: Annotated[str, {"__template_metadata__": {"kind": "llm"}}] = field(
+        default="anthropic/claude-3-5-sonnet-20240620",
+        metadata={
+            "description": "The name of the language model to use for the agent's main interactions. "
+            "Should be in the form: provider/model-name.",
+            "json_schema_extra": {"langgraph_nodes": ["call_model"]},
+        },
+    )
+
+```
+
+## Iterating on prompts
+
+### Node Configuration
+
+With this set up, running your graph and viewing in LangGraph Studio will result in the graph rendering like such.
+
+**Note the configuration icon in the top right corner of the `call_model` node**:
+
+![Graph in Studio](../img/studio_graph_with_configuration.png){width=1200}
+
+Clicking this icon will open a modal where you can edit the configuration for all of the fields associated with the `call_model` node. From here, you can save your changes and apply them to the graph. Note that these values reflect the currently active assistant, and saving will update the assistant with the new values.
+
+![Configuration modal](../img/studio_node_configuration.png){width=1200}
+
+### Playground
+
+LangGraph Studio also supports prompt engineering through an integration with the LangSmith Playground. To do so:
 
 1. Open an existing thread or create a new one.
 2. Within the thread log, any nodes that have made an LLM call will have a "View LLM Runs" button. Clicking this will open a popover with the LLM runs for that node.
@@ -8,8 +135,6 @@ In LangGraph Studio you can iterate on the prompts used within your graph by uti
 
 ![Playground in Studio](../img/studio_playground.png){width=1200}
 
-
-
 From here you can edit the prompt, test different model configurations and re-run just this LLM call without having to re-run the entire graph. When you are happy with your changes, you can copy the updated prompt back into your graph.
 
 For more information on how to use the LangSmith Playground, see the [LangSmith Playground documentation](https://docs.smith.langchain.com/prompt_engineering/how_to_guides#playground).

docs/docs/concepts/agentic_concepts.md

@@ -14,7 +14,7 @@ As a result, there are many different types of [agent architectures](https://blo
 
 ## Router
 
-A router allows an LLM to select a single step from a specified set of options. This is an agent architecture that exhibits a relatively limited level of control because the LLM usually focuses on making a single decision and produces a specific output from limited set of pre-defined options. Routers typically employ a few different concepts to achieve this.
+A router allows an LLM to select a single step from a specified set of options. This is an agent architecture that exhibits a relatively limited level of control because the LLM usually focuses on making a single decision and produces a specific output from a limited set of pre-defined options. Routers typically employ a few different concepts to achieve this.
 
 ### Structured Output
 

docs/docs/concepts/langgraph_studio.md

@@ -103,7 +103,7 @@ def routing_function(state: GraphState) -> Literal["node_b","node_c"]:
 ### Studio Desktop FAQs
 
 !!! warning "Deprecation Warning"
-In order to support a wider range of platforms and users, we now recommend following the above instructions to connect to LangGraph Studio using the development server instead of the desktop app.
+    In order to support a wider range of platforms and users, we now recommend following the above instructions to connect to LangGraph Studio using the development server instead of the desktop app.
 
 The LangGraph Studio Desktop App is a standalone application that allows you to connect to your LangGraph application and visualize and interact with your graph. It is available for MacOS only and requires Docker to be installed.
 

docs/docs/concepts/multi_agent.md

@@ -89,7 +89,7 @@ def transfer_to_bob(state):
     )
 ```
 
-This is a special case of updating the graph state from tools where in addition the state update, the control flow is included as well.
+This is a special case of updating the graph state from tools where, in addition to the state update, the control flow is included as well.
 
 !!! important
 
@@ -235,7 +235,7 @@ supervisor = create_react_agent(model, tools)
 
 ### Hierarchical
 
-As you add more agents to your system, it might become too hard for the supervisor to manage all of them. The supervisor might start making poor decisions about which agent to call next, the context might become too complex for a single supervisor to keep track of. In other words, you end up with the same problems that motivated the multi-agent architecture in the first place.
+As you add more agents to your system, it might become too hard for the supervisor to manage all of them. The supervisor might start making poor decisions about which agent to call next, or the context might become too complex for a single supervisor to keep track of. In other words, you end up with the same problems that motivated the multi-agent architecture in the first place.
 
 To address this, you can design your system _hierarchically_. For example, you can create separate, specialized teams of agents managed by individual supervisors, and a top-level supervisor to manage the teams.
 
@@ -339,9 +339,9 @@ builder.add_edge("agent_1", "agent_2")
 
 ## Communication between agents
 
-The most important thing when building multi-agent systems is figuring out how the agents communicate. There are few different considerations:
+The most important thing when building multi-agent systems is figuring out how the agents communicate. There are a few different considerations:
 
-- Do agents communicate via [**via graph state or via tool calls**](#graph-state-vs-tool-calls)?
+- Do agents communicate [**via graph state or via tool calls**](#graph-state-vs-tool-calls)?
 - What if two agents have [**different state schemas**](#different-state-schemas)?
 - How to communicate over a [**shared message list**](#shared-message-list)?
 

docs/docs/concepts/persistence.md

@@ -32,7 +32,7 @@ from typing_extensions import TypedDict
 from operator import add
 
 class State(TypedDict):
-    foo: int
+    foo: str
     bar: Annotated[list[str], add]
 
 def node_a(state: State):
@@ -232,7 +232,7 @@ from langgraph.store.memory import InMemoryStore
 in_memory_store = InMemoryStore()
 ```
 
-Memories are namespaced by a `tuple`, which in this specific example will be `(<user_id>, "memories")`. The namespace can be any length and represent anything, does not have be user specific.
+Memories are namespaced by a `tuple`, which in this specific example will be `(<user_id>, "memories")`. The namespace can be any length and represent anything, does not have to be user specific.
 
 ```python 
 user_id = "1"
@@ -387,6 +387,9 @@ We can access the memories and use them in our model call.
 def call_model(state: MessagesState, config: RunnableConfig, *, store: BaseStore):
     # Get the user id from the config
     user_id = config["configurable"]["user_id"]
+
+    # Namespace the memory
+    namespace = (user_id, "memories")
     
     # Search based on the most recent message
     memories = store.search(

docs/docs/how-tos/command.ipynb

@@ -33,7 +33,7 @@
     "    )\n",
     "```\n",
     "\n",
-    "If you are using [subgraphs](#subgraphs), you might want to navigate from a node a subgraph to a different subgraph (i.e. a different node in the parent graph). To do so, you can specify `graph=Command.PARENT` in `Command`:\n",
+    "If you are using [subgraphs](#subgraphs), you might want to navigate from a node within a subgraph to a different subgraph (i.e. a different node in the parent graph). To do so, you can specify `graph=Command.PARENT` in `Command`:\n",
     "\n",
     "```python\n",
     "def my_node(state: State) -> Command[Literal[\"my_other_node\"]]:\n",

docs/docs/how-tos/create-react-agent-memory.ipynb

@@ -122,20 +122,18 @@
     "\n",
     "# For this tutorial we will use custom tool that returns pre-defined values for weather in two cities (NYC & SF)\n",
     "\n",
-    "from typing import Literal\n",
-    "\n",
     "from langchain_core.tools import tool\n",
     "\n",
     "\n",
     "@tool\n",
-    "def get_weather(city: Literal[\"nyc\", \"sf\"]):\n",
+    "def get_weather(location: str) -> str:\n",
     "    \"\"\"Use this to get weather information.\"\"\"\n",
-    "    if city == \"nyc\":\n",
+    "    if any([city in location.lower() for city in [\"nyc\", \"new york city\"]]):\n",
     "        return \"It might be cloudy in nyc\"\n",
-    "    elif city == \"sf\":\n",
+    "    elif any([city in location.lower() for city in [\"sf\", \"san francisco\"]]):\n",
     "        return \"It's always sunny in sf\"\n",
     "    else:\n",
-    "        raise AssertionError(\"Unknown city\")\n",
+    "        return f\"I am not sure what the weather is in {location}\"\n",
     "\n",
     "\n",
     "tools = [get_weather]\n",
@@ -220,7 +218,7 @@
    "id": "838a043f-90ad-4e69-9d1d-6e22db2c346c",
    "metadata": {},
    "source": [
-    "Notice that when we pass the same the same thread ID, the chat history is preserved"
+    "Notice that when we pass the same thread ID, the chat history is preserved."
    ]
   },
   {

docs/docs/how-tos/human_in_the_loop/wait-user-input.ipynb

@@ -397,7 +397,8 @@
     "# We define a fake node to ask the human\n",
     "def ask_human(state):\n",
     "    tool_call_id = state[\"messages\"][-1].tool_calls[0][\"id\"]\n",
-    "    location = interrupt(\"Please provide your location:\")\n",
+    "    ask = AskHuman.model_validate(state[\"messages\"][-1].tool_calls[0][\"args\"])\n",
+    "    location = interrupt(ask.question)\n",
     "    tool_message = [{\"tool_call_id\": tool_call_id, \"type\": \"tool\", \"content\": location}]\n",
     "    return {\"messages\": tool_message}\n",
     "\n",
@@ -491,7 +492,7 @@
     "        \"messages\": [\n",
     "            (\n",
     "                \"user\",\n",
-    "                \"Use the search tool to ask the user where they are, then look up the weather there\",\n",
+    "                \"Ask the user where they are, then look up the weather there\",\n",
     "            )\n",
     "        ]\n",
     "    },\n",

docs/docs/how-tos/index.md

@@ -198,7 +198,6 @@ Learn how to set up your app for deployment to LangGraph Platform:
 - [How to test locally](../cloud/deployment/test_locally.md)
 - [How to rebuild graph at runtime](../cloud/deployment/graph_rebuild.md)
 - [How to use LangGraph Platform to deploy CrewAI, AutoGen, and other frameworks](autogen-langgraph-platform.ipynb)
-- [How to integrate LangGraph into your React application](../cloud/how-tos/use_stream_react.md)
 
 ### Deployment
 
@@ -257,6 +256,13 @@ Streaming the results of your LLM application is vital for ensuring a good user
 - [How to stream in debug mode](../cloud/how-tos/stream_debug.md)
 - [How to stream multiple modes](../cloud/how-tos/stream_multiple.md)
 
+### Frontend and Generative UI
+
+With LangGraph Platform you can integrate LangGraph agents into your React applications and colocate UI components with your agent code. 
+
+- [How to integrate LangGraph into your React application](../cloud/how-tos/use_stream_react.md)
+- [How to implement Generative User Interfaces with LangGraph](../cloud/how-tos/generative_ui_react.md)
+
 ### Human-in-the-loop
 
 When designing complex graphs, relying entirely on the LLM for decision-making can be risky, particularly when it involves tools that interact with files, APIs, or databases. These interactions may lead to unintended data access or modifications, depending on the use case. To mitigate these risks, LangGraph allows you to integrate human-in-the-loop behavior, ensuring your LLM applications operate as intended without undesirable outcomes.

docs/docs/how-tos/memory/add-summary-conversation-history.ipynb

@@ -10,6 +10,7 @@
     "One of the most common use cases for persistence is to use it to keep track of conversation history. This is great - it makes it easy to continue conversations. As conversations get longer and longer, however, this conversation history can build up and take up more and more of the context window. This can often be undesirable as it leads to more expensive and longer calls to the LLM, and potentially ones that error. One way to work around that is to create a summary of the conversation to date, and use that with the past N messages. This guide will go through an example of how to do that.\n",
     "\n",
     "This will involve a few steps:\n",
+    "\n",
     "- Check if the conversation is too long (can be done by checking number of messages or length of messages)\n",
     "- If yes, the create summary (will need a prompt for this)\n",
     "- Then remove all except the last N messages\n",
@@ -98,7 +99,7 @@
     "from typing import Literal\n",
     "\n",
     "from langchain_anthropic import ChatAnthropic\n",
-    "from langchain_core.messages import SystemMessage, RemoveMessage\n",
+    "from langchain_core.messages import SystemMessage, RemoveMessage, HumanMessage\n",
     "from langgraph.checkpoint.memory import MemorySaver\n",
     "from langgraph.graph import MessagesState, StateGraph, START, END\n",
     "\n",

docs/docs/how-tos/memory/manage-conversation-history.ipynb

@@ -7,7 +7,7 @@
    "source": [
     "# How to manage conversation history\n",
     "\n",
-    "One of the most common use cases for persistence is to use it to keep track of conversation history. This is great - it makes it easy to continue conversations. As conversations get longer and longer, however, this conversation history can build up and take up more and more of the context window. This can often be undesirable as it leads to more expensive and longer calls to the LLM, and potentially ones that error. In order to prevent this from happening, you need to probably manage the conversation history.\n",
+    "One of the most common use cases for persistence is to use it to keep track of conversation history. This is great - it makes it easy to continue conversations. As conversations get longer and longer, however, this conversation history can build up and take up more and more of the context window. This can often be undesirable as it leads to more expensive and longer calls to the LLM, and potentially ones that error. In order to prevent this from happening, you need to properly manage the conversation history.\n",
     "\n",
     "Note: this guide focuses on how to do this in LangGraph, where you can fully customize how this is done. If you want a more off-the-shelf solution, you can look into functionality provided in LangChain:\n",
     "\n",

docs/docs/how-tos/pass-run-time-values-to-tools.ipynb

@@ -38,7 +38,7 @@
     "    </p>\n",
     "</div> \n",
     "\n",
-    "The core technique the examples below is to **annotate** a parameter as \"injected\", meaning it will be injected by your program and should not be seen or populated by the LLM. Let the following codesnippet serve as a tl;dr:\n",
+    "The core technique in the examples below is to **annotate** a parameter as \"injected\", meaning it will be injected by your program and should not be seen or populated by the LLM. Let the following codesnippet serve as a tl;dr:\n",
     "\n",
     "```python\n",
     "from typing import Annotated\n",

docs/docs/how-tos/react-agent-structured-output.ipynb

@@ -65,7 +65,7 @@
     "\n",
     "**Pros and Cons**\n",
     "\n",
-    "The benefit to this format is that you only need one LLM, and can save money and latency because of this. The downside to this option is that it isn't guaranteed that the single LLM will call the correct tool when you want it to. We can help the LLM by setting `tool_choice` to `any` when we use `bind_tools` which forces the LLM to select at least one tool at every turn, but this is far from a fool proof strategy. In addition, another downside is that the agent might call *multiple* tools, so we need to check for this explicitly in our routing function (or if we are using OpenAI we an set `parallell_tool_calling=False` to ensure only one tool is called at a time).\n",
+    "The benefit to this format is that you only need one LLM, and can save money and latency because of this. The downside to this option is that it isn't guaranteed that the single LLM will call the correct tool when you want it to. We can help the LLM by setting `tool_choice` to `any` when we use `bind_tools` which forces the LLM to select at least one tool at every turn, but this is far from a foolproof strategy. In addition, another downside is that the agent might call *multiple* tools, so we need to check for this explicitly in our routing function (or if we are using OpenAI we can set `parallell_tool_calling=False` to ensure only one tool is called at a time).\n",
     "\n",
     "**Option 2**\n",
     "\n",

docs/docs/how-tos/state-model.ipynb

@@ -266,6 +266,235 @@
     "    print(\"An exception was raised because bad_node sets `a` to an integer.\")\n",
     "    print(e)"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "2270bc3c",
+   "metadata": {},
+   "source": [
+    "## Multiple Nodes\n",
+    "\n",
+    "Run-time validation will also work in a multi-node graph. In the example below `bad_node` updates `a` to an integer. \n",
+    "\n",
+    "Because run-time validation occurs on **inputs**, the validation error will occur when `ok_node` is called (not when `bad_node` returns an update to the state which is inconsistent with the schema)."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d832cdcc",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langgraph.graph import StateGraph, START, END\n",
+    "from typing_extensions import TypedDict\n",
+    "\n",
+    "from pydantic import BaseModel\n",
+    "\n",
+    "\n",
+    "# The overall state of the graph (this is the public state shared across nodes)\n",
+    "class OverallState(BaseModel):\n",
+    "    a: str\n",
+    "\n",
+    "\n",
+    "def bad_node(state: OverallState):\n",
+    "    return {\n",
+    "        \"a\": 123  # Invalid\n",
+    "    }\n",
+    "\n",
+    "\n",
+    "def ok_node(state: OverallState):\n",
+    "    return {\"a\": \"goodbye\"}\n",
+    "\n",
+    "\n",
+    "# Build the state graph\n",
+    "builder = StateGraph(OverallState)\n",
+    "builder.add_node(bad_node)\n",
+    "builder.add_node(ok_node)\n",
+    "builder.add_edge(START, \"bad_node\")\n",
+    "builder.add_edge(\"bad_node\", \"ok_node\")\n",
+    "builder.add_edge(\"ok_node\", END)\n",
+    "graph = builder.compile()\n",
+    "\n",
+    "# Test the graph with a valid input\n",
+    "try:\n",
+    "    graph.invoke({\"a\": \"hello\"})\n",
+    "except Exception as e:\n",
+    "    print(\"An exception was raised because bad_node sets `a` to an integer.\")\n",
+    "    print(e)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "456b1f77",
+   "metadata": {},
+   "source": [
+    "## Advanced Pydantic Model Usage\n",
+    "\n",
+    "This section covers more advanced topics when using Pydantic models with LangGraph.\n",
+    "\n",
+    "### Serialization Behavior\n",
+    "\n",
+    "When using Pydantic models as state schemas, it's important to understand how serialization works, especially when:\n",
+    "- Passing Pydantic objects as inputs\n",
+    "- Receiving outputs from the graph\n",
+    "- Working with nested Pydantic models\n",
+    "\n",
+    "Let's see these behaviors in action:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0e919cdc",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langgraph.graph import StateGraph, START, END\n",
+    "from pydantic import BaseModel\n",
+    "\n",
+    "\n",
+    "class NestedModel(BaseModel):\n",
+    "    value: str\n",
+    "\n",
+    "\n",
+    "class ComplexState(BaseModel):\n",
+    "    text: str\n",
+    "    count: int\n",
+    "    nested: NestedModel\n",
+    "\n",
+    "\n",
+    "def process_node(state: ComplexState):\n",
+    "    # Node receives a validated Pydantic object\n",
+    "    print(f\"Input state type: {type(state)}\")\n",
+    "    print(f\"Nested type: {type(state.nested)}\")\n",
+    "\n",
+    "    # Return a dictionary update\n",
+    "    return {\"text\": state.text + \" processed\", \"count\": state.count + 1}\n",
+    "\n",
+    "\n",
+    "# Build the graph\n",
+    "builder = StateGraph(ComplexState)\n",
+    "builder.add_node(\"process\", process_node)\n",
+    "builder.add_edge(START, \"process\")\n",
+    "builder.add_edge(\"process\", END)\n",
+    "graph = builder.compile()\n",
+    "\n",
+    "# Create a Pydantic instance for input\n",
+    "input_state = ComplexState(text=\"hello\", count=0, nested=NestedModel(value=\"test\"))\n",
+    "print(f\"Input object type: {type(input_state)}\")\n",
+    "\n",
+    "# Invoke graph with a Pydantic instance\n",
+    "result = graph.invoke(input_state)\n",
+    "print(f\"Output type: {type(result)}\")\n",
+    "print(f\"Output content: {result}\")\n",
+    "\n",
+    "# Convert back to Pydantic model if needed\n",
+    "output_model = ComplexState(**result)\n",
+    "print(f\"Converted back to Pydantic: {type(output_model)}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "f13f28ce",
+   "metadata": {},
+   "source": [
+    "### Runtime Type Coercion\n",
+    "\n",
+    "Pydantic performs runtime type coercion for certain data types. This can be helpful but also lead to unexpected behavior if you're not aware of it."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "faf59316",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langgraph.graph import StateGraph, START, END\n",
+    "from pydantic import BaseModel\n",
+    "\n",
+    "\n",
+    "class CoercionExample(BaseModel):\n",
+    "    # Pydantic will coerce string numbers to integers\n",
+    "    number: int\n",
+    "    # Pydantic will parse string booleans to bool\n",
+    "    flag: bool\n",
+    "\n",
+    "\n",
+    "def inspect_node(state: CoercionExample):\n",
+    "    print(f\"number: {state.number} (type: {type(state.number)})\")\n",
+    "    print(f\"flag: {state.flag} (type: {type(state.flag)})\")\n",
+    "    return {}\n",
+    "\n",
+    "\n",
+    "builder = StateGraph(CoercionExample)\n",
+    "builder.add_node(\"inspect\", inspect_node)\n",
+    "builder.add_edge(START, \"inspect\")\n",
+    "builder.add_edge(\"inspect\", END)\n",
+    "graph = builder.compile()\n",
+    "\n",
+    "# Demonstrate coercion with string inputs that will be converted\n",
+    "result = graph.invoke({\"number\": \"42\", \"flag\": \"true\"})\n",
+    "\n",
+    "# This would fail with a validation error\n",
+    "try:\n",
+    "    graph.invoke({\"number\": \"not-a-number\", \"flag\": \"true\"})\n",
+    "except Exception as e:\n",
+    "    print(f\"\\nExpected validation error: {e}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "2844475b",
+   "metadata": {},
+   "source": [
+    "### Working with Message Models\n",
+    "\n",
+    "When working with LangChain message types in your state schema, there are important considerations for serialization. You should use `AnyMessage` (rather than `BaseMessage`) for proper serialization/deserialization when using message objects over the wire:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "bd0734b0",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langgraph.graph import StateGraph, START, END\n",
+    "from pydantic import BaseModel\n",
+    "from langchain_core.messages import HumanMessage, AIMessage, AnyMessage\n",
+    "from typing import List\n",
+    "\n",
+    "\n",
+    "class ChatState(BaseModel):\n",
+    "    messages: List[AnyMessage]\n",
+    "    context: str\n",
+    "\n",
+    "\n",
+    "def add_message(state: ChatState):\n",
+    "    return {\"messages\": state.messages + [AIMessage(content=\"Hello there!\")]}\n",
+    "\n",
+    "\n",
+    "builder = StateGraph(ChatState)\n",
+    "builder.add_node(\"add_message\", add_message)\n",
+    "builder.add_edge(START, \"add_message\")\n",
+    "builder.add_edge(\"add_message\", END)\n",
+    "graph = builder.compile()\n",
+    "\n",
+    "# Create input with a message\n",
+    "initial_state = ChatState(\n",
+    "    messages=[HumanMessage(content=\"Hi\")], context=\"Customer support chat\"\n",
+    ")\n",
+    "\n",
+    "result = graph.invoke(initial_state)\n",
+    "print(f\"Output: {result}\")\n",
+    "\n",
+    "# Convert back to Pydantic model to see message types\n",
+    "output_model = ChatState(**result)\n",
+    "for i, msg in enumerate(output_model.messages):\n",
+    "    print(f\"Message {i}: {type(msg).__name__} - {msg.content}\")"
+   ]
   }
  ],
  "metadata": {

docs/docs/how-tos/update-state-from-tools.ipynb

@@ -210,7 +210,7 @@
    "id": "cbb06aea-6654-4245-91f8-af6e8f2b5377",
    "metadata": {},
    "source": [
-    "Let's now add personalization: we'll respond differently to the user based on the state values AFTER the state has been updated from the tool. To achieve this, let's define a function that will dynamically construct the system prompt based on the graph state. It will be called ever time the LLM is called and the function output will be passed to the LLM:"
+    "Let's now add personalization: we'll respond differently to the user based on the state values AFTER the state has been updated from the tool. To achieve this, let's define a function that will dynamically construct the system prompt based on the graph state. It will be called every time the LLM is called and the function output will be passed to the LLM:"
    ]
   },
   {

docs/docs/index.md

@@ -3,4 +3,26 @@ hide_comments: true
 title: Home
 ---
 
+<script>
+  // This script only runs in MkDocs, not on GitHub
+  var hideGitHubVersion = function() {
+    document.querySelectorAll('.github-only').forEach(el => el.style.display = 'none');
+  };
+
+  // Handle both initial load and subsequent navigation
+  document.addEventListener('DOMContentLoaded', hideGitHubVersion);
+  document$.subscribe(hideGitHubVersion);
+</script>
+
+<p class="mkdocs-only">
+  <img class="logo-light" src="static/wordmark_dark.svg" alt="LangGraph Logo" width="80%">
+  <img class="logo-dark" src="static/wordmark_light.svg" alt="LangGraph Logo" width="80%">
+</p>
+
+<style>
+h1 {
+  display: none;
+}
+</style>
+
 {!../README.md!}

docs/docs/llms.txt

@@ -0,0 +1,191 @@
+# LangGraph
+
+## Quickstart
+
+These guides are designed to help you get started with LangGraph.
+
+- [LangGraph Quickstart](https://langchain-ai.github.io/langgraph/tutorials/introduction/): Build a chatbot that can use tools and keep track of conversation history. Add human-in-the-loop capabilities and explore how time-travel works.
+- [Common Workflows](https://langchain-ai.github.io/langgraph/tutorials/workflows/): Overview of the most common workflows using LLMs implemented with LangGraph.
+- [LangGraph Server Quickstart](https://langchain-ai.github.io/langgraph/tutorials/langgraph-platform/local-server/): Launch a LangGraph server locally and interact with it using REST API and LangGraph Studio Web UI.
+- [Deploy with LangGraph Cloud Quickstart](https://langchain-ai.github.io/langgraph/cloud/quick_start/): Deploy a LangGraph app using LangGraph Cloud.
+
+## Concepts
+
+These guides provide explanations of the key concepts behind the LangGraph framework.
+
+- [Why LangGraph?](https://langchain-ai.github.io/langgraph/concepts/high_level/): Motivation for LangGraph, a library for building agentic applications with LLMs.
+- [LangGraph Glossary](https://langchain-ai.github.io/langgraph/concepts/low_level/): LangGraph workflows are designed as graphs, with nodes representing different components and edges representing the flow of information between them. This guide provides an overview of the key concepts associated with LangGraph graph primitives.
+- [Common Agentic Patterns](https://langchain-ai.github.io/langgraph/concepts/agentic_concepts/): An agent uses an LLM to pick its own control flow to solve more complex problems! Agents are a key building block in many LLM applications. This guide explains the different types of agent architectures and how they can be used to control the flow of an application.
+- [Multi-Agent Systems](https://langchain-ai.github.io/langgraph/concepts/multi_agent/): Complex LLM applications can often be broken down into multiple agents, each responsible for a different part of the application. This guide explains common patterns for building multi-agent systems.
+- [Breakpoints](https://langchain-ai.github.io/langgraph/concepts/breakpoints/): Breakpoints allow pausing the execution of a graph at specific points. Breakpoints allow stepping through graph execution for debugging purposes.
+- [Human-in-the-Loop](https://langchain-ai.github.io/langgraph/concepts/human_in_the_loop/): Explains different ways of integrating human feedback into a LangGraph application.
+- [Time Travel](https://langchain-ai.github.io/langgraph/concepts/time-travel/): Time travel allows you to replay past actions in your LangGraph application to explore alternative paths and debug issues.
+- [Persistence](https://langchain-ai.github.io/langgraph/concepts/persistence/): LangGraph has a built-in persistence layer, implemented through checkpointers. This persistence layer helps to support powerful capabilities like human-in-the-loop, memory, time travel, and fault-tolerance.
+- [Memory](https://langchain-ai.github.io/langgraph/concepts/memory/): Memory in AI applications refers to the ability to process, store, and effectively recall information from past interactions. With memory, your agents can learn from feedback and adapt to users' preferences.
+- [Streaming](https://langchain-ai.github.io/langgraph/concepts/streaming/): Streaming is crucial for enhancing the responsiveness of applications built on LLMs. By displaying output progressively, even before a complete response is ready, streaming significantly improves user experience (UX), particularly when dealing with the latency of LLMs.
+- [Functional API](https://langchain-ai.github.io/langgraph/concepts/functional_api/): `@entrypoint` and `@task` decorators that allow you to add LangGraph functionality to an existing codebase.
+- [Durable Execution](https://langchain-ai.github.io/langgraph/concepts/durable_execution/): LangGraph's built-in [persistence](https://langchain-ai.github.io/langgraph/concepts/persistence/) layer provides durable execution for workflows, ensuring that the state of each execution step is saved to a durable store. 
+- [Pregel](https://langchain-ai.github.io/langgraph/concepts/pregel/): Pregel is LangGraph's runtime, which is responsible for managing the execution of LangGraph applications.
+- [FAQ](https://langchain-ai.github.io/langgraph/concepts/faq/): Frequently asked questions about LangGraph.
+
+## How-tos
+
+Here you’ll find answers to “How do I...?” types of questions. 
+
+These guides are **goal-oriented** and concrete. 
+
+They're meant to help you complete a specific task.
+
+### Graph API Basics
+
+- [How to update graph state from nodes](https://langchain-ai.github.io/langgraph/how-tos/state-reducers/)
+- [How to create a sequence of steps](https://langchain-ai.github.io/langgraph/how-tos/sequence/)
+- [How to create branches for parallel execution](https://langchain-ai.github.io/langgraph/how-tos/branching/)
+- [How to create and control loops with recursion limits](https://langchain-ai.github.io/langgraph/how-tos/recursion-limit/)
+- [How to visualize your graph](https://langchain-ai.github.io/langgraph/how-tos/visualization/)
+
+### Fine-grained Control
+
+These guides demonstrate LangGraph features that grant fine-grained control over the execution of your graph.
+
+- [How to create map-reduce branches for parallel execution](https://langchain-ai.github.io/langgraph/how-tos/map-reduce/)
+- [How to update state and jump to nodes in graphs and subgraphs](https://langchain-ai.github.io/langgraph/how-tos/command/)
+- [How to add runtime configuration to your graph](https://langchain-ai.github.io/langgraph/how-tos/configuration/)
+- [How to add node retries](https://langchain-ai.github.io/langgraph/how-tos/node-retries/)
+- [How to return state before hitting recursion limit](https://langchain-ai.github.io/langgraph/how-tos/return-when-recursion-limit-hits/)
+
+### Persistence
+ 
+Persistence makes it easy to persist state across graph runs (per-thread persistence) and across threads (cross-thread persistence). 
+
+These how-to guides show how to add persistence to your graph.
+
+- [How to add thread-level persistence to your graph](https://langchain-ai.github.io/langgraph/how-tos/persistence/)
+- [How to add thread-level persistence to a subgraph](https://langchain-ai.github.io/langgraph/how-tos/subgraph-persistence/)
+- [How to add cross-thread persistence to your graph](https://langchain-ai.github.io/langgraph/how-tos/cross-thread-persistence/)
+- [How to use Postgres checkpointer for persistence](https://langchain-ai.github.io/langgraph/how-tos/persistence_postgres/)
+- [How to use MongoDB checkpointer for persistence](https://langchain-ai.github.io/langgraph/how-tos/persistence_mongodb/)
+- [How to create a custom checkpointer using Redis](https://langchain-ai.github.io/langgraph/how-tos/persistence_redis/)
+
+See the below guides for how-to add persistence to your workflow using the [Functional API](https://langchain-ai.github.io/langgraph/concepts/functional_api/):
+
+- [How to add thread-level persistence (functional API)](https://langchain-ai.github.io/langgraph/how-tos/persistence-functional/)
+- [How to add cross-thread persistence (functional API)](https://langchain-ai.github.io/langgraph/how-tos/cross-thread-persistence-functional/)
+
+### Memory
+
+LangGraph makes it easy to manage conversation memory in your graph. These how-to guides show how to implement different strategies for that.
+
+- [How to manage conversation history](https://langchain-ai.github.io/langgraph/how-tos/memory/manage-conversation-history/)
+- [How to delete messages](https://langchain-ai.github.io/langgraph/how-tos/memory/delete-messages/)
+- [How to add summary conversation memory](https://langchain-ai.github.io/langgraph/how-tos/memory/add-summary-conversation-history/)
+- [How to add long-term memory (cross-thread)](https://langchain-ai.github.io/langgraph/how-tos/memory/cross-thread-persistence/)
+- [How to use semantic search for long-term memory](https://langchain-ai.github.io/langgraph/how-tos/memory/semantic-search/)
+
+### Human-in-the-loop
+
+Human-in-the-loop functionality allows you to involve humans in the decision-making process of your graph.
+
+These how-to guides show how to implement human-in-the-loop workflows in your graph.
+
+- [How to wait for user input](https://langchain-ai.github.io/langgraph/how-tos/human_in_the_loop/wait-user-input/): A basic example that shows how to implement a human-in-the-loop workflow in your graph using the `interrupt` function.
+- [How to review tool calls](https://langchain-ai.github.io/langgraph/how-tos/human_in_the_loop/review-tool-calls/): Incorporate human-in-the-loop for reviewing/editing/accepting tool call requests before they executed using the `interrupt` function.
+- [How to add static breakpoints](https://langchain-ai.github.io/langgraph/how-tos/human_in_the_loop/breakpoints/): Use for debugging purposes. For human-in-the-loop workflows, we recommend the [`interrupt` function](https://langchain-ai.github.io/langgraph/reference/types/#langgraph.types.interrupt) instead.
+- [How to edit graph state](https://langchain-ai.github.io/langgraph/how-tos/human_in_the_loop/edit-graph-state/): Edit graph state using `graph.update_state` method. Use this if implementing a **human-in-the-loop** workflow via **static breakpoints**.
+
+See the below guides for how-to implement human-in-the-loop workflows with the Functional API.
+
+- [How to wait for user input (Functional API)](https://langchain-ai.github.io/langgraph/how-tos/wait-user-input-functional/)
+- [How to review tool calls (Functional API)](https://langchain-ai.github.io/langgraph/how-tos/review-tool-calls-functional/)
+
+### Time Travel
+
+[Time travel](https://langchain-ai.github.io/langgraph/concepts/time-travel/) allows you to replay past actions in your LangGraph application to explore alternative paths and debug issues. These how-to guides show how to use time travel in your graph.
+
+- [How to view and update past graph state](https://langchain-ai.github.io/langgraph/how-tos/time-travel/)
+
+### Streaming
+
+[Streaming](https://langchain-ai.github.io/langgraph/concepts/streaming/) is crucial for enhancing the responsiveness of applications built on LLMs. By displaying output progressively, even before a complete response is ready, streaming significantly improves user experience (UX), particularly when dealing with the latency of LLMs.
+
+- [How to stream](https://langchain-ai.github.io/langgraph/how-tos/streaming/)
+- [How to stream LLM tokens](https://langchain-ai.github.io/langgraph/how-tos/streaming-tokens/)
+- [How to stream LLM tokens from specific nodes](https://langchain-ai.github.io/langgraph/how-tos/streaming-specific-nodes/)
+- [How to stream data from within a tool](https://langchain-ai.github.io/langgraph/how-tos/streaming-events-from-within-tools/)
+- [How to stream from subgraphs](https://langchain-ai.github.io/langgraph/how-tos/streaming-subgraphs/)
+- [How to disable streaming for models that don't support it](https://langchain-ai.github.io/langgraph/how-tos/disable-streaming/)
+
+### Tool calling
+
+[Tool calling](https://python.langchain.com/docs/concepts/tool_calling/) is a type of [chat model](https://python.langchain.com/docs/concepts/chat_models/) API.
+
+It accepts tool schemas, along with messages, as input and returns invocations of those tools as part of the output message.
+
+These how-to guides show common patterns for tool calling with LangGraph:
+
+- [How to call tools using ToolNode](https://langchain-ai.github.io/langgraph/how-tos/tool-calling/)
+- [How to handle tool calling errors](https://langchain-ai.github.io/langgraph/how-tos/tool-calling-errors/)
+- [How to pass runtime values to tools](https://langchain-ai.github.io/langgraph/how-tos/pass-run-time-values-to-tools/)
+- [How to pass config to tools](https://langchain-ai.github.io/langgraph/how-tos/pass-config-to-tools/)
+- [How to update graph state from tools](https://langchain-ai.github.io/langgraph/how-tos/update-state-from-tools/)
+- [How to handle large numbers of tools](https://langchain-ai.github.io/langgraph/how-tos/many-tools/)
+
+### Subgraphs
+
+Subgraphs allow you to reuse an existing graph from another graph. 
+
+These how-to guides show how to use subgraphs:
+
+- [How to use subgraphs](https://langchain-ai.github.io/langgraph/how-tos/subgraph/)
+- [How to view and update state in subgraphs](https://langchain-ai.github.io/langgraph/how-tos/subgraphs-manage-state/)
+- [How to transform inputs and outputs of a subgraph](https://langchain-ai.github.io/langgraph/how-tos/subgraph-transform-state/)
+
+### Multi-agent
+
+Multi-agent systems are useful to break down complex LLM applications into multiple agents, each responsible for a different part of the application. 
+
+These how-to guides show how to implement multi-agent systems in LangGraph:
+
+- [How to implement handoffs between agents](https://langchain-ai.github.io/langgraph/how-tos/agent-handoffs/)
+- [How to build a multi-agent network](https://langchain-ai.github.io/langgraph/how-tos/multi-agent-network/)
+- [How to add multi-turn conversation in a multi-agent application](https://langchain-ai.github.io/langgraph/how-tos/multi-agent-multi-turn-convo/)
+
+### State Management
+
+- [How to use Pydantic model as graph state](https://langchain-ai.github.io/langgraph/how-tos/state-model/)
+- [How to define input/output schema for your graph](https://langchain-ai.github.io/langgraph/how-tos/input_output_schema/)
+- [How to pass private state between nodes inside the graph](https://langchain-ai.github.io/langgraph/how-tos/pass_private_state/)
+
+### Other
+
+- [How to run graph asynchronously](https://langchain-ai.github.io/langgraph/how-tos/async/)
+- [How to force tool-calling agent to structure output](https://langchain-ai.github.io/langgraph/how-tos/react-agent-structured-output/)
+- [How to pass custom LangSmith run ID for graph runs](https://langchain-ai.github.io/langgraph/how-tos/run-id-langsmith/)
+- [How to integrate LangGraph with AutoGen, CrewAI, and other frameworks](https://langchain-ai.github.io/langgraph/how-tos/autogen-integration/)
+
+## Use cases 
+
+Explore practical implementations tailored for specific scenarios:
+
+### Chatbots
+
+- [Customer Support](https://langchain-ai.github.io/langgraph/tutorials/customer-support/customer-support/): Build a multi-functional support bot for flights, hotels, and car rentals.
+- [Prompt Generation from User Requirements](https://langchain-ai.github.io/langgraph/tutorials/chatbots/information-gather-prompting/): Build an information gathering chatbot.
+- [Code Assistant](https://langchain-ai.github.io/langgraph/tutorials/code_assistant/langgraph_code_assistant/): Build a code analysis and generation assistant.
+
+### RAG
+
+- [Agentic RAG](https://langchain-ai.github.io/langgraph/tutorials/rag/langgraph_agentic_rag/): Use an agent to figure out how to retrieve the most relevant information before using the retrieved information to answer the user's question.
+- [Adaptive RAG](https://langchain-ai.github.io/langgraph/tutorials/rag/langgraph_adaptive_rag/): Adaptive RAG is a strategy for RAG that unites (1) query analysis with (2) active / self-corrective RAG. Implementation of: https://arxiv.org/abs/2403.14403
+    - For a version that uses a local LLM: [Adaptive RAG using local LLMs](https://langchain-ai.github.io/langgraph/tutorials/rag/langgraph_adaptive_rag_local/)
+- [Corrective RAG](https://langchain-ai.github.io/langgraph/tutorials/rag/langgraph_crag/): Uses an LLM to grade the quality of the retrieved information from the given source, and if the quality is low, it will try to retrieve the information from another source. Implementation of: https://arxiv.org/pdf/2401.15884.pdf 
+    - For a version that uses a local LLM: [Corrective RAG using local LLMs](https://langchain-ai.github.io/langgraph/tutorials/rag/langgraph_crag_local/)
+- [Self-RAG](https://langchain-ai.github.io/langgraph/tutorials/rag/langgraph_self_rag/): Self-RAG is a strategy for RAG that incorporates self-reflection / self-grading on retrieved documents and generations. Implementation of https://arxiv.org/abs/2310.11511.
+    - For a version that uses a local LLM: [Self-RAG using local LLMs](https://langchain-ai.github.io/langgraph/tutorials/rag/langgraph_self_rag_local/) 
+- [SQL Agent](https://langchain-ai.github.io/langgraph/tutorials/sql-agent/): Build a SQL agent that can answer questions about a SQL database.
+
+### Multi-Agent Systems
+
+- [Network](https://langchain-ai.github.io/langgraph/tutorials/multi_agent/multi-agent-collaboration/): Enable two or more agents to collaborate on a task
+- [Supervisor](https://langchain-ai.github.io/langgraph/tutorials/multi_agent/agent_supervisor/): Use an LLM to orchestrate and delegate to individual agents
+- [Hierarchical Teams](https://langchain-ai.github.io/langgraph/tutorials/multi_agent/hierarchical_agent_teams/): Orchestrate nested teams of agents to solve problems

docs/mkdocs.yml

@@ -85,7 +85,7 @@ plugins:
 
 nav:
   - Home:
-    - Introduction: index.md
+    - index.md
     - Get started:
       - Learn the basics: tutorials/introduction.ipynb
       - Deployment:
@@ -231,6 +231,7 @@ nav:
                 - cloud/how-tos/stream_debug.md
                 - cloud/how-tos/stream_multiple.md
                 - cloud/how-tos/use_stream_react.md
+                - cloud/how-tos/generative_ui_react.md
             - Human-in-the-loop:
                 - Human-in-the-loop: how-tos#human-in-the-loop_1
                 - cloud/how-tos/human_in_the_loop_breakpoint.md
@@ -359,7 +360,7 @@ nav:
     - Resources:
       # NOTE: prebuilt.md is auto-generated by `make build-prebuilt`
       - Prebuilt Agents: prebuilt.md
-      - Adopters: adopters.md
+      - Companies using LangGraph: adopters.md
       - FAQ: concepts/faq.md
       - Troubleshooting:
           - Troubleshooting: troubleshooting/errors/index.md

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

@@ -2,6 +2,7 @@ import asyncio
 import logging
 from collections.abc import AsyncIterator, Iterable, Sequence
 from contextlib import asynccontextmanager
+from types import TracebackType
 from typing import Any, Callable, Optional, Union, cast
 
 import orjson
@@ -20,11 +21,12 @@ from langgraph.store.base import (
 )
 from langgraph.store.base.batch import AsyncBatchedBaseStore
 from langgraph.store.postgres.base import (
-    _PLACEHOLDER,
+    PLACEHOLDER,
     BasePostgresStore,
     PoolConfig,
     PostgresIndexConfig,
     Row,
+    TTLConfig,
     _decode_ns_bytes,
     _ensure_index_config,
     _group_ops,
@@ -106,6 +108,11 @@ class AsyncPostgresStore(AsyncBatchedBaseStore, BasePostgresStore[_ainternal.Con
         Semantic search is disabled by default. You can enable it by providing an `index` configuration
         when creating the store. Without this configuration, all `index` arguments passed to
         `put` or `aput` will have no effect.
+
+    Note:
+        If you provide a TTL configuration, you must explicitly call `start_ttl_sweeper()` to begin
+        the background task that removes expired items. Call `stop_ttl_sweeper()` to properly
+        clean up resources when you're done with the store.
     """
 
     __slots__ = (
@@ -115,7 +122,11 @@ class AsyncPostgresStore(AsyncBatchedBaseStore, BasePostgresStore[_ainternal.Con
         "supports_pipeline",
         "index_config",
         "embeddings",
+        "ttl_config",
+        "_ttl_sweeper_task",
+        "_ttl_stop_event",
     )
+    supports_ttl: bool = True
 
     def __init__(
         self,
@@ -126,6 +137,7 @@ class AsyncPostgresStore(AsyncBatchedBaseStore, BasePostgresStore[_ainternal.Con
             Callable[[Union[bytes, orjson.Fragment]], dict[str, Any]]
         ] = None,
         index: Optional[PostgresIndexConfig] = None,
+        ttl: Optional[TTLConfig] = None,
     ) -> None:
         if isinstance(conn, AsyncConnectionPool) and pipe is not None:
             raise ValueError(
@@ -141,10 +153,13 @@ class AsyncPostgresStore(AsyncBatchedBaseStore, BasePostgresStore[_ainternal.Con
         self.index_config = index
         if self.index_config:
             self.embeddings, self.index_config = _ensure_index_config(self.index_config)
-
         else:
             self.embeddings = None
 
+        self.ttl_config = ttl
+        self._ttl_sweeper_task: Optional[asyncio.Task[None]] = None
+        self._ttl_stop_event = asyncio.Event()
+
     async def abatch(self, ops: Iterable[Op]) -> list[Result]:
         grouped_ops, num_ops = _group_ops(ops)
         results: list[Result] = [None] * num_ops
@@ -167,6 +182,7 @@ class AsyncPostgresStore(AsyncBatchedBaseStore, BasePostgresStore[_ainternal.Con
         pipeline: bool = False,
         pool_config: Optional[PoolConfig] = None,
         index: Optional[PostgresIndexConfig] = None,
+        ttl: Optional[TTLConfig] = None,
     ) -> AsyncIterator["AsyncPostgresStore"]:
         """Create a new AsyncPostgresStore instance from a connection string.
 
@@ -198,16 +214,16 @@ class AsyncPostgresStore(AsyncBatchedBaseStore, BasePostgresStore[_ainternal.Con
                     **cast(dict, pc),
                 ),
             ) as pool:
-                yield cls(conn=pool, index=index)
+                yield cls(conn=pool, index=index, ttl=ttl)
         else:
             async with await AsyncConnection.connect(
                 conn_string, autocommit=True, prepare_threshold=0, row_factory=dict_row
             ) as conn:
                 if pipeline:
                     async with conn.pipeline() as pipe:
-                        yield cls(conn=conn, pipe=pipe, index=index)
+                        yield cls(conn=conn, pipe=pipe, index=index, ttl=ttl)
                 else:
-                    yield cls(conn=conn, index=index)
+                    yield cls(conn=conn, index=index, ttl=ttl)
 
     async def setup(self) -> None:
         """Set up the store database asynchronously.
@@ -256,6 +272,119 @@ class AsyncPostgresStore(AsyncBatchedBaseStore, BasePostgresStore[_ainternal.Con
                         "INSERT INTO vector_migrations (v) VALUES (%s)", (v,)
                     )
 
+    async def sweep_ttl(self) -> int:
+        """Delete expired store items based on TTL.
+
+        Returns:
+            int: The number of deleted items.
+        """
+        async with self._cursor() as cur:
+            await cur.execute(
+                """
+                DELETE FROM store
+                WHERE expires_at IS NOT NULL AND expires_at < NOW()
+                """
+            )
+            deleted_count = cur.rowcount
+            return deleted_count
+
+    async def start_ttl_sweeper(
+        self, sweep_interval_minutes: Optional[int] = None
+    ) -> asyncio.Task[None]:
+        """Periodically delete expired store items based on TTL.
+
+        Returns:
+            Task that can be awaited or cancelled.
+        """
+        if not self.ttl_config:
+            return asyncio.create_task(asyncio.sleep(0))
+
+        if self._ttl_sweeper_task is not None and not self._ttl_sweeper_task.done():
+            return self._ttl_sweeper_task
+
+        self._ttl_stop_event.clear()
+
+        interval = float(
+            sweep_interval_minutes or self.ttl_config.get("sweep_interval_minutes") or 5
+        )
+        logger.info(f"Starting store TTL sweeper with interval {interval} minutes")
+
+        async def _sweep_loop() -> None:
+            while not self._ttl_stop_event.is_set():
+                try:
+                    try:
+                        await asyncio.wait_for(
+                            self._ttl_stop_event.wait(),
+                            timeout=interval * 60,
+                        )
+                        break
+                    except asyncio.TimeoutError:
+                        pass
+
+                    expired_items = await self.sweep_ttl()
+                    if expired_items > 0:
+                        logger.info(f"Store swept {expired_items} expired items")
+                except asyncio.CancelledError:
+                    break
+                except Exception as exc:
+                    logger.exception("Store TTL sweep iteration failed", exc_info=exc)
+
+        task = asyncio.create_task(_sweep_loop())
+        task.set_name("ttl_sweeper")
+        self._ttl_sweeper_task = task
+        return task
+
+    async def stop_ttl_sweeper(self, timeout: Optional[float] = None) -> bool:
+        """Stop the TTL sweeper task if it's running.
+
+        Args:
+            timeout: Maximum time to wait for the task to stop, in seconds.
+                If None, wait indefinitely.
+
+        Returns:
+            bool: True if the task was successfully stopped or wasn't running,
+                False if the timeout was reached before the task stopped.
+        """
+        if self._ttl_sweeper_task is None or self._ttl_sweeper_task.done():
+            return True
+
+        logger.info("Stopping TTL sweeper task")
+        self._ttl_stop_event.set()
+
+        if timeout is not None:
+            try:
+                await asyncio.wait_for(self._ttl_sweeper_task, timeout=timeout)
+                success = True
+            except asyncio.TimeoutError:
+                success = False
+        else:
+            await self._ttl_sweeper_task
+            success = True
+
+        if success:
+            self._ttl_sweeper_task = None
+            logger.info("TTL sweeper task stopped")
+        else:
+            logger.warning("Timed out waiting for TTL sweeper task to stop")
+
+        return success
+
+    async def __aenter__(self) -> "AsyncPostgresStore":
+        return self
+
+    async def __aexit__(
+        self,
+        exc_type: Optional[type[BaseException]],
+        exc_val: Optional[BaseException],
+        exc_tb: Optional["TracebackType"],
+    ) -> None:
+        # Ensure the TTL sweeper task is stopped when exiting the context
+        if hasattr(self, "_ttl_sweeper_task") and self._ttl_sweeper_task is not None:
+            # Set the event to signal the task to stop
+            self._ttl_stop_event.set()
+            # We don't wait for the task to complete here to avoid blocking
+            # The task will clean up itself gracefully
+
     async def _execute_batch(
         self,
         grouped_ops: dict,
@@ -360,7 +489,7 @@ class AsyncPostgresStore(AsyncBatchedBaseStore, BasePostgresStore[_ainternal.Con
             for (idx, _), vector in zip(embedding_requests, vectors):
                 _paramslist = queries[idx][1]
                 for i in range(len(_paramslist)):
-                    if _paramslist[i] is _PLACEHOLDER:
+                    if _paramslist[i] is PLACEHOLDER:
                         _paramslist[i] = vector
 
         for (idx, _), (query, params) in zip(search_ops, queries):

libs/checkpoint-postgres/langgraph/store/postgres/base.py

@@ -1,4 +1,5 @@
 import asyncio
+import concurrent.futures
 import json
 import logging
 import threading
@@ -39,6 +40,7 @@ from langgraph.store.base import (
     Result,
     SearchItem,
     SearchOp,
+    TTLConfig,
     ensure_embeddings,
     get_text_at_path,
     tokenize_path,
@@ -73,6 +75,17 @@ CREATE TABLE IF NOT EXISTS store (
     """
 -- For faster lookups by prefix
 CREATE INDEX CONCURRENTLY IF NOT EXISTS store_prefix_idx ON store USING btree (prefix text_pattern_ops);
+""",
+    """
+-- Add expires_at column to store table
+ALTER TABLE store
+ADD COLUMN IF NOT EXISTS expires_at TIMESTAMP WITH TIME ZONE,
+ADD COLUMN IF NOT EXISTS ttl_minutes INT;
+""",
+    """
+-- Add indexes for efficient TTL sweeping
+CREATE INDEX IF NOT EXISTS idx_store_expires_at ON store (expires_at)
+WHERE expires_at IS NOT NULL;
 """,
 ]
 
@@ -224,20 +237,55 @@ class BasePostgresStore(Generic[C]):
         self,
         get_ops: Sequence[tuple[int, GetOp]],
     ) -> list[tuple[str, tuple, tuple[str, ...], list]]:
+        """
+        Build queries to fetch (and optionally refresh the TTL of) multiple keys per namespace.
+
+        Each returned element is a tuple of:
+        (sql_query_string, sql_params, namespace, items_for_this_namespace)
+
+        where items_for_this_namespace is the original list of (idx, key, refresh_ttl).
+        """
+
         namespace_groups = defaultdict(list)
+        refresh_ttls = defaultdict(list)
         for idx, op in get_ops:
             namespace_groups[op.namespace].append((idx, op.key))
+            refresh_ttls[op.namespace].append(op.refresh_ttl)
+
         results = []
         for namespace, items in namespace_groups.items():
             _, keys = zip(*items)
-            keys_to_query = ",".join(["%s"] * len(keys))
-            query = f"""
-                SELECT key, value, created_at, updated_at
-                FROM store
-                WHERE prefix = %s AND key IN ({keys_to_query})
+            this_refresh_ttls = refresh_ttls[namespace]
+
+            query = """
+                WITH passed_in AS (
+                    SELECT unnest(%s::text[]) AS key,
+                        unnest(%s::bool[])  AS do_refresh
+                ),
+                updated AS (
+                    UPDATE store s
+                    SET expires_at = NOW() + (s.ttl_minutes || ' minutes')::interval
+                    FROM passed_in p
+                    WHERE s.prefix = %s
+                    AND s.key    = p.key
+                    AND p.do_refresh = TRUE
+                    AND s.ttl_minutes IS NOT NULL
+                    RETURNING s.key
+                )
+                SELECT s.key, s.value, s.created_at, s.updated_at
+                FROM store s
+                JOIN passed_in p ON s.key = p.key
+                WHERE s.prefix = %s
             """
-            params = (_namespace_to_text(namespace), *keys)
+            ns_text = _namespace_to_text(namespace)
+            params = (
+                list(keys),  # -> unnest(%s::text[])
+                list(this_refresh_ttls),  # -> unnest(%s::bool[])
+                ns_text,  # -> prefix = %s (for UPDATE)
+                ns_text,  # -> prefix = %s (for final SELECT)
+            )
             results.append((query, params, namespace, items))
+
         return results
 
     def _prepare_batch_PUT_queries(
@@ -247,7 +295,6 @@ class BasePostgresStore(Generic[C]):
         list[tuple[str, Sequence]],
         Optional[tuple[str, Sequence[tuple[str, str, str, str]]]],
     ]:
-        # Last-write wins
         dedupped_ops: dict[tuple[tuple[str, ...], str], PutOp] = {}
         for _, op in put_ops:
             dedupped_ops[(op.namespace, op.key)] = op
@@ -281,15 +328,26 @@ class BasePostgresStore(Generic[C]):
             insertion_params = []
             vector_values = []
             embedding_request_params = []
+            # Handle TTL expiration
 
             # First handle main store insertions
             for op in inserts:
-                values.append("(%s, %s, %s, CURRENT_TIMESTAMP, CURRENT_TIMESTAMP)")
+                if op.ttl is not None:
+                    expires_at_str = f"NOW() + INTERVAL '{op.ttl*60} seconds'"
+                    ttl_minutes = op.ttl
+                else:
+                    expires_at_str = "NULL"
+                    ttl_minutes = None
+
+                values.append(
+                    f"(%s, %s, %s, CURRENT_TIMESTAMP, CURRENT_TIMESTAMP, {expires_at_str}, %s)"
+                )
                 insertion_params.extend(
                     [
                         _namespace_to_text(op.namespace),
                         op.key,
                         Jsonb(cast(dict, op.value)),
+                        ttl_minutes,
                     ]
                 )
 
@@ -303,7 +361,7 @@ class BasePostgresStore(Generic[C]):
                     k = op.key
 
                     if op.index is None:
-                        paths = self.index_config["__tokenized_fields"]
+                        paths = cast(dict, self.index_config)["__tokenized_fields"]
                     else:
                         paths = [(ix, tokenize_path(ix)) for ix in op.index]
 
@@ -318,11 +376,13 @@ class BasePostgresStore(Generic[C]):
 
             values_str = ",".join(values)
             query = f"""
-                INSERT INTO store (prefix, key, value, created_at, updated_at)
+                INSERT INTO store (prefix, key, value, created_at, updated_at, expires_at, ttl_minutes)
                 VALUES {values_str}
                 ON CONFLICT (prefix, key) DO UPDATE
                 SET value = EXCLUDED.value,
-                    updated_at = CURRENT_TIMESTAMP
+                    updated_at = CURRENT_TIMESTAMP,
+                    expires_at = EXCLUDED.expires_at,
+                    ttl_minutes = EXCLUDED.ttl_minutes
             """
             queries.append((query, insertion_params))
 
@@ -346,117 +406,151 @@ class BasePostgresStore(Generic[C]):
         list[tuple[str, list[Union[None, str, list[float]]]]],  # queries, params
         list[tuple[int, str]],  # idx, query_text pairs to embed
     ]:
+        """
+        Build per-SearchOp SQL queries (with optional TTL refresh) plus embedding requests.
+        Returns:
+        - queries: list of (SQL, param_list)
+        - embedding_requests: list of (original_index_in_search_ops, text_query)
+        """
+
         queries = []
         embedding_requests = []
-
         for idx, (_, op) in enumerate(search_ops):
-            # Build filter conditions first
             filter_params = []
-            filter_conditions = []
+            filter_clauses = []
             if op.filter:
                 for key, value in op.filter.items():
                     if isinstance(value, dict):
                         for op_name, val in value.items():
-                            condition, filter_params_ = self._get_filter_condition(
+                            condition, params_ = self._get_filter_condition(
                                 key, op_name, val
                             )
-                            filter_conditions.append(condition)
-                            filter_params.extend(filter_params_)
+                            filter_clauses.append(condition)
+                            filter_params.extend(params_)
                     else:
-                        filter_conditions.append("value->%s = %s::jsonb")
-                        filter_params.extend([key, json.dumps(value)])
+                        filter_clauses.append("value->%s = %s::jsonb")
+                        filter_params.extend([key, orjson.dumps(value).decode("utf-8")])
+
+            ns_condition = "TRUE"
+            ns_param: Optional[Sequence[Union[str]]] = None
+            if op.namespace_prefix:
+                ns_condition = "store.prefix LIKE %s"
+                ns_param = (f"{_namespace_to_text(op.namespace_prefix)}%",)
+            else:
+                ns_param = ()
+
+            extra_filters = (
+                " AND " + " AND ".join(filter_clauses) if filter_clauses else ""
+            )
 
-            # Vector search branch
             if op.query and self.index_config:
+                # We'll embed the text later, so record the request.
                 embedding_requests.append((idx, op.query))
 
-                score_operator, post_operator = _get_distance_operator(self)
+                score_operator, post_operator = get_distance_operator(self)
+                post_operator = post_operator.replace("scored", "uniq")
                 vector_type = (
                     cast(PostgresIndexConfig, self.index_config)
                     .get("ann_index_config", {})
                     .get("vector_type", "vector")
                 )
 
+                # For hamming bit vectors, or “regular” vectors
                 if (
                     vector_type == "bit"
-                    and self.index_config.get("distance_type") == "hamming"
+                    and cast(dict, self.index_config).get("distance_type") == "hamming"
                 ):
                     score_operator = score_operator % (
                         "%s",
-                        self.index_config["dims"],
+                        cast(dict, self.index_config)["dims"],
                     )
                 else:
-                    score_operator = score_operator % (
-                        "%s",
-                        vector_type,
-                    )
+                    score_operator = score_operator % ("%s", vector_type)
 
-                vectors_per_doc_estimate = self.index_config["__estimated_num_vectors"]
+                vectors_per_doc_estimate = cast(dict, self.index_config)[
+                    "__estimated_num_vectors"
+                ]
                 expanded_limit = (op.limit * vectors_per_doc_estimate * 2) + 1
 
-                # Vector search with CTE for proper score handling
-                filter_str = (
-                    ""
-                    if not filter_conditions
-                    else " AND " + " AND ".join(filter_conditions)
-                )
-                if op.namespace_prefix:
-                    prefix_filter_str = f"WHERE s.prefix LIKE %s {filter_str} "
-                    ns_args: Sequence = (f"{_namespace_to_text(op.namespace_prefix)}%",)
-                else:
-                    ns_args = ()
-                    if filter_str:
-                        prefix_filter_str = f"WHERE {filter_str} "
-                    else:
-                        prefix_filter_str = ""
-
-                base_query = f"""
-                    WITH scored AS (
-                        SELECT s.prefix, s.key, s.value, s.created_at, s.updated_at, {score_operator} AS neg_score
-                        FROM store s
-                        JOIN store_vectors sv ON s.prefix = sv.prefix AND s.key = sv.key
-                        {prefix_filter_str}
-                        ORDER BY {score_operator} ASC 
+                # “sub_scored” does the main vector search
+                # Then we do DISTINCT ON to drop duplicates if your store can have them
+                # Finally we limit & offset
+                vector_search_cte = f"""
+                        SELECT store.prefix, store.key, store.value, store.created_at, store.updated_at,
+                            {score_operator} AS neg_score
+                        FROM store
+                        JOIN store_vectors sv ON store.prefix = sv.prefix AND store.key = sv.key
+                        WHERE {ns_condition} {extra_filters}
+                        ORDER BY {score_operator} ASC
                         LIMIT %s
-                    )
-                    SELECT * FROM (
-                        SELECT DISTINCT ON (prefix, key) 
-                            prefix, key, value, created_at, updated_at, {post_operator} as score 
-                        FROM scored 
-                        ORDER BY prefix, key, score DESC
-                    ) AS unique_docs
-                    ORDER BY score DESC
-                    LIMIT %s
-                    OFFSET %s
-                """
-                params = [
-                    _PLACEHOLDER,  # Vector placeholder
-                    *ns_args,
+                    """
+
+                search_results_sql = f"""
+                        WITH scored AS (
+                            {vector_search_cte}
+                        )
+                        SELECT uniq.prefix, uniq.key, uniq.value, uniq.created_at, uniq.updated_at,
+                            {post_operator} AS score
+                        FROM (
+                            SELECT DISTINCT ON (scored.prefix, scored.key)
+                                scored.prefix, scored.key, scored.value, scored.created_at, scored.updated_at, scored.neg_score
+                            FROM scored
+                            ORDER BY scored.prefix, scored.key, scored.neg_score ASC
+                        ) uniq
+                        ORDER BY score DESC
+                        LIMIT %s
+                        OFFSET %s
+                    """
+
+                search_results_params = [
+                    PLACEHOLDER,
+                    *ns_param,
                     *filter_params,
-                    _PLACEHOLDER,
+                    PLACEHOLDER,
                     expanded_limit,
                     op.limit,
                     op.offset,
                 ]
 
-            # Regular search branch
             else:
-                base_query = """
-                    SELECT prefix, key, value, created_at, updated_at
-                    FROM store
-                    WHERE prefix LIKE %s
-                """
-                params = [f"{_namespace_to_text(op.namespace_prefix)}%"]
+                base_query = f"""
+                        SELECT store.prefix, store.key, store.value, store.created_at, store.updated_at, NULL AS score
+                        FROM store
+                        WHERE {ns_condition} {extra_filters}
+                        ORDER BY store.updated_at DESC
+                        LIMIT %s
+                        OFFSET %s
+                    """
+                search_results_sql = base_query
+                search_results_params = [
+                    *ns_param,
+                    *filter_params,
+                    op.limit,
+                    op.offset,
+                ]
 
-                if filter_conditions:
-                    params.extend(filter_params)
-                    base_query += " AND " + " AND ".join(filter_conditions)
-
-                base_query += " ORDER BY updated_at DESC"
-                base_query += " LIMIT %s OFFSET %s"
-                params.extend([op.limit, op.offset])
-
-            queries.append((base_query, params))
+            if op.refresh_ttl:
+                # Wrap entire primary query in a CTE, then perform "update_at"
+                final_sql = f"""
+                        WITH search_results AS (
+                            {search_results_sql}
+                        ),
+                        updated AS (
+                            UPDATE store s
+                            SET expires_at = NOW() + (s.ttl_minutes || ' minutes')::interval
+                            FROM search_results sr
+                            WHERE s.prefix = sr.prefix
+                            AND s.key = sr.key
+                            AND s.ttl_minutes IS NOT NULL
+                        )
+                        SELECT sr.prefix, sr.key, sr.value, sr.created_at, sr.updated_at, sr.score
+                        FROM search_results sr
+                    """
+                final_params = search_results_params[:]  # copy
+            else:
+                final_sql = search_results_sql
+                final_params = search_results_params
+            queries.append((final_sql, final_params))
 
         return queries, embedding_requests
 
@@ -602,6 +696,11 @@ class PostgresStore(BaseStore, BasePostgresStore[_pg_internal.Conn]):
         Make sure to call `setup()` before first use to create necessary tables and indexes.
         The pgvector extension must be available to use vector search.
 
+    Note:
+        If you provide a TTL configuration, you must explicitly call `start_ttl_sweeper()` to begin
+        the background thread that removes expired items. Call `stop_ttl_sweeper()` to properly
+        clean up resources when you're done with the store.
+
     """
 
     __slots__ = (
@@ -611,7 +710,10 @@ class PostgresStore(BaseStore, BasePostgresStore[_pg_internal.Conn]):
         "supports_pipeline",
         "index_config",
         "embeddings",
+        "_ttl_sweeper_thread",
+        "_ttl_stop_event",
     )
+    supports_ttl: bool = True
 
     def __init__(
         self,
@@ -622,6 +724,7 @@ class PostgresStore(BaseStore, BasePostgresStore[_pg_internal.Conn]):
             Callable[[Union[bytes, orjson.Fragment]], dict[str, Any]]
         ] = None,
         index: Optional[PostgresIndexConfig] = None,
+        ttl: Optional[TTLConfig] = None,
     ) -> None:
         super().__init__()
         self._deserializer = deserializer
@@ -634,6 +737,9 @@ class PostgresStore(BaseStore, BasePostgresStore[_pg_internal.Conn]):
             self.embeddings, self.index_config = _ensure_index_config(self.index_config)
         else:
             self.embeddings = None
+        self.ttl_config = ttl
+        self._ttl_sweeper_thread: Optional[threading.Thread] = None
+        self._ttl_stop_event = threading.Event()
 
     @classmethod
     @contextmanager
@@ -644,6 +750,7 @@ class PostgresStore(BaseStore, BasePostgresStore[_pg_internal.Conn]):
         pipeline: bool = False,
         pool_config: Optional[PoolConfig] = None,
         index: Optional[PostgresIndexConfig] = None,
+        ttl: Optional[TTLConfig] = None,
     ) -> Iterator["PostgresStore"]:
         """Create a new PostgresStore instance from a connection string.
 
@@ -675,16 +782,123 @@ class PostgresStore(BaseStore, BasePostgresStore[_pg_internal.Conn]):
                     **cast(dict, pc),
                 ),
             ) as pool:
-                yield cls(conn=pool, index=index)
+                yield cls(conn=pool, index=index, ttl=ttl)
         else:
             with Connection.connect(
                 conn_string, autocommit=True, prepare_threshold=0, row_factory=dict_row
             ) as conn:
                 if pipeline:
                     with conn.pipeline() as pipe:
-                        yield cls(conn, pipe=pipe, index=index)
+                        yield cls(conn, pipe=pipe, index=index, ttl=ttl)
                 else:
-                    yield cls(conn, index=index)
+                    yield cls(conn, index=index, ttl=ttl)
+
+    def sweep_ttl(self) -> int:
+        """Delete expired store items based on TTL.
+
+        Returns:
+            int: The number of deleted items.
+        """
+        with self._cursor() as cur:
+            cur.execute(
+                """
+                DELETE FROM store
+                WHERE expires_at IS NOT NULL AND expires_at < NOW()
+                """
+            )
+            deleted_count = cur.rowcount
+            return deleted_count
+
+    def start_ttl_sweeper(
+        self, sweep_interval_minutes: Optional[int] = None
+    ) -> concurrent.futures.Future[None]:
+        """Periodically delete expired store items based on TTL.
+
+        Returns:
+            Future that can be waited on or cancelled.
+        """
+        if not self.ttl_config:
+            future: concurrent.futures.Future[None] = concurrent.futures.Future()
+            future.set_result(None)
+            return future
+
+        if self._ttl_sweeper_thread and self._ttl_sweeper_thread.is_alive():
+            logger.info("TTL sweeper thread is already running")
+            # Return a future that can be used to cancel the existing thread
+            future = concurrent.futures.Future()
+            future.add_done_callback(
+                lambda f: self._ttl_stop_event.set() if f.cancelled() else None
+            )
+            return future
+
+        self._ttl_stop_event.clear()
+
+        interval = float(
+            sweep_interval_minutes or self.ttl_config.get("sweep_interval_minutes") or 5
+        )
+        logger.info(f"Starting store TTL sweeper with interval {interval} minutes")
+
+        future = concurrent.futures.Future()
+
+        def _sweep_loop() -> None:
+            try:
+                while not self._ttl_stop_event.is_set():
+                    if self._ttl_stop_event.wait(interval * 60):
+                        break
+
+                    try:
+                        expired_items = self.sweep_ttl()
+                        if expired_items > 0:
+                            logger.info(f"Store swept {expired_items} expired items")
+                    except Exception as exc:
+                        logger.exception(
+                            "Store TTL sweep iteration failed", exc_info=exc
+                        )
+                future.set_result(None)
+            except Exception as exc:
+                future.set_exception(exc)
+
+        thread = threading.Thread(target=_sweep_loop, daemon=True, name="ttl-sweeper")
+        self._ttl_sweeper_thread = thread
+        thread.start()
+
+        future.add_done_callback(
+            lambda f: self._ttl_stop_event.set() if f.cancelled() else None
+        )
+        return future
+
+    def stop_ttl_sweeper(self, timeout: Optional[float] = None) -> bool:
+        """Stop the TTL sweeper thread if it's running.
+
+        Args:
+            timeout: Maximum time to wait for the thread to stop, in seconds.
+                If None, wait indefinitely.
+
+        Returns:
+            bool: True if the thread was successfully stopped or wasn't running,
+                False if the timeout was reached before the thread stopped.
+        """
+        if not self._ttl_sweeper_thread or not self._ttl_sweeper_thread.is_alive():
+            return True
+
+        logger.info("Stopping TTL sweeper thread")
+        self._ttl_stop_event.set()
+
+        self._ttl_sweeper_thread.join(timeout)
+        success = not self._ttl_sweeper_thread.is_alive()
+
+        if success:
+            self._ttl_sweeper_thread = None
+            logger.info("TTL sweeper thread stopped")
+        else:
+            logger.warning("Timed out waiting for TTL sweeper thread to stop")
+
+        return success
+
+    def __del__(self) -> None:
+        """Ensure the TTL sweeper thread is stopped when the object is garbage collected."""
+        if hasattr(self, "_ttl_stop_event") and hasattr(self, "_ttl_sweeper_thread"):
+            self.stop_ttl_sweeper(timeout=0.1)
 
     @contextmanager
     def _cursor(self, *, pipeline: bool = False) -> Iterator[Cursor[DictRow]]:
@@ -828,7 +1042,7 @@ class PostgresStore(BaseStore, BasePostgresStore[_pg_internal.Conn]):
             for (idx, _), embedding in zip(embedding_requests, embeddings):
                 _paramslist = queries[idx][1]
                 for i in range(len(_paramslist)):
-                    if _paramslist[i] is _PLACEHOLDER:
+                    if _paramslist[i] is PLACEHOLDER:
                         _paramslist[i] = embedding
 
         for (idx, _), (query, params) in zip(search_ops, queries):
@@ -883,8 +1097,14 @@ class PostgresStore(BaseStore, BasePostgresStore[_pg_internal.Conn]):
         with self._cursor() as cur:
             version = _get_version(cur, table="store_migrations")
             for v, sql in enumerate(self.MIGRATIONS[version + 1 :], start=version + 1):
-                cur.execute(sql)
-                cur.execute("INSERT INTO store_migrations (v) VALUES (%s)", (v,))
+                try:
+                    cur.execute(sql)
+                    cur.execute("INSERT INTO store_migrations (v) VALUES (%s)", (v,))
+                except Exception as e:
+                    logger.error(
+                        f"Failed to apply migration {v}.\nSql={sql}\nError={e}"
+                    )
+                    raise
 
             if self.index_config:
                 version = _get_version(cur, table="vector_migrations")
@@ -1055,7 +1275,7 @@ def _decode_ns_bytes(namespace: Union[str, bytes, list]) -> tuple[str, ...]:
     return tuple(namespace.split("."))
 
 
-def _get_distance_operator(store: Any) -> tuple[str, str]:
+def get_distance_operator(store: Any) -> tuple[str, str]:
     """Get the distance operator and score expression based on config."""
     # Note: Today, we are not using ANN indices due to restrictions
     # on PGVector's support for mixing vector and non-vector filters
@@ -1121,4 +1341,4 @@ def _ensure_index_config(
     return embeddings, index_config
 
 
-_PLACEHOLDER = object()
+PLACEHOLDER = object()

libs/checkpoint-postgres/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langgraph-checkpoint-postgres"
-version = "2.0.15"
+version = "2.0.19"
 description = "Library with a Postgres implementation of LangGraph checkpoint saver."
 authors = []
 license = "MIT"
@@ -10,7 +10,7 @@ packages = [{ include = "langgraph" }]
 
 [tool.poetry.dependencies]
 python = "^3.9.0,<4.0"
-langgraph-checkpoint = "^2.0.15"
+langgraph-checkpoint = "^2.0.21"
 orjson = ">=3.10.1"
 psycopg = "^3.2.0"
 psycopg-pool = "^3.2.0"

libs/checkpoint-postgres/tests/test_async_store.py

@@ -26,6 +26,9 @@ from tests.conftest import (
     CharacterEmbeddings,
 )
 
+TTL_SECONDS = 6
+TTL_MINUTES = TTL_SECONDS / 60
+
 
 @pytest.fixture(scope="function", params=["default", "pipe", "pool"])
 async def store(request) -> AsyncIterator[AsyncPostgresStore]:
@@ -42,28 +45,54 @@ async def store(request) -> AsyncIterator[AsyncPostgresStore]:
 
     conn_string = f"{uri_base}/{database}{query_params}"
     admin_conn_string = DEFAULT_URI
-
+    ttl_config = {
+        "default_ttl": TTL_MINUTES,
+        "refresh_on_read": True,
+        "sweep_interval_minutes": TTL_MINUTES / 2,
+    }
     async with await AsyncConnection.connect(
         admin_conn_string, autocommit=True
     ) as conn:
         await conn.execute(f"CREATE DATABASE {database}")
     try:
-        async with AsyncPostgresStore.from_conn_string(conn_string) as store:
+        async with AsyncPostgresStore.from_conn_string(
+            conn_string, ttl=ttl_config
+        ) as store:
+            store.MIGRATIONS = [
+                (
+                    mig.replace("ttl_minutes INT;", "ttl_minutes FLOAT;")
+                    if isinstance(mig, str)
+                    else mig
+                )
+                for mig in store.MIGRATIONS
+            ]
             await store.setup()
+            async with store._cursor() as cur:
+                # drop the migration index
+                await cur.execute("DROP TABLE IF EXISTS store_migrations")
+            await store.setup()  # Will fail if migrations aren't idempotent
 
         if request.param == "pipe":
             async with AsyncPostgresStore.from_conn_string(
-                conn_string, pipeline=True
+                conn_string, pipeline=True, ttl=ttl_config
             ) as store:
+                await store.start_ttl_sweeper()
                 yield store
+                await store.stop_ttl_sweeper()
         elif request.param == "pool":
             async with AsyncPostgresStore.from_conn_string(
-                conn_string, pool_config={"min_size": 1, "max_size": 10}
+                conn_string, pool_config={"min_size": 1, "max_size": 10}, ttl=ttl_config
             ) as store:
+                await store.start_ttl_sweeper()
                 yield store
+                await store.stop_ttl_sweeper()
         else:  # default
-            async with AsyncPostgresStore.from_conn_string(conn_string) as store:
+            async with AsyncPostgresStore.from_conn_string(
+                conn_string, ttl=ttl_config
+            ) as store:
+                await store.start_ttl_sweeper()
                 yield store
+                await store.stop_ttl_sweeper()
     finally:
         async with await AsyncConnection.connect(
             admin_conn_string, autocommit=True
@@ -635,3 +664,28 @@ async def test_search_sorting(
         assert len(set(r.key for r in results)) == 10
         assert results[0].key == "M"
         assert results[0].score > results[1].score
+
+
+async def test_store_ttl(store):
+    # Assumes a TTL of 1 minute = 60 seconds
+    ns = ("foo",)
+    await store.start_ttl_sweeper()
+    await store.aput(
+        ns,
+        key="item1",
+        value={"foo": "bar"},
+        ttl=TTL_MINUTES,  # type: ignore
+    )
+    await asyncio.sleep(TTL_SECONDS - 2)
+    res = await store.aget(ns, key="item1", refresh_ttl=True)
+    assert res is not None
+    await asyncio.sleep(TTL_SECONDS - 2)
+    results = await store.asearch(ns, query="foo", refresh_ttl=True)
+    assert len(results) == 1
+    await asyncio.sleep(TTL_SECONDS - 2)
+    res = await store.aget(ns, key="item1", refresh_ttl=False)
+    assert res is not None
+    await asyncio.sleep(TTL_SECONDS - 1)
+    # Now has been (TTL_SECONDS-2)*2 > TTL_SECONDS + TTL_SECONDS/2
+    results = await store.asearch(ns, query="bar", refresh_ttl=False)
+    assert len(results) == 0

libs/checkpoint-postgres/tests/test_store.py

@@ -1,6 +1,7 @@
 # type: ignore
 
 import re
+import time
 from contextlib import contextmanager
 from typing import Any, Optional
 from uuid import uuid4
@@ -24,6 +25,9 @@ from tests.conftest import (
     CharacterEmbeddings,
 )
 
+TTL_SECONDS = 6
+TTL_MINUTES = TTL_SECONDS / 60
+
 
 @pytest.fixture(scope="function", params=["default", "pipe", "pool"])
 def store(request) -> PostgresStore:
@@ -32,29 +36,56 @@ def store(request) -> PostgresStore:
     uri_base = "/".join(uri_parts[:-1])
     query_params = ""
     if "?" in uri_parts[-1]:
-        db_name, query_params = uri_parts[-1].split("?", 1)
+        _, query_params = uri_parts[-1].split("?", 1)
         query_params = "?" + query_params
 
     conn_string = f"{uri_base}/{database}{query_params}"
     admin_conn_string = DEFAULT_URI
-
+    ttl_config = {
+        "default_ttl": TTL_MINUTES,
+        "refresh_on_read": True,
+        "sweep_interval_minutes": TTL_MINUTES / 2,
+    }
     with Connection.connect(admin_conn_string, autocommit=True) as conn:
         conn.execute(f"CREATE DATABASE {database}")
     try:
-        with PostgresStore.from_conn_string(conn_string) as store:
+        with PostgresStore.from_conn_string(conn_string, ttl=ttl_config) as store:
+            store.MIGRATIONS = [
+                (
+                    mig.replace("ttl_minutes INT;", "ttl_minutes FLOAT;")
+                    if isinstance(mig, str)
+                    else mig
+                )
+                for mig in store.MIGRATIONS
+            ]
             store.setup()
 
         if request.param == "pipe":
-            with PostgresStore.from_conn_string(conn_string, pipeline=True) as store:
+            with PostgresStore.from_conn_string(
+                conn_string,
+                pipeline=True,
+                ttl=ttl_config,
+            ) as store:
+                store.start_ttl_sweeper()
                 yield store
+
+                store.stop_ttl_sweeper()
         elif request.param == "pool":
             with PostgresStore.from_conn_string(
-                conn_string, pool_config={"min_size": 1, "max_size": 10}
+                conn_string,
+                pool_config={"min_size": 1, "max_size": 10},
+                ttl=ttl_config,
             ) as store:
+                store.start_ttl_sweeper()
                 yield store
+
+                store.stop_ttl_sweeper()
         else:  # default
-            with PostgresStore.from_conn_string(conn_string) as store:
+            with PostgresStore.from_conn_string(conn_string, ttl=ttl_config) as store:
+                store.start_ttl_sweeper()
                 yield store
+
+                store.stop_ttl_sweeper()
     finally:
         with Connection.connect(admin_conn_string, autocommit=True) as conn:
             conn.execute(f"DROP DATABASE {database}")
@@ -220,134 +251,127 @@ def test_batch_list_namespaces_ops(store: PostgresStore) -> None:
     assert all(ns[-1] == "public" for ns in results[2])
 
 
-class TestPostgresStore:
-    @pytest.fixture(autouse=True)
-    def setup(self) -> None:
-        with PostgresStore.from_conn_string(DEFAULT_URI) as store:
-            store.setup()
+def test_basic_store_ops(store) -> None:
+    namespace = ("test", "documents")
+    item_id = "doc1"
+    item_value = {"title": "Test Document", "content": "Hello, World!"}
 
-    def test_basic_store_ops(self) -> None:
-        with PostgresStore.from_conn_string(DEFAULT_URI) as store:
-            namespace = ("test", "documents")
-            item_id = "doc1"
-            item_value = {"title": "Test Document", "content": "Hello, World!"}
+    store.put(namespace, item_id, item_value)
+    item = store.get(namespace, item_id)
 
-            store.put(namespace, item_id, item_value)
-            item = store.get(namespace, item_id)
+    assert item
+    assert item.namespace == namespace
+    assert item.key == item_id
+    assert item.value == item_value
 
-            assert item
-            assert item.namespace == namespace
-            assert item.key == item_id
-            assert item.value == item_value
+    # Test update
+    updated_value = {"title": "Updated Document", "content": "Hello, Updated!"}
+    store.put(namespace, item_id, updated_value)
+    updated_item = store.get(namespace, item_id)
 
-            # Test update
-            updated_value = {"title": "Updated Document", "content": "Hello, Updated!"}
-            store.put(namespace, item_id, updated_value)
-            updated_item = store.get(namespace, item_id)
+    assert updated_item.value == updated_value
+    assert updated_item.updated_at > item.updated_at
 
-            assert updated_item.value == updated_value
-            assert updated_item.updated_at > item.updated_at
+    # Test get from non-existent namespace
+    different_namespace = ("test", "other_documents")
+    item_in_different_namespace = store.get(different_namespace, item_id)
+    assert item_in_different_namespace is None
 
-            # Test get from non-existent namespace
-            different_namespace = ("test", "other_documents")
-            item_in_different_namespace = store.get(different_namespace, item_id)
-            assert item_in_different_namespace is None
+    # Test delete
+    store.delete(namespace, item_id)
+    deleted_item = store.get(namespace, item_id)
+    assert deleted_item is None
 
-            # Test delete
-            store.delete(namespace, item_id)
-            deleted_item = store.get(namespace, item_id)
-            assert deleted_item is None
 
-    def test_list_namespaces(self) -> None:
-        with PostgresStore.from_conn_string(DEFAULT_URI) as store:
-            # Create test data with various namespaces
-            test_namespaces = [
-                ("test", "documents", "public"),
-                ("test", "documents", "private"),
-                ("test", "images", "public"),
-                ("test", "images", "private"),
-                ("prod", "documents", "public"),
-                ("prod", "documents", "private"),
-            ]
+def test_list_namespaces(store) -> None:
+    # Create test data with various namespaces
+    test_namespaces = [
+        ("test", "documents", "public"),
+        ("test", "documents", "private"),
+        ("test", "images", "public"),
+        ("test", "images", "private"),
+        ("prod", "documents", "public"),
+        ("prod", "documents", "private"),
+    ]
 
-            # Insert test data
-            for namespace in test_namespaces:
-                store.put(namespace, "dummy", {"content": "dummy"})
+    # Insert test data
+    for namespace in test_namespaces:
+        store.put(namespace, "dummy", {"content": "dummy"})
 
-            # Test listing with various filters
-            all_namespaces = store.list_namespaces()
-            assert len(all_namespaces) == len(test_namespaces)
+    # Test listing with various filters
+    all_namespaces = store.list_namespaces()
+    assert len(all_namespaces) == len(test_namespaces)
 
-            # Test prefix filtering
-            test_prefix_namespaces = store.list_namespaces(prefix=["test"])
-            assert len(test_prefix_namespaces) == 4
-            assert all(ns[0] == "test" for ns in test_prefix_namespaces)
+    # Test prefix filtering
+    test_prefix_namespaces = store.list_namespaces(prefix=["test"])
+    assert len(test_prefix_namespaces) == 4
+    assert all(ns[0] == "test" for ns in test_prefix_namespaces)
 
-            # Test suffix filtering
-            public_namespaces = store.list_namespaces(suffix=["public"])
-            assert len(public_namespaces) == 3
-            assert all(ns[-1] == "public" for ns in public_namespaces)
+    # Test suffix filtering
+    public_namespaces = store.list_namespaces(suffix=["public"])
+    assert len(public_namespaces) == 3
+    assert all(ns[-1] == "public" for ns in public_namespaces)
 
-            # Test max depth
-            depth_2_namespaces = store.list_namespaces(max_depth=2)
-            assert all(len(ns) <= 2 for ns in depth_2_namespaces)
+    # Test max depth
+    depth_2_namespaces = store.list_namespaces(max_depth=2)
+    assert all(len(ns) <= 2 for ns in depth_2_namespaces)
 
-            # Test pagination
-            paginated_namespaces = store.list_namespaces(limit=3)
-            assert len(paginated_namespaces) == 3
+    # Test pagination
+    paginated_namespaces = store.list_namespaces(limit=3)
+    assert len(paginated_namespaces) == 3
 
-            # Cleanup
-            for namespace in test_namespaces:
-                store.delete(namespace, "dummy")
+    # Cleanup
+    for namespace in test_namespaces:
+        store.delete(namespace, "dummy")
 
-    def test_search(self) -> None:
-        with PostgresStore.from_conn_string(DEFAULT_URI) as store:
-            # Create test data
-            test_data = [
-                (
-                    ("test", "docs"),
-                    "doc1",
-                    {"title": "First Doc", "author": "Alice", "tags": ["important"]},
-                ),
-                (
-                    ("test", "docs"),
-                    "doc2",
-                    {"title": "Second Doc", "author": "Bob", "tags": ["draft"]},
-                ),
-                (
-                    ("test", "images"),
-                    "img1",
-                    {"title": "Image 1", "author": "Alice", "tags": ["final"]},
-                ),
-            ]
 
-            for namespace, key, value in test_data:
-                store.put(namespace, key, value)
+def test_search(store) -> None:
+    # Create test data
+    test_data = [
+        (
+            ("test", "docs"),
+            "doc1",
+            {"title": "First Doc", "author": "Alice", "tags": ["important"]},
+        ),
+        (
+            ("test", "docs"),
+            "doc2",
+            {"title": "Second Doc", "author": "Bob", "tags": ["draft"]},
+        ),
+        (
+            ("test", "images"),
+            "img1",
+            {"title": "Image 1", "author": "Alice", "tags": ["final"]},
+        ),
+    ]
 
-            # Test basic search
-            all_items = store.search(["test"])
-            assert len(all_items) == 3
+    for namespace, key, value in test_data:
+        store.put(namespace, key, value)
 
-            # Test namespace filtering
-            docs_items = store.search(["test", "docs"])
-            assert len(docs_items) == 2
-            assert all(item.namespace == ("test", "docs") for item in docs_items)
+    # Test basic search
+    all_items = store.search(["test"])
+    assert len(all_items) == 3
 
-            # Test value filtering
-            alice_items = store.search(["test"], filter={"author": "Alice"})
-            assert len(alice_items) == 2
-            assert all(item.value["author"] == "Alice" for item in alice_items)
+    # Test namespace filtering
+    docs_items = store.search(["test", "docs"])
+    assert len(docs_items) == 2
+    assert all(item.namespace == ("test", "docs") for item in docs_items)
 
-            # Test pagination
-            paginated_items = store.search(["test"], limit=2)
-            assert len(paginated_items) == 2
+    # Test value filtering
+    alice_items = store.search(["test"], filter={"author": "Alice"})
+    assert len(alice_items) == 2
+    assert all(item.value["author"] == "Alice" for item in alice_items)
 
-            offset_items = store.search(["test"], offset=2)
-            assert len(offset_items) == 1
+    # Test pagination
+    paginated_items = store.search(["test"], limit=2)
+    assert len(paginated_items) == 2
 
-            # Cleanup
-            for namespace, key, _ in test_data:
-                store.delete(namespace, key)
+    offset_items = store.search(["test"], offset=2)
+    assert len(offset_items) == 1
+
+    # Cleanup
+    for namespace, key, _ in test_data:
+        store.delete(namespace, key)
 
 
 @contextmanager
@@ -356,6 +380,7 @@ def _create_vector_store(
     distance_type: str,
     fake_embeddings: Embeddings,
     text_fields: Optional[list[str]] = None,
+    enable_ttl: bool = True,
 ) -> PostgresStore:
     """Create a store with vector search enabled."""
     database = f"test_{uuid4().hex[:16]}"
@@ -385,23 +410,32 @@ def _create_vector_store(
         with PostgresStore.from_conn_string(
             conn_string,
             index=index_config,
+            ttl={"default_ttl": 2, "refresh_on_read": True} if enable_ttl else None,
         ) as store:
             store.setup()
+            with store._cursor() as cur:
+                # drop the migration index
+                cur.execute("DROP TABLE IF EXISTS store_migrations")
+            store.setup()  # Will fail if migrations aren't idempotent
             yield store
     finally:
         with Connection.connect(admin_conn_string, autocommit=True) as conn:
             conn.execute(f"DROP DATABASE {database}")
 
 
+_vector_params = [
+    (vector_type, distance_type, True)
+    for vector_type in VECTOR_TYPES
+    for distance_type in (
+        ["hamming"] if vector_type == "bit" else ["l2", "inner_product", "cosine"]
+    )
+]
+_vector_params += [(*_vector_params[-1][:2], False)]
+
+
 @pytest.fixture(
     scope="function",
-    params=[
-        (vector_type, distance_type)
-        for vector_type in VECTOR_TYPES
-        for distance_type in (
-            ["hamming"] if vector_type == "bit" else ["l2", "inner_product", "cosine"]
-        )
-    ],
+    params=_vector_params,
     ids=lambda p: f"{p[0]}_{p[1]}",
 )
 def vector_store(
@@ -409,8 +443,10 @@ def vector_store(
     fake_embeddings: Embeddings,
 ) -> PostgresStore:
     """Create a store with vector search enabled."""
-    vector_type, distance_type = request.param
-    with _create_vector_store(vector_type, distance_type, fake_embeddings) as store:
+    vector_type, distance_type, enable_ttl = request.param
+    with _create_vector_store(
+        vector_type, distance_type, fake_embeddings, enable_ttl=enable_ttl
+    ) as store:
         yield store
 
 
@@ -474,7 +510,10 @@ def test_vector_update_with_embedding(vector_store: PostgresStore) -> None:
     assert not any(r.key == "doc4" for r in results_new)
 
 
-def test_vector_search_with_filters(vector_store: PostgresStore) -> None:
+@pytest.mark.parametrize("refresh_ttl", [True, False])
+def test_vector_search_with_filters(
+    vector_store: PostgresStore, refresh_ttl: bool
+) -> None:
     """Test combining vector search with filters."""
     # Insert test documents
     docs = [
@@ -487,16 +526,23 @@ def test_vector_search_with_filters(vector_store: PostgresStore) -> None:
     for key, value in docs:
         vector_store.put(("test",), key, value)
 
-    results = vector_store.search(("test",), query="apple", filter={"color": "red"})
+    results = vector_store.search(
+        ("test",), query="apple", filter={"color": "red"}, refresh_ttl=refresh_ttl
+    )
     assert len(results) == 2
     assert results[0].key == "doc1"
 
-    results = vector_store.search(("test",), query="car", filter={"color": "red"})
+    results = vector_store.search(
+        ("test",), query="car", filter={"color": "red"}, refresh_ttl=refresh_ttl
+    )
     assert len(results) == 2
     assert results[0].key == "doc2"
 
     results = vector_store.search(
-        ("test",), query="bbbbluuu", filter={"score": {"$gt": 3.2}}
+        ("test",),
+        query="bbbbluuu",
+        filter={"score": {"$gt": 3.2}},
+        refresh_ttl=refresh_ttl,
     )
     assert len(results) == 3
     assert results[0].key == "doc4"
@@ -688,7 +734,7 @@ def test_embed_with_path_operation_config(
         store.put(("test",), "doc5", doc5, index=False)
         results = store.search(("test",))
         assert len(results) == 3
-        assert all(r.score is None for r in results)
+        assert all(r.score is None for r in results), f"{results}"
         assert any(r.key == "doc5" for r in results)
 
         results = store.search(("test",), query="hhh")
@@ -790,3 +836,27 @@ def test_nonnull_migrations() -> None:
     for migration in PostgresStore.MIGRATIONS:
         statement = _leading_comment_remover.sub("", migration).split()[0]
         assert statement.strip()
+
+
+def test_store_ttl(store):
+    # Assumes a TTL of 1 minute = 60 seconds
+    ns = ("foo",)
+    store.put(
+        ns,
+        key="item1",
+        value={"foo": "bar"},
+        ttl=TTL_MINUTES,  # type: ignore
+    )
+    time.sleep(TTL_SECONDS - 2)
+    res = store.get(ns, key="item1", refresh_ttl=True)
+    assert res is not None
+    time.sleep(TTL_SECONDS - 2)
+    results = store.search(ns, query="foo", refresh_ttl=True)
+    assert len(results) == 1
+    time.sleep(TTL_SECONDS - 2)
+    res = store.get(ns, key="item1", refresh_ttl=False)
+    assert res is not None
+    time.sleep(TTL_SECONDS - 1)
+    # Now has been (TTL_SECONDS-2)*2 > TTL_SECONDS + TTL_SECONDS/2
+    res = store.search(ns, query="bar", refresh_ttl=False)
+    assert len(res) == 0

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

@@ -56,7 +56,10 @@ class SqliteSaver(BaseCheckpointSaver[str]):
         >>> builder.add_node("add_one", lambda x: x + 1)
         >>> builder.set_entry_point("add_one")
         >>> builder.set_finish_point("add_one")
-        >>> conn = sqlite3.connect("checkpoints.sqlite")
+        >>> # Create a new SqliteSaver instance
+        >>> # Note: check_same_thread=False is OK as the implementation uses a lock
+        >>> # to ensure thread safety.
+        >>> conn = sqlite3.connect("checkpoints.sqlite", check_same_thread=False)
         >>> memory = SqliteSaver(conn)
         >>> graph = builder.compile(checkpointer=memory)
         >>> config = {"configurable": {"thread_id": "1"}}

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

@@ -530,6 +530,7 @@ class AsyncSqliteSaver(BaseCheckpointSaver[str]):
                     for idx, (channel, value) in enumerate(writes)
                 ],
             )
+            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-sqlite/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langgraph-checkpoint-sqlite"
-version = "2.0.5"
+version = "2.0.6"
 description = "Library with a SQLite implementation of LangGraph checkpoint saver."
 authors = []
 license = "MIT"

libs/checkpoint/langgraph/checkpoint/serde/base.py

@@ -45,3 +45,18 @@ def maybe_add_typed_methods(serde: SerializerProtocol) -> SerializerProtocol:
         return SerializerCompat(serde)
 
     return serde
+
+
+class CipherProtocol(Protocol):
+    """Protocol for encryption and decryption of data.
+    - `encrypt`: Encrypt plaintext.
+    - `decrypt`: Decrypt ciphertext.
+    """
+
+    def encrypt(self, plaintext: bytes) -> tuple[str, bytes]:
+        """Encrypt plaintext. Returns a tuple (cipher name, ciphertext)."""
+        ...
+
+    def decrypt(self, ciphername: str, ciphertext: bytes) -> bytes:
+        """Decrypt ciphertext. Returns the plaintext."""
+        ...

libs/checkpoint/langgraph/checkpoint/serde/encrypted.py

@@ -0,0 +1,86 @@
+import os
+from typing import Any
+
+from langgraph.checkpoint.serde.base import CipherProtocol, SerializerProtocol
+from langgraph.checkpoint.serde.jsonplus import JsonPlusSerializer
+
+
+class EncryptedSerializer(SerializerProtocol):
+    """Serializer that encrypts and decrypts data using an encryption protocol."""
+
+    def __init__(
+        self, cipher: CipherProtocol, serde: SerializerProtocol = JsonPlusSerializer()
+    ) -> None:
+        self.cipher = cipher
+        self.serde = serde
+
+    def dumps(self, obj: Any) -> bytes:
+        return self.serde.dumps(obj)
+
+    def loads(self, data: bytes) -> Any:
+        return self.serde.loads(data)
+
+    def dumps_typed(self, obj: Any) -> tuple[str, bytes]:
+        """Serialize an object to a tuple (type, bytes) and encrypt the bytes."""
+        # serialize data
+        typ, data = self.serde.dumps_typed(obj)
+        # encrypt data
+        ciphername, ciphertext = self.cipher.encrypt(data)
+        # add cipher name to type
+        return f"{typ}+{ciphername}", ciphertext
+
+    def loads_typed(self, data: tuple[str, bytes]) -> Any:
+        enc_cipher, ciphertext = data
+        # unencrypted data
+        if "+" not in enc_cipher:
+            return self.serde.loads_typed(data)
+        # extract cipher name
+        typ, ciphername = enc_cipher.split("+", 1)
+        # decrypt data
+        decrypted_data = self.cipher.decrypt(ciphername, ciphertext)
+        # deserialize data
+        return self.serde.loads_typed((typ, decrypted_data))
+
+    @classmethod
+    def from_pycryptodome_aes(
+        cls, serde: SerializerProtocol = JsonPlusSerializer(), **kwargs: Any
+    ) -> "EncryptedSerializer":
+        """Create an EncryptedSerializer using AES encryption."""
+        try:
+            from Crypto.Cipher import AES  # type: ignore
+        except ImportError:
+            raise ImportError(
+                "Pycryptodome is not installed. Please install it with `pip install pycryptodome`."
+            ) from None
+
+        # check if AES key is provided
+        if "key" in kwargs:
+            key: bytes = kwargs.pop("key")
+        else:
+            key_str = os.getenv("LANGGRAPH_AES_KEY")
+            if key_str is None:
+                raise ValueError("LANGGRAPH_AES_KEY environment variable is not set.")
+            key = key_str.encode()
+            if len(key) not in (16, 24, 32):
+                raise ValueError("LANGGRAPH_AES_KEY must be 16, 24, or 32 bytes long.")
+
+        # set default mode to EAX if not provided
+        if kwargs.get("mode") is None:
+            kwargs["mode"] = AES.MODE_EAX
+
+        class PycryptodomeAesCipher(CipherProtocol):
+            def encrypt(self, plaintext: bytes) -> tuple[str, bytes]:
+                cipher = AES.new(key, **kwargs)
+                ciphertext, tag = cipher.encrypt_and_digest(plaintext)
+                return "aes", cipher.nonce + tag + ciphertext
+
+            def decrypt(self, ciphername: str, ciphertext: bytes) -> bytes:
+                assert ciphername == "aes", f"Unsupported cipher: {ciphername}"
+                nonce = ciphertext[:16]
+                tag = ciphertext[16:32]
+                actual_ciphertext = ciphertext[32:]
+
+                cipher = AES.new(key, **kwargs, nonce=nonce)
+                return cipher.decrypt_and_verify(actual_ciphertext, tag)
+
+        return cls(PycryptodomeAesCipher(), serde)

libs/checkpoint/langgraph/checkpoint/serde/jsonplus.py

@@ -487,7 +487,12 @@ def _msgpack_ext_hook(code: int, data: bytes) -> Any:
             except Exception:
                 return cls.construct(**tup[2])
         except Exception:
-            return
+            # for pydantic objects we can't find/reconstruct
+            # let's return the kwargs dict instead
+            try:
+                return tup[2]
+            except NameError:
+                return
     elif code == EXT_PYDANTIC_V2:
         try:
             tup = msgpack.unpackb(
@@ -500,7 +505,12 @@ def _msgpack_ext_hook(code: int, data: bytes) -> Any:
             except Exception:
                 return cls.model_construct(**tup[2])
         except Exception:
-            return
+            # for pydantic objects we can't find/reconstruct
+            # let's return the kwargs dict instead
+            try:
+                return tup[2]
+            except NameError:
+                return
 
 
 def _msgpack_enc(data: Any) -> bytes:

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

@@ -11,9 +11,19 @@ Core types:
 
 from abc import ABC, abstractmethod
 from datetime import datetime
-from typing import Any, Iterable, Literal, NamedTuple, Optional, TypedDict, Union, cast
+from typing import (
+    Any,
+    Iterable,
+    Literal,
+    NamedTuple,
+    Optional,
+    TypedDict,
+    Union,
+    cast,
+)
 
 from langchain_core.embeddings import Embeddings
+from typing_extensions import override
 
 from langgraph.store.base.embed import (
     AEmbeddingsFunc,
@@ -24,6 +34,20 @@ from langgraph.store.base.embed import (
 )
 
 
+class NotProvided:
+    """Sentinel singleton."""
+
+    def __bool__(self) -> Literal[False]:
+        return False
+
+    @override
+    def __repr__(self) -> str:
+        return "NOT_GIVEN"
+
+
+NOT_PROVIDED = NotProvided()
+
+
 class Item:
     """Represents a stored item with metadata.
 
@@ -59,7 +83,7 @@ class Item:
             else created_at
         )
         self.updated_at = (
-            datetime.fromisoformat(cast(str, created_at))
+            datetime.fromisoformat(cast(str, updated_at))
             if isinstance(updated_at, str)
             else updated_at
         )
@@ -166,6 +190,13 @@ class GetOp(NamedTuple):
         "doc456"  # For a document
         ```
     """
+    refresh_ttl: bool = True
+    """Whether to refresh TTLs for the returned item.
+
+    If no TTL was specified for the original item(s),
+    or if TTL support is not enabled for your adapter,
+    this argument is ignored.
+    """
 
 
 class SearchOp(NamedTuple):
@@ -260,6 +291,13 @@ class SearchOp(NamedTuple):
         - "technical documentation about REST APIs"
         - "machine learning papers from 2023"
     """
+    refresh_ttl: bool = True
+    """Whether to refresh TTLs for the returned item.
+
+    If no TTL was specified for the original item(s),
+    or if TTL support is not enabled for your adapter,
+    this argument is ignored.
+    """
 
 
 # Type representing a namespace path that can include wildcards
@@ -463,6 +501,15 @@ class PutOp(NamedTuple):
         ]
         ```
     """
+    ttl: Optional[float] = None
+    """Controls the TTL (time-to-live) for the item in minutes.
+
+    If provided, and if the store you are using supports this feature, the item
+    will expire this many minutes after it was last accessed. The expiration timer
+    refreshes on both read operations (get/search) and write operations (put/update).
+    When the TTL expires, the item will be scheduled for deletion on a best-effort basis.
+    Defaults to None (no expiration).
+    """
 
 
 Op = Union[GetOp, SearchOp, PutOp, ListNamespacesOp]
@@ -473,6 +520,31 @@ class InvalidNamespaceError(ValueError):
     """Provided namespace is invalid."""
 
 
+class TTLConfig(TypedDict, total=False):
+    """Configuration for TTL (time-to-live) behavior in the store."""
+
+    refresh_on_read: bool
+    """Default behavior for refreshing TTLs on read operations (GET and SEARCH).
+    
+    If True, TTLs will be refreshed on read operations (get/search) by default.
+    This can be overridden per-operation by explicitly setting refresh_ttl.
+    Defaults to True if not configured.
+    """
+    default_ttl: Optional[float]
+    """Default TTL (time-to-live) in minutes for new items.
+    
+    If provided, new items will expire after this many minutes after their last access.
+    The expiration timer refreshes on both read and write operations.
+    Defaults to None (no expiration).
+    """
+    sweep_interval_minutes: Optional[int]
+    """Interval in minutes between TTL sweep operations.
+    
+    If provided, the store will periodically delete expired items based on TTL.
+    Defaults to None (no sweeping).
+    """
+
+
 class IndexConfig(TypedDict, total=False):
     """Configuration for indexing documents for semantic search in the store.
 
@@ -612,8 +684,14 @@ class BaseStore(ABC):
         by providing an `index` configuration at creation time. Without this
         configuration, semantic search is disabled and any `index` arguments
         to storage operations will have no effect.
+
+        Similarly, TTL (time-to-live) support is disabled by default.
+        Subclasses must explicitly set `supports_ttl = True` to enable this feature.
     """
 
+    supports_ttl: bool = False
+    ttl_config: Optional[TTLConfig] = None
+
     __slots__ = ("__weakref__",)
 
     @abstractmethod
@@ -640,17 +718,28 @@ class BaseStore(ABC):
             The order of results matches the order of input operations.
         """
 
-    def get(self, namespace: tuple[str, ...], key: str) -> Optional[Item]:
+    def get(
+        self,
+        namespace: tuple[str, ...],
+        key: str,
+        *,
+        refresh_ttl: Optional[bool] = None,
+    ) -> Optional[Item]:
         """Retrieve a single item.
 
         Args:
             namespace: Hierarchical path for the item.
             key: Unique identifier within the namespace.
+            refresh_ttl: Whether to refresh TTLs for the returned item.
+                If None (default), uses the store's default refresh_ttl setting.
+                If no TTL is specified, this argument is ignored.
 
         Returns:
             The retrieved item or None if not found.
         """
-        return self.batch([GetOp(namespace, key)])[0]
+        return self.batch(
+            [GetOp(namespace, str(key), _ensure_refresh(self.ttl_config, refresh_ttl))]
+        )[0]
 
     def search(
         self,
@@ -661,6 +750,7 @@ class BaseStore(ABC):
         filter: Optional[dict[str, Any]] = None,
         limit: int = 10,
         offset: int = 0,
+        refresh_ttl: Optional[bool] = None,
     ) -> list[SearchItem]:
         """Search for items within a namespace prefix.
 
@@ -670,6 +760,8 @@ class BaseStore(ABC):
             filter: Key-value pairs to filter results.
             limit: Maximum number of items to return.
             offset: Number of items to skip before returning results.
+            refresh_ttl: Whether to refresh TTLs for the returned items.
+                If no TTL is specified, this argument is ignored.
 
         Returns:
             List of items matching the search criteria.
@@ -707,7 +799,18 @@ class BaseStore(ABC):
             Note: Natural language search support depends on your store implementation
             and requires proper embedding configuration.
         """
-        return self.batch([SearchOp(namespace_prefix, filter, limit, offset, query)])[0]
+        return self.batch(
+            [
+                SearchOp(
+                    namespace_prefix,
+                    filter,
+                    limit,
+                    offset,
+                    query,
+                    _ensure_refresh(self.ttl_config, refresh_ttl),
+                )
+            ]
+        )[0]
 
     def put(
         self,
@@ -715,6 +818,8 @@ class BaseStore(ABC):
         key: str,
         value: dict[str, Any],
         index: Optional[Union[Literal[False], list[str]]] = None,
+        *,
+        ttl: Union[Optional[float], "NotProvided"] = NOT_PROVIDED,
     ) -> None:
         """Store or update an item in the store.
 
@@ -735,12 +840,20 @@ class BaseStore(ABC):
                     - Nested fields: "metadata.title"
                     - Array access: "chapters[*].content" (each indexed separately)
                     - Specific indices: "authors[0].name"
+            ttl: Time to live in minutes. Support for this argument depends on your store adapter.
+                If specified, the item will expire after this many minutes from when it was last accessed.
+                None means no expiration. Expired runs will be deleted opportunistically.
+                By default, the expiration timer refreshes on both read operations (get/search)
+                and write operations (put/update), whenever the item is included in the operation.
 
         Note:
             Indexing support depends on your store implementation.
             If you do not initialize the store with indexing capabilities,
             the `index` parameter will be ignored.
 
+            Similarly, TTL support depends on the specific store implementation.
+            Some implementations may not support expiration of items.
+
         ???+ example "Examples"
             Store item. Indexing depends on how you configure the store.
             ```python
@@ -759,7 +872,22 @@ class BaseStore(ABC):
             ```
         """
         _validate_namespace(namespace)
-        self.batch([PutOp(namespace, key, value, index=index)])
+        if ttl not in (NOT_PROVIDED, None) and not self.supports_ttl:
+            raise NotImplementedError(
+                f"TTL is not supported by {self.__class__.__name__}. "
+                f"Use a store implementation that supports TTL or set ttl=None."
+            )
+        self.batch(
+            [
+                PutOp(
+                    namespace,
+                    str(key),
+                    value,
+                    index=index,
+                    ttl=_ensure_ttl(self.ttl_config, ttl),
+                )
+            ]
+        )
 
     def delete(self, namespace: tuple[str, ...], key: str) -> None:
         """Delete an item.
@@ -768,7 +896,7 @@ class BaseStore(ABC):
             namespace: Hierarchical path for the item.
             key: Unique identifier within the namespace.
         """
-        self.batch([PutOp(namespace, key, None)])
+        self.batch([PutOp(namespace, str(key), None, ttl=None)])
 
     def list_namespaces(
         self,
@@ -823,7 +951,13 @@ class BaseStore(ABC):
         )
         return self.batch([op])[0]
 
-    async def aget(self, namespace: tuple[str, ...], key: str) -> Optional[Item]:
+    async def aget(
+        self,
+        namespace: tuple[str, ...],
+        key: str,
+        *,
+        refresh_ttl: Optional[bool] = None,
+    ) -> Optional[Item]:
         """Asynchronously retrieve a single item.
 
         Args:
@@ -833,7 +967,17 @@ class BaseStore(ABC):
         Returns:
             The retrieved item or None if not found.
         """
-        return (await self.abatch([GetOp(namespace, key)]))[0]
+        return (
+            await self.abatch(
+                [
+                    GetOp(
+                        namespace,
+                        str(key),
+                        _ensure_refresh(self.ttl_config, refresh_ttl),
+                    )
+                ]
+            )
+        )[0]
 
     async def asearch(
         self,
@@ -844,6 +988,7 @@ class BaseStore(ABC):
         filter: Optional[dict[str, Any]] = None,
         limit: int = 10,
         offset: int = 0,
+        refresh_ttl: Optional[bool] = None,
     ) -> list[SearchItem]:
         """Asynchronously search for items within a namespace prefix.
 
@@ -853,6 +998,9 @@ class BaseStore(ABC):
             filter: Key-value pairs to filter results.
             limit: Maximum number of items to return.
             offset: Number of items to skip before returning results.
+            refresh_ttl: Whether to refresh TTLs for the returned items.
+                If None (default), uses the store's TTLConfig.refresh_default setting.
+                If TTLConfig is not provided or no TTL is specified, this argument is ignored.
 
         Returns:
             List of items matching the search criteria.
@@ -892,7 +1040,16 @@ class BaseStore(ABC):
         """
         return (
             await self.abatch(
-                [SearchOp(namespace_prefix, filter, limit, offset, query)]
+                [
+                    SearchOp(
+                        namespace_prefix,
+                        filter,
+                        limit,
+                        offset,
+                        query,
+                        _ensure_refresh(self.ttl_config, refresh_ttl),
+                    )
+                ]
             )
         )[0]
 
@@ -902,6 +1059,8 @@ class BaseStore(ABC):
         key: str,
         value: dict[str, Any],
         index: Optional[Union[Literal[False], list[str]]] = None,
+        *,
+        ttl: Union[Optional[float], "NotProvided"] = NOT_PROVIDED,
     ) -> None:
         """Asynchronously store or update an item in the store.
 
@@ -922,12 +1081,20 @@ class BaseStore(ABC):
                     - Nested fields: "metadata.title"
                     - Array access: "chapters[*].content" (each indexed separately)
                     - Specific indices: "authors[0].name"
+            ttl: Time to live in minutes. Support for this argument depends on your store adapter.
+                If specified, the item will expire after this many minutes from when it was last accessed.
+                None means no expiration. Expired runs will be deleted opportunistically.
+                By default, the expiration timer refreshes on both read operations (get/search)
+                and write operations (put/update), whenever the item is included in the operation.
 
         Note:
             Indexing support depends on your store implementation.
             If you do not initialize the store with indexing capabilities,
             the `index` parameter will be ignored.
 
+            Similarly, TTL support depends on the specific store implementation.
+            Some implementations may not support expiration of items.
+
         ???+ example "Examples"
             Store item. Indexing depends on how you configure the store.
             ```python
@@ -954,7 +1121,22 @@ class BaseStore(ABC):
             ```
         """
         _validate_namespace(namespace)
-        await self.abatch([PutOp(namespace, key, value, index=index)])
+        if ttl not in (NOT_PROVIDED, None) and not self.supports_ttl:
+            raise NotImplementedError(
+                f"TTL is not supported by {self.__class__.__name__}. "
+                f"Use a store implementation that supports TTL or set ttl=None."
+            )
+        await self.abatch(
+            [
+                PutOp(
+                    namespace,
+                    str(key),
+                    value,
+                    index=index,
+                    ttl=_ensure_ttl(self.ttl_config, ttl),
+                )
+            ]
+        )
 
     async def adelete(self, namespace: tuple[str, ...], key: str) -> None:
         """Asynchronously delete an item.
@@ -963,7 +1145,7 @@ class BaseStore(ABC):
             namespace: Hierarchical path for the item.
             key: Unique identifier within the namespace.
         """
-        await self.abatch([PutOp(namespace, key, None)])
+        await self.abatch([PutOp(namespace, str(key), None)])
 
     async def alist_namespaces(
         self,
@@ -1043,6 +1225,27 @@ def _validate_namespace(namespace: tuple[str, ...]) -> None:
         )
 
 
+def _ensure_refresh(
+    ttl_config: Optional[TTLConfig], refresh_ttl: Optional[bool] = None
+) -> bool:
+    if refresh_ttl is not None:
+        return refresh_ttl
+    if ttl_config is not None:
+        return ttl_config.get("refresh_on_read", True)
+    return True
+
+
+def _ensure_ttl(
+    ttl_config: Optional[TTLConfig],
+    ttl: Union[Optional[float], "NotProvided"] = NOT_PROVIDED,
+) -> Optional[float]:
+    if ttl is NOT_PROVIDED:
+        if ttl_config:
+            return ttl_config.get("default_ttl")
+        return None
+    return ttl
+
+
 __all__ = [
     "BaseStore",
     "Item",

libs/checkpoint/langgraph/store/base/batch.py

@@ -5,17 +5,21 @@ from collections.abc import Iterable
 from typing import Any, Callable, Literal, Optional, TypeVar, Union
 
 from langgraph.store.base import (
+    NOT_PROVIDED,
     BaseStore,
     GetOp,
     Item,
     ListNamespacesOp,
     MatchCondition,
     NamespacePath,
+    NotProvided,
     Op,
     PutOp,
     Result,
     SearchItem,
     SearchOp,
+    _ensure_refresh,
+    _ensure_ttl,
     _validate_namespace,
 )
 
@@ -68,10 +72,21 @@ class AsyncBatchedBaseStore(BaseStore):
         self,
         namespace: tuple[str, ...],
         key: str,
+        *,
+        refresh_ttl: Optional[bool] = None,
     ) -> Optional[Item]:
         assert not self._task.done()
         fut = self._loop.create_future()
-        self._aqueue.put_nowait((fut, GetOp(namespace, key)))
+        self._aqueue.put_nowait(
+            (
+                fut,
+                GetOp(
+                    namespace,
+                    key,
+                    refresh_ttl=_ensure_refresh(self.ttl_config, refresh_ttl),
+                ),
+            )
+        )
         return await fut
 
     async def asearch(
@@ -83,11 +98,22 @@ class AsyncBatchedBaseStore(BaseStore):
         filter: Optional[dict[str, Any]] = None,
         limit: int = 10,
         offset: int = 0,
+        refresh_ttl: Optional[bool] = None,
     ) -> list[SearchItem]:
         assert not self._task.done()
         fut = self._loop.create_future()
         self._aqueue.put_nowait(
-            (fut, SearchOp(namespace_prefix, filter, limit, offset, query))
+            (
+                fut,
+                SearchOp(
+                    namespace_prefix,
+                    filter,
+                    limit,
+                    offset,
+                    query,
+                    refresh_ttl=_ensure_refresh(self.ttl_config, refresh_ttl),
+                ),
+            )
         )
         return await fut
 
@@ -97,11 +123,20 @@ class AsyncBatchedBaseStore(BaseStore):
         key: str,
         value: dict[str, Any],
         index: Optional[Union[Literal[False], list[str]]] = None,
+        *,
+        ttl: Union[Optional[float], "NotProvided"] = NOT_PROVIDED,
     ) -> None:
         assert not self._task.done()
         _validate_namespace(namespace)
         fut = self._loop.create_future()
-        self._aqueue.put_nowait((fut, PutOp(namespace, key, value, index)))
+        self._aqueue.put_nowait(
+            (
+                fut,
+                PutOp(
+                    namespace, key, value, index, ttl=_ensure_ttl(self.ttl_config, ttl)
+                ),
+            )
+        )
         return await fut
 
     async def adelete(
@@ -149,9 +184,11 @@ class AsyncBatchedBaseStore(BaseStore):
         self,
         namespace: tuple[str, ...],
         key: str,
+        *,
+        refresh_ttl: Optional[bool] = None,
     ) -> Optional[Item]:
         return asyncio.run_coroutine_threadsafe(
-            self.aget(namespace, key=key), self._loop
+            self.aget(namespace, key=key, refresh_ttl=refresh_ttl), self._loop
         ).result()
 
     @_check_loop
@@ -164,10 +201,16 @@ class AsyncBatchedBaseStore(BaseStore):
         filter: Optional[dict[str, Any]] = None,
         limit: int = 10,
         offset: int = 0,
+        refresh_ttl: Optional[bool] = None,
     ) -> list[SearchItem]:
         return asyncio.run_coroutine_threadsafe(
             self.asearch(
-                namespace_prefix, query=query, filter=filter, limit=limit, offset=offset
+                namespace_prefix,
+                query=query,
+                filter=filter,
+                limit=limit,
+                offset=offset,
+                refresh_ttl=refresh_ttl,
             ),
             self._loop,
         ).result()
@@ -179,10 +222,19 @@ class AsyncBatchedBaseStore(BaseStore):
         key: str,
         value: dict[str, Any],
         index: Optional[Union[Literal[False], list[str]]] = None,
+        *,
+        ttl: Union[Optional[float], "NotProvided"] = NOT_PROVIDED,
     ) -> None:
         _validate_namespace(namespace)
         asyncio.run_coroutine_threadsafe(
-            self.aput(namespace, key=key, value=value, index=index), self._loop
+            self.aput(
+                namespace,
+                key=key,
+                value=value,
+                index=index,
+                ttl=_ensure_ttl(self.ttl_config, ttl),
+            ),
+            self._loop,
         ).result()
 
     @_check_loop

libs/checkpoint/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langgraph-checkpoint"
-version = "2.0.16"
+version = "2.0.21"
 description = "Library with base interfaces for LangGraph checkpoint savers."
 authors = []
 license = "MIT"

libs/checkpoint/tests/test_jsonplus.py

@@ -130,7 +130,7 @@ def test_serde_jsonplus() -> None:
             key="my-key",
             namespace=("a", "name", " "),
             created_at=datetime(2024, 9, 24, 17, 29, 10, 128397),
-            updated_at=datetime(2024, 9, 24, 17, 29, 10, 128397),
+            updated_at=datetime(2024, 9, 24, 17, 29, 11, 128397),
         ),
     }
 

libs/checkpoint/tests/test_store.py

@@ -148,8 +148,8 @@ async def test_async_batch_store(mocker: MockerFixture) -> None:
     assert abatch.call_count == 1
     assert [tuple(c.args[0]) for c in abatch.call_args_list] == [
         (
-            GetOp(("a",), "b"),
-            GetOp(("c",), "d"),
+            GetOp(("a",), "b", refresh_ttl=True),
+            GetOp(("c",), "d", refresh_ttl=True),
         ),
     ]
 
@@ -467,8 +467,8 @@ async def test_async_batch_store_deduplication(mocker: MockerFixture) -> None:
     assert len(abatch.call_args_list) == 1
     ops = list(abatch.call_args_list[0].args[1])
     assert len(ops) == 2
-    assert GetOp(("test",), "same") in ops
-    assert GetOp(("test",), "different") in ops
+    assert GetOp(("test",), "same", refresh_ttl=True) in ops
+    assert GetOp(("test",), "different", refresh_ttl=True) in ops
 
     abatch.reset_mock()
 

libs/cli/Makefile

@@ -1,4 +1,4 @@
-.PHONY: test lint format test-integration
+.PHONY: test lint format test-integration update-schema
 
 ######################
 # TESTING AND COVERAGE
@@ -31,3 +31,6 @@ lint lint_diff lint_package lint_tests:
 format format_diff:
 	poetry run ruff format $(PYTHON_FILES)
 	poetry run ruff check --select I --fix $(PYTHON_FILES)
+
+update-schema:
+	poetry run python generate_schema.py

libs/cli/README.md

@@ -79,7 +79,7 @@ The CLI uses a `langgraph.json` configuration file with these key settings:
 }
 ```
 
-See the [full documentation](https://langchain-ai.github.io/langgraph/docs/cloud/reference/cli.html) for detailed configuration options.
+See the [full documentation](https://langchain-ai.github.io/langgraph/cloud/reference/cli/) for detailed configuration options.
 
 ## Development
 

libs/cli/generate_schema.py

@@ -0,0 +1,226 @@
+#!/usr/bin/env python3
+"""
+Script to generate a JSON schema for the langgraph-cli Config class.
+
+This script creates a schema.json file that can be referenced in langgraph.json files
+to provide IDE autocompletion and validation.
+"""
+
+import inspect
+import json
+import textwrap
+from pathlib import Path
+
+import msgspec
+
+from langgraph_cli.config import (
+    AuthConfig,
+    Config,
+    CorsConfig,
+    HttpConfig,
+    IndexConfig,
+    SecurityConfig,
+    StoreConfig,
+)
+
+
+def add_descriptions_to_schema(schema, cls):
+    """Add docstring descriptions to the schema properties."""
+    if schema.get("description"):
+        schema["description"] = inspect.cleandoc(schema["description"])
+    elif class_doc := inspect.getdoc(cls):
+        schema["description"] = inspect.cleandoc(class_doc)
+    # Get attribute docstrings from the class
+    attr_docs = {}
+
+    # Also check class annotations for docstrings
+    source_lines = inspect.getsourcelines(cls)[0]
+    current_attr = None
+    docstring_lines = []
+
+    for line in source_lines:
+        line = line.strip()
+
+        # Check for attribute definition (TypedDict style)
+        if ":" in line and not line.startswith("#") and not line.startswith('"""'):
+            parts = line.split(":", 1)
+            if len(parts) == 2 and parts[0].strip().isidentifier():
+                # If we were collecting a docstring, save it for the previous attribute
+                if current_attr and docstring_lines:
+                    attr_docs[current_attr] = "\n".join(docstring_lines).strip('"')
+                    docstring_lines = []
+
+                current_attr = parts[0].strip()
+
+        # Check for docstring after attribute
+        elif line.startswith('"""') and current_attr:
+            # Start or end of a docstring
+            if len(line) > 3 and line.endswith('"""'):
+                # Single line docstring
+                attr_docs[current_attr] = line.strip('"')
+                current_attr = None
+            elif docstring_lines:
+                # End of multi-line docstring
+                docstring_lines.append(line.rstrip('"'))
+                attr_docs[current_attr] = "\n".join(docstring_lines).strip('"')
+                docstring_lines = []
+                current_attr = None
+            else:
+                # Start of multi-line docstring
+                docstring_lines.append(line.lstrip('"'))
+
+        # Continue multi-line docstring
+        elif docstring_lines and current_attr:
+            docstring_lines.append(line.strip('"'))
+
+    # Add the last docstring if there is one
+    if current_attr and docstring_lines:
+        attr_docs[current_attr] = "\n".join(docstring_lines).strip('"')
+
+    # Add descriptions to properties
+    if "properties" in schema:
+        for prop_name, prop_schema in schema["properties"].items():
+            # First try to get from attribute docstrings
+            if prop_name in attr_docs and "description" not in prop_schema:
+                prop_schema["description"] = textwrap.dedent(attr_docs[prop_name])
+            # Fall back to class docstring parsing
+            elif class_doc:
+                for line in class_doc.split("\n"):
+                    if line.strip().startswith(
+                        f"{prop_name}:"
+                    ) or line.strip().startswith(f'"{prop_name}"'):
+                        description = line.split(":", 1)[1].strip()
+                        if description and "description" not in prop_schema:
+                            prop_schema["description"] = description
+                            break
+
+    # Recursively process nested definitions
+    if "$defs" in schema:
+        for def_name, def_schema in schema["$defs"].items():
+            # Find the class that corresponds to this definition
+            for potential_cls in [
+                Config,
+                StoreConfig,
+                IndexConfig,
+                AuthConfig,
+                SecurityConfig,
+                HttpConfig,
+                CorsConfig,
+            ]:
+                if potential_cls.__name__ == def_name:
+                    add_descriptions_to_schema(def_schema, potential_cls)
+                    break
+
+    return schema
+
+
+def generate_schema():
+    """Generate a JSON schema for the Config class using msgspec."""
+    # Generate the basic schema
+    schema = msgspec.json.schema(Config)
+
+    # Add title and description
+    schema["title"] = "LangGraph CLI Configuration"
+    schema["description"] = "Configuration schema for langgraph-cli"
+
+    # Add docstring descriptions
+    schema = add_descriptions_to_schema(schema, Config)
+
+    # Add constraint that only one of python_version or node_version should be specified
+    config_schema = schema["$defs"]["Config"]
+
+    # Create two subschemas: one with python_version and one with node_version
+    # Define properties specific to Python projects
+    python_specific_props = ["python_version", "pip_config_file"]
+    # Define properties specific to Node.js projects
+    node_specific_props = ["node_version"]
+    # Define properties common to both project types
+    common_props = [
+        k
+        for k in config_schema["properties"]
+        if k not in python_specific_props and k not in node_specific_props
+    ]
+
+    # Create Python schema with python_version and pip_config_file
+    python_schema = {
+        "type": "object",
+        "properties": {
+            # Include Python-specific properties
+            **{k: config_schema["properties"][k].copy() for k in python_specific_props},
+            # Include common properties
+            **{k: config_schema["properties"][k].copy() for k in common_props},
+        },
+        "required": ["dependencies", "graphs"],
+    }
+
+    # Add enum constraint for python_version
+    if "python_version" in python_schema["properties"]:
+        python_schema["properties"]["python_version"]["enum"] = ["3.11", "3.12"]
+
+    # Create Node.js schema with node_version
+    node_schema = {
+        "type": "object",
+        "properties": {
+            # Include Node-specific properties
+            **{k: config_schema["properties"][k].copy() for k in node_specific_props},
+            # Include common properties
+            **{k: config_schema["properties"][k].copy() for k in common_props},
+        },
+        "required": ["node_version", "graphs"],
+    }
+
+    # Add enum constraint for node_version
+    if "node_version" in node_schema["properties"]:
+        node_schema["properties"]["node_version"]["anyOf"] = [
+            {"type": "string", "enum": ["20"]},
+            {"type": "null"},
+        ]
+
+    # Replace the Config schema with a oneOf constraint
+    config_schema["oneOf"] = [python_schema, node_schema]
+
+    # Remove the properties field as it's now defined in the oneOf subschemas
+    if "properties" in config_schema:
+        del config_schema["properties"]
+
+    return schema
+
+
+def main():
+    """Generate the schema and write it to a file."""
+    schema = generate_schema()
+
+    # Add versioning to the schema
+    import importlib.metadata
+
+    try:
+        version = importlib.metadata.version("langgraph_cli").split(".")
+        schema_version = f"v{version[0]}"
+    except importlib.metadata.PackageNotFoundError:
+        schema_version = "v1"
+
+    # Add version to schema
+    schema["version"] = schema_version
+
+    config_dir = Path(__file__).parent / "schemas"
+
+    # Create versioned schema file
+    versioned_path = config_dir / f"schema.{schema_version}.json"
+    with open(versioned_path, "w") as f:
+        json.dump(schema, f, indent=2)
+
+    # Also create a latest version
+    latest_path = config_dir / "schema.json"
+    with open(latest_path, "w") as f:
+        json.dump(schema, f, indent=2)
+
+    print(f"Schema written to {versioned_path} and {latest_path}")
+    print(
+        f"You can now add '$schema: https://raw.githubusercontent.com/langchain-ai/langgraph/refs/heads/main/libs/cli/schemas/schema.json'"
+        f" or '$schema: https://raw.githubusercontent.com/langchain-ai/langgraph/refs/heads/main/libs/cli/schemas/schema.{schema_version}.json'"
+        " to your langgraph.json files"
+    )
+
+
+if __name__ == "__main__":
+    main()

libs/cli/langgraph_cli/config.py

@@ -3,7 +3,7 @@ import os
 import pathlib
 import textwrap
 from collections import Counter
-from typing import NamedTuple, Optional, TypedDict, Union
+from typing import Any, NamedTuple, Optional, TypedDict, Union
 
 import click
 
@@ -11,13 +11,43 @@ MIN_NODE_VERSION = "20"
 MIN_PYTHON_VERSION = "3.11"
 
 
+class TTLConfig(TypedDict, total=False):
+    """Configuration for TTL (time-to-live) behavior in the store."""
+
+    refresh_on_read: bool
+    """Default behavior for refreshing TTLs on read operations (GET and SEARCH).
+    
+    If True, TTLs will be refreshed on read operations (get/search) by default.
+    This can be overridden per-operation by explicitly setting refresh_ttl.
+    Defaults to True if not configured.
+    """
+    default_ttl: Optional[float]
+    """Optional. Default TTL (time-to-live) in minutes for new items.
+    
+    If provided, all new items will have this TTL unless explicitly overridden.
+    If omitted, items will have no TTL by default.
+    """
+    sweep_interval_minutes: Optional[int]
+    """Optional. Interval in minutes between TTL sweep iterations.
+    
+    If provided, the store will periodically delete expired items based on the TTL.
+    If omitted, no automatic sweeping will occur.
+    """
+
+
 class IndexConfig(TypedDict, total=False):
-    """Configuration for indexing documents for semantic search in the store."""
+    """Configuration for indexing documents for semantic search in the store.
+
+    This governs how text is converted into embeddings and stored for vector-based lookups.
+    """
 
     dims: int
-    """Number of dimensions in the embedding vectors.
+    """Required. Dimensionality of the embedding vectors you will store.
     
-    Common embedding models have the following dimensions:
+    Must match the output dimension of your selected embedding model or custom embed function.
+    If mismatched, you will likely encounter shape/size errors when inserting or querying vectors.
+    
+    Common embedding model output dimensions:
         - openai:text-embedding-3-large: 3072
         - openai:text-embedding-3-small: 1536
         - openai:text-embedding-ada-002: 1536
@@ -28,42 +58,130 @@ class IndexConfig(TypedDict, total=False):
     """
 
     embed: str
-    """Optional model (string) to generate embeddings from text or path to model or function.
+    """Required. Identifier or reference to the embedding model or a custom embedding function.
     
-    Examples:
+    The format can vary:
+      - "<provider>:<model_name>" for recognized providers (e.g., "openai:text-embedding-3-large")
+      - "path/to/module.py:function_name" for your own local embedding function
+      - "my_custom_embed" if it's a known alias in your system
+
+     Examples:
         - "openai:text-embedding-3-large"
         - "cohere:embed-multilingual-v3.0"
-        - "src/app.py:embeddings
+        - "src/app.py:embeddings"
+    
+    Note: Must return embeddings of dimension `dims`.
     """
 
     fields: Optional[list[str]]
-    """Fields to extract text from for embedding generation.
+    """Optional. List of JSON fields to extract before generating embeddings.
     
-    Defaults to the root ["$"], which embeds the json object as a whole.
+    Defaults to ["$"], which means the entire JSON object is embedded as one piece of text.
+    If you provide multiple fields (e.g. ["title", "content"]), each is extracted and embedded separately,
+    often saving token usage if you only care about certain parts of the data.
+    
+    Example:
+        fields=["title", "abstract", "author.biography"]
     """
 
 
 class StoreConfig(TypedDict, total=False):
-    embed: Optional[IndexConfig]
-    """Configuration for vector embeddings in store."""
+    """Configuration for the built-in long-term memory store.
+
+    This store can optionally perform semantic search. If you omit `index`,
+    the store will just handle traditional (non-embedded) data without vector lookups.
+    """
+
+    index: Optional[IndexConfig]
+    """Optional. Defines the vector-based semantic search configuration.
+    
+    If provided, the store will:
+      - Generate embeddings according to `index.embed`
+      - Enforce the embedding dimension given by `index.dims`
+      - Embed only specified JSON fields (if any) from `index.fields`
+    
+    If omitted, no vector index is initialized.
+    """
+
+    ttl: Optional[TTLConfig]
+    """Optional. Defines the TTL (time-to-live) behavior configuration.
+    
+    If provided, the store will apply TTL settings according to the configuration.
+    If omitted, no TTL behavior is configured.
+    """
 
 
 class SecurityConfig(TypedDict, total=False):
-    securitySchemes: dict
-    security: list
+    """Configuration for OpenAPI security definitions and requirements.
+
+    Useful for specifying global or path-level authentication and authorization flows
+    (e.g., OAuth2, API key headers, etc.).
+    """
+
+    securitySchemes: dict[str, dict[str, Any]]
+    """Required. Dict describing each security scheme recognized by your OpenAPI spec.
+    
+    Keys are scheme names (e.g. "OAuth2", "ApiKeyAuth") and values are their definitions.
+    Example:
+        {
+            "OAuth2": {
+                "type": "oauth2",
+                "flows": {
+                    "password": {
+                        "tokenUrl": "/token",
+                        "scopes": {"read": "Read data", "write": "Write data"}
+                    }
+                }
+            }
+        }
+    """
+    security: list[dict[str, list[str]]]
+    """Optional. Global security requirements across all endpoints.
+    
+    Each element in the list maps a security scheme (e.g. "OAuth2") to a list of scopes (e.g. ["read", "write"]).
+    Example:
+        [
+            {"OAuth2": ["read", "write"]},
+            {"ApiKeyAuth": []}
+        ]
+    """
     # path => {method => security}
-    paths: dict[str, dict[str, list]]
+    paths: dict[str, dict[str, list[dict[str, list[str]]]]]
+    """Optional. Path-specific security overrides.
+    
+    Keys are path templates (e.g., "/items/{item_id}"), mapping to:
+      - Keys that are HTTP methods (e.g., "GET", "POST"),
+      - Values are lists of security definitions (just like `security`) for that method.
+    
+    Example:
+        {
+            "/private_data": {
+                "GET": [{"OAuth2": ["read"]}],
+                "POST": [{"OAuth2": ["write"]}]
+            }
+        }
+    """
 
 
 class AuthConfig(TypedDict, total=False):
-    path: str
-    """Path to the authentication function in a Python file."""
-    disable_studio_auth: bool
-    """Whether to disable auth when connecting from the LangSmith Studio."""
-    openapi: SecurityConfig
-    """The schema to use for updating the openapi spec.
+    """Configuration for custom authentication logic and how it integrates into the OpenAPI spec."""
 
-    Example:
+    path: str
+    """Required. Path to an instance of the Auth() class that implements custom authentication.
+    
+    Format: "path/to/file.py:my_auth"
+    """
+    disable_studio_auth: bool
+    """Optional. Whether to disable LangSmith API-key authentication for requests originating the Studio. 
+    
+    Defaults to False, meaning that if a particular header is set, the server will verify the `x-api-key` header
+    value is a valid API key for the deployment's workspace. If True, all requests will go through your custom
+    authentication logic, regardless of origin of the request.
+    """
+    openapi: SecurityConfig
+    """Required. Detailed security configuration that merges into your deployment's OpenAPI spec.
+    
+    Example (OAuth2):
         {
             "securitySchemes": {
                 "OAuth2": {
@@ -71,88 +189,185 @@ class AuthConfig(TypedDict, total=False):
                     "flows": {
                         "password": {
                             "tokenUrl": "/token",
-                            "scopes": {
-                                "me": "Read information about the current user",
-                                "items": "Access to create and manage items"
-                            }
+                            "scopes": {"me": "Read user info", "items": "Manage items"}
                         }
                     }
                 }
             },
             "security": [
-                {"OAuth2": ["me"]}  # Default security requirement for all endpoints
+                {"OAuth2": ["me"]}
             ]
         }
     """
 
 
 class CorsConfig(TypedDict, total=False):
+    """Specifies Cross-Origin Resource Sharing (CORS) rules for your server.
+
+    If omitted, defaults are typically very restrictive (often no cross-origin requests).
+    Configure carefully if you want to allow usage from browsers hosted on other domains.
+    """
+
     allow_origins: list[str]
+    """Optional. List of allowed origins (e.g., "https://example.com").
+    
+    Default is often an empty list (no external origins). 
+    Use "*" only if you trust all origins, as that bypasses most restrictions.
+    """
     allow_methods: list[str]
+    """Optional. HTTP methods permitted for cross-origin requests (e.g. ["GET", "POST"]).
+    
+    Default might be ["GET", "POST", "OPTIONS"] depending on your server framework.
+    """
     allow_headers: list[str]
+    """Optional. HTTP headers that can be used in cross-origin requests (e.g. ["Content-Type", "Authorization"])."""
     allow_credentials: bool
+    """Optional. If True, cross-origin requests can include credentials (cookies, auth headers).
+    
+    Default False to avoid accidentally exposing secured endpoints to untrusted sites.
+    """
     allow_origin_regex: str
+    """Optional. A regex pattern for matching allowed origins, used if you have dynamic subdomains.
+    
+    Example: "^https://.*\.mycompany\.com$"
+    """
     expose_headers: list[str]
+    """Optional. List of headers that browsers are allowed to read from the response in cross-origin contexts."""
     max_age: int
+    """Optional. How many seconds the browser may cache preflight responses.
+    
+    Default might be 600 (10 minutes). Larger values reduce preflight requests but can cause stale configurations.
+    """
 
 
 class HttpConfig(TypedDict, total=False):
+    """Configuration for the built-in HTTP server that powers your deployment's routes and endpoints."""
+
     app: str
-    """Import path for a custom Starlette/FastAPI app to mount"""
+    """Optional. Import path to a custom Starlette/FastAPI application to mount.
+    
+    Format: "path/to/module.py:app_var"
+    If provided, it can override or extend the default routes.
+    """
     disable_assistants: bool
-    """Disable /assistants routes"""
+    """Optional. If True, /assistants routes are removed from the server.
+    
+    Default is False (meaning /assistants is enabled).
+    """
     disable_threads: bool
-    """Disable /threads routes"""
+    """Optional. If True, /threads routes are removed.
+    
+    Default is False.
+    """
     disable_runs: bool
-    """Disable /runs routes"""
+    """Optional. If True, /runs routes are removed.
+    
+    Default is False.
+    """
     disable_store: bool
-    """Disable /store routes"""
+    """Optional. If True, /store routes are removed, disabling direct store interactions via HTTP.
+    
+    Default is False.
+    """
     disable_meta: bool
-    """Disable /ok, /info, /metrics, and /docs routes"""
+    """Optional. If True, all meta endpoints (/ok, /info, /metrics, /docs) are disabled.
+    
+    Default is False.
+    """
     cors: Optional[CorsConfig]
-    """Cross-Origin Resource Sharing (CORS) configuration"""
+    """Optional. Defines CORS restrictions. If omitted, no special rules are set and 
+    cross-origin behavior depends on default server settings.
+    """
 
 
 class Config(TypedDict, total=False):
-    """Configuration for langgraph-cli."""
+    """Top-level config for langgraph-cli or similar deployment tooling."""
 
     python_version: str
-    """Python version to use."""
+    """Optional. Python version in 'major.minor' format (e.g. '3.11'). 
+    Must be at least 3.11 or greater for this deployment to function properly.
+    """
 
     node_version: Optional[str]
-    """Node.js version to use."""
+    """Optional. Node.js version as a major version (e.g. '20'), if your deployment needs Node.
+    Must be >= 20 if provided.
+    """
 
     pip_config_file: Optional[str]
-    """Path to a pip configuration file."""
+    """Optional. Path to a pip config file (e.g., "/etc/pip.conf" or "pip.ini") for controlling
+    package installation (custom indices, credentials, etc.).
+    
+    Only relevant if Python dependencies are installed via pip. If omitted, default pip settings are used.
+    """
 
     dockerfile_lines: list[str]
-    """Additional lines to add to the Dockerfile."""
+    """Optional. Additional Docker instructions that will be appended to your base Dockerfile.
+    
+    Useful for installing OS packages, setting environment variables, etc. 
+    Example:
+        dockerfile_lines=[
+            "RUN apt-get update && apt-get install -y libmagic-dev",
+            "ENV MY_CUSTOM_VAR=hello_world"
+        ]
+    """
 
     dependencies: list[str]
-    """Additional Python dependencies to install."""
+    """List of Python dependencies to install, either from PyPI or local paths.
+    
+    Examples:
+      - "." or "./src" if you have a local Python package
+      - str (aka "anthropic") for a PyPI package
+      - "git+https://github.com/org/repo.git@main" for a Git-based package
+    Defaults to an empty list, meaning no additional packages installed beyond your base environment.
+    """
 
     graphs: dict[str, str]
-    """Mapping of graph names to their definitions."""
+    """Optional. Named definitions of graphs, each pointing to a Python object.
+
+    
+    Graphs can be StateGraph, @entrypoint, or any other Pregel object OR they can point to (async) context
+    managers that accept a single configuration argument (of type RunnableConfig) and return a pregel object
+    (instance of Stategraph, etc.).
+    
+    Keys are graph names, values are "path/to/file.py:object_name".
+    Example:
+        {
+            "mygraph": "graphs/my_graph.py:graph_definition",
+            "anothergraph": "graphs/another.py:get_graph"
+        }
+    """
 
     env: Union[dict[str, str], str]
-    """Environment variables to set.
-
-    If a dictionary is provided, the keys are environment variable names
-    and the values are the corresponding environment variable values.
-
-    If a string is provided, it is interpreted as a path to a file containing
-    environment variables in the format KEY=VALUE, with one environment variable
-    per line.
+    """Optional. Environment variables to set for your deployment.
+    
+    - If given as a dict, keys are variable names and values are their values.
+    - If given as a string, it must be a path to a file containing lines in KEY=VALUE format.
+    
+    Example as a dict:
+        env={"API_TOKEN": "abc123", "DEBUG": "true"}
+    Example as a file path:
+        env=".env"
     """
 
     store: Optional[StoreConfig]
-    """Configuration for vector embeddings in store."""
+    """Optional. Configuration for the built-in long-term memory store, including semantic search indexing.
+    
+    If omitted, no vector index is set up (the object store will still be present, however).
+    """
 
     auth: Optional[AuthConfig]
-    """Configuration for authentication."""
+    """Optional. Custom authentication config, including the path to your Python auth logic and 
+    the OpenAPI security definitions it uses.
+    """
 
     http: Optional[HttpConfig]
-    """Configuration for HTTP server."""
+    """Optional. Configuration for the built-in HTTP server, controlling which custom routes are exposed
+    and how cross-origin requests are handled.
+    """
+
+    ui: Optional[dict[str, str]]
+    """Optional. Named definitions of UI components emitted by the agent, each pointing to a JS/TS file.
+    """
 
 
 def _parse_version(version_str: str) -> tuple[int, int]:
@@ -189,6 +404,7 @@ def validate_config(config: Config) -> Config:
             "store": config.get("store"),
             "auth": config.get("auth"),
             "http": config.get("http"),
+            "ui": config.get("ui"),
         }
         if config.get("node_version")
         else {
@@ -201,6 +417,7 @@ def validate_config(config: Config) -> Config:
             "store": config.get("store"),
             "auth": config.get("auth"),
             "http": config.get("http"),
+            "ui": config.get("ui"),
         }
     )
 
@@ -687,9 +904,11 @@ def python_config_to_docker(
     pip_pkgs_str = f"RUN {pip_install} {' '.join(pypi_deps)}" if pypi_deps else ""
     if local_deps.pip_reqs:
         pip_reqs_str = os.linesep.join(
-            f"COPY --from=__outer_{reqpath.name} requirements.txt {destpath}"
-            if reqpath.parent in local_deps.additional_contexts
-            else f"ADD {reqpath.relative_to(config_path.parent)} {destpath}"
+            (
+                f"COPY --from=__outer_{reqpath.name} requirements.txt {destpath}"
+                if reqpath.parent in local_deps.additional_contexts
+                else f"ADD {reqpath.relative_to(config_path.parent)} {destpath}"
+            )
             for reqpath, destpath in local_deps.pip_reqs
         )
         pip_reqs_str += f'{os.linesep}RUN {pip_install} {" ".join("-r " + r for _,r in local_deps.pip_reqs)}'
@@ -724,13 +943,15 @@ RUN set -ex && \\
     )
 
     local_pkgs_str = os.linesep.join(
-        f"""# -- Adding local package {relpath} --
+        (
+            f"""# -- Adding local package {relpath} --
 COPY --from={name} . /deps/{name}
 # -- End of local package {relpath} --"""
-        if fullpath in local_deps.additional_contexts
-        else f"""# -- Adding local package {relpath} --
+            if fullpath in local_deps.additional_contexts
+            else f"""# -- Adding local package {relpath} --
 ADD {relpath} /deps/{name}
 # -- End of local package {relpath} --"""
+        )
         for fullpath, (relpath, name) in local_deps.real_pkgs.items()
     )
 
@@ -845,6 +1066,7 @@ ADD . {faux_path}
 RUN cd {faux_path} && {install_cmd}
 {env_additional_config}
 ENV LANGSERVE_GRAPHS='{json.dumps(config["graphs"])}'
+{f"ENV LANGGRAPH_UI='{json.dumps(config['ui'])}'" if config.get("ui") else ""}
 
 WORKDIR {faux_path}
 

libs/cli/poetry.lock

@@ -707,6 +707,58 @@ files = [
     {file = "msgpack-1.1.0.tar.gz", hash = "sha256:dd432ccc2c72b914e4cb77afce64aab761c1137cc698be3984eee260bcb2896e"},
 ]
 
+[[package]]
+name = "msgspec"
+version = "0.19.0"
+description = "A fast serialization and validation library, with builtin support for JSON, MessagePack, YAML, and TOML."
+optional = false
+python-versions = ">=3.9"
+files = [
+    {file = "msgspec-0.19.0-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:d8dd848ee7ca7c8153462557655570156c2be94e79acec3561cf379581343259"},
+    {file = "msgspec-0.19.0-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:0553bbc77662e5708fe66aa75e7bd3e4b0f209709c48b299afd791d711a93c36"},
+    {file = "msgspec-0.19.0-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:fe2c4bf29bf4e89790b3117470dea2c20b59932772483082c468b990d45fb947"},
+    {file = "msgspec-0.19.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:00e87ecfa9795ee5214861eab8326b0e75475c2e68a384002aa135ea2a27d909"},
+    {file = "msgspec-0.19.0-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:3c4ec642689da44618f68c90855a10edbc6ac3ff7c1d94395446c65a776e712a"},
+    {file = "msgspec-0.19.0-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:2719647625320b60e2d8af06b35f5b12d4f4d281db30a15a1df22adb2295f633"},
+    {file = "msgspec-0.19.0-cp310-cp310-win_amd64.whl", hash = "sha256:695b832d0091edd86eeb535cd39e45f3919f48d997685f7ac31acb15e0a2ed90"},
+    {file = "msgspec-0.19.0-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:aa77046904db764b0462036bc63ef71f02b75b8f72e9c9dd4c447d6da1ed8f8e"},
+    {file = "msgspec-0.19.0-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:047cfa8675eb3bad68722cfe95c60e7afabf84d1bd8938979dd2b92e9e4a9551"},
+    {file = "msgspec-0.19.0-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:e78f46ff39a427e10b4a61614a2777ad69559cc8d603a7c05681f5a595ea98f7"},
+    {file = "msgspec-0.19.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6c7adf191e4bd3be0e9231c3b6dc20cf1199ada2af523885efc2ed218eafd011"},
+    {file = "msgspec-0.19.0-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:f04cad4385e20be7c7176bb8ae3dca54a08e9756cfc97bcdb4f18560c3042063"},
+    {file = "msgspec-0.19.0-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:45c8fb410670b3b7eb884d44a75589377c341ec1392b778311acdbfa55187716"},
+    {file = "msgspec-0.19.0-cp311-cp311-win_amd64.whl", hash = "sha256:70eaef4934b87193a27d802534dc466778ad8d536e296ae2f9334e182ac27b6c"},
+    {file = "msgspec-0.19.0-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:f98bd8962ad549c27d63845b50af3f53ec468b6318400c9f1adfe8b092d7b62f"},
+    {file = "msgspec-0.19.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:43bbb237feab761b815ed9df43b266114203f53596f9b6e6f00ebd79d178cdf2"},
+    {file = "msgspec-0.19.0-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:4cfc033c02c3e0aec52b71710d7f84cb3ca5eb407ab2ad23d75631153fdb1f12"},
+    {file = "msgspec-0.19.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:d911c442571605e17658ca2b416fd8579c5050ac9adc5e00c2cb3126c97f73bc"},
+    {file = "msgspec-0.19.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:757b501fa57e24896cf40a831442b19a864f56d253679f34f260dcb002524a6c"},
+    {file = "msgspec-0.19.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:5f0f65f29b45e2816d8bded36e6b837a4bf5fb60ec4bc3c625fa2c6da4124537"},
+    {file = "msgspec-0.19.0-cp312-cp312-win_amd64.whl", hash = "sha256:067f0de1c33cfa0b6a8206562efdf6be5985b988b53dd244a8e06f993f27c8c0"},
+    {file = "msgspec-0.19.0-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:f12d30dd6266557aaaf0aa0f9580a9a8fbeadfa83699c487713e355ec5f0bd86"},
+    {file = "msgspec-0.19.0-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:82b2c42c1b9ebc89e822e7e13bbe9d17ede0c23c187469fdd9505afd5a481314"},
+    {file = "msgspec-0.19.0-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:19746b50be214a54239aab822964f2ac81e38b0055cca94808359d779338c10e"},
+    {file = "msgspec-0.19.0-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:60ef4bdb0ec8e4ad62e5a1f95230c08efb1f64f32e6e8dd2ced685bcc73858b5"},
+    {file = "msgspec-0.19.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:ac7f7c377c122b649f7545810c6cd1b47586e3aa3059126ce3516ac7ccc6a6a9"},
+    {file = "msgspec-0.19.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:a5bc1472223a643f5ffb5bf46ccdede7f9795078194f14edd69e3aab7020d327"},
+    {file = "msgspec-0.19.0-cp313-cp313-win_amd64.whl", hash = "sha256:317050bc0f7739cb30d257ff09152ca309bf5a369854bbf1e57dffc310c1f20f"},
+    {file = "msgspec-0.19.0-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:15c1e86fff77184c20a2932cd9742bf33fe23125fa3fcf332df9ad2f7d483044"},
+    {file = "msgspec-0.19.0-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:3b5541b2b3294e5ffabe31a09d604e23a88533ace36ac288fa32a420aa38d229"},
+    {file = "msgspec-0.19.0-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0f5c043ace7962ef188746e83b99faaa9e3e699ab857ca3f367b309c8e2c6b12"},
+    {file = "msgspec-0.19.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ca06aa08e39bf57e39a258e1996474f84d0dd8130d486c00bec26d797b8c5446"},
+    {file = "msgspec-0.19.0-cp39-cp39-musllinux_1_2_aarch64.whl", hash = "sha256:e695dad6897896e9384cf5e2687d9ae9feaef50e802f93602d35458e20d1fb19"},
+    {file = "msgspec-0.19.0-cp39-cp39-musllinux_1_2_x86_64.whl", hash = "sha256:3be5c02e1fee57b54130316a08fe40cca53af92999a302a6054cd451700ea7db"},
+    {file = "msgspec-0.19.0-cp39-cp39-win_amd64.whl", hash = "sha256:0684573a821be3c749912acf5848cce78af4298345cb2d7a8b8948a0a5a27cfe"},
+    {file = "msgspec-0.19.0.tar.gz", hash = "sha256:604037e7cd475345848116e89c553aa9a233259733ab51986ac924ab1b976f8e"},
+]
+
+[package.extras]
+dev = ["attrs", "coverage", "eval-type-backport", "furo", "ipython", "msgpack", "mypy", "pre-commit", "pyright", "pytest", "pyyaml", "sphinx", "sphinx-copybutton", "sphinx-design", "tomli", "tomli_w"]
+doc = ["furo", "ipython", "sphinx", "sphinx-copybutton", "sphinx-design"]
+test = ["attrs", "eval-type-backport", "msgpack", "pytest", "pyyaml", "tomli", "tomli_w"]
+toml = ["tomli", "tomli_w"]
+yaml = ["pyyaml"]
+
 [[package]]
 name = "mypy"
 version = "1.15.0"
@@ -1665,4 +1717,4 @@ inmem = ["langgraph-api", "python-dotenv"]
 [metadata]
 lock-version = "2.0"
 python-versions = "^3.9.0,<4.0"
-content-hash = "04a0b0e203ae00f30cca7e454c632be0b41cbb4820d0c05315a346d514b5f98e"
+content-hash = "d0e2bdcb600ad031867413025fcc58bb162609209359d63ca99a77060cf8cbb4"

libs/cli/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langgraph-cli"
-version = "0.1.74"
+version = "0.1.77"
 description = "CLI for interacting with LangGraph API"
 authors = []
 license = "MIT"
@@ -25,6 +25,7 @@ pytest-asyncio = "^0.21.1"
 pytest-mock = "^3.11.1"
 pytest-watch = "^4.2.0"
 mypy = "^1.10.0"
+msgspec = "^0.19.0"
 
 [tool.poetry.extras]
 inmem = ["langgraph-api", "python-dotenv"]

libs/cli/schemas/schema.json

@@ -0,0 +1,480 @@
+{
+  "$ref": "#/$defs/Config",
+  "$defs": {
+    "Config": {
+      "title": "Config",
+      "description": "Top-level config for langgraph-cli or similar deployment tooling.",
+      "type": "object",
+      "required": [],
+      "oneOf": [
+        {
+          "type": "object",
+          "properties": {
+            "python_version": {
+              "type": "string",
+              "description": "Optional. Python version in 'major.minor' format (e.g. '3.11').\nMust be at least 3.11 or greater for this deployment to function properly.\n",
+              "enum": [
+                "3.11",
+                "3.12"
+              ]
+            },
+            "pip_config_file": {
+              "anyOf": [
+                {
+                  "type": "string"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Path to a pip config file (e.g., \"/etc/pip.conf\" or \"pip.ini\") for controlling\npackage installation (custom indices, credentials, etc.).\n\nOnly relevant if Python dependencies are installed via pip. If omitted, default pip settings are used.\n"
+            },
+            "auth": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/AuthConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Custom authentication config, including the path to your Python auth logic and\nthe OpenAPI security definitions it uses.\n"
+            },
+            "dependencies": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "List of Python dependencies to install, either from PyPI or local paths.\n"
+            },
+            "dockerfile_lines": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "Optional. Additional Docker instructions that will be appended to your base Dockerfile.\n\nUseful for installing OS packages, setting environment variables, etc."
+            },
+            "env": {
+              "anyOf": [
+                {
+                  "type": "object",
+                  "additionalProperties": {
+                    "type": "string"
+                  }
+                },
+                {
+                  "type": "string"
+                }
+              ],
+              "description": "Optional. Environment variables to set for your deployment.\n\n- If given as a dict, keys are variable names and values are their values.\n- If given as a string, it must be a path to a file containing lines in KEY=VALUE format.\n\nenv=\".env\n"
+            },
+            "graphs": {
+              "type": "object",
+              "additionalProperties": {
+                "type": "string"
+              },
+              "description": "Optional. Named definitions of graphs, each pointing to a Python object.\n\n\nGraphs can be StateGraph, @entrypoint, or any other Pregel object OR they can point to (async) context\nmanagers that accept a single configuration argument (of type RunnableConfig) and return a pregel object\n(instance of Stategraph, etc.).\n"
+            },
+            "http": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/HttpConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Configuration for the built-in HTTP server, controlling which custom routes are exposed\nand how cross-origin requests are handled.\n"
+            },
+            "store": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/StoreConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Configuration for the built-in long-term memory store, including semantic search indexing.\n\nIf omitted, no vector index is set up (the object store will still be present, however).\n"
+            },
+            "ui": {
+              "anyOf": [
+                {
+                  "type": "object",
+                  "additionalProperties": {
+                    "type": "string"
+                  }
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Named definitions of UI components emitted by the agent, each pointing to a JS/TS file.\n"
+            }
+          },
+          "required": [
+            "dependencies",
+            "graphs"
+          ]
+        },
+        {
+          "type": "object",
+          "properties": {
+            "node_version": {
+              "anyOf": [
+                {
+                  "type": "string",
+                  "enum": [
+                    "20"
+                  ]
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Node.js version as a major version (e.g. '20'), if your deployment needs Node.\nMust be >= 20 if provided.\n"
+            },
+            "auth": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/AuthConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Custom authentication config, including the path to your Python auth logic and\nthe OpenAPI security definitions it uses.\n"
+            },
+            "dependencies": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "List of Python dependencies to install, either from PyPI or local paths.\n"
+            },
+            "dockerfile_lines": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "Optional. Additional Docker instructions that will be appended to your base Dockerfile.\n\nUseful for installing OS packages, setting environment variables, etc."
+            },
+            "env": {
+              "anyOf": [
+                {
+                  "type": "object",
+                  "additionalProperties": {
+                    "type": "string"
+                  }
+                },
+                {
+                  "type": "string"
+                }
+              ],
+              "description": "Optional. Environment variables to set for your deployment.\n\n- If given as a dict, keys are variable names and values are their values.\n- If given as a string, it must be a path to a file containing lines in KEY=VALUE format.\n\nenv=\".env\n"
+            },
+            "graphs": {
+              "type": "object",
+              "additionalProperties": {
+                "type": "string"
+              },
+              "description": "Optional. Named definitions of graphs, each pointing to a Python object.\n\n\nGraphs can be StateGraph, @entrypoint, or any other Pregel object OR they can point to (async) context\nmanagers that accept a single configuration argument (of type RunnableConfig) and return a pregel object\n(instance of Stategraph, etc.).\n"
+            },
+            "http": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/HttpConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Configuration for the built-in HTTP server, controlling which custom routes are exposed\nand how cross-origin requests are handled.\n"
+            },
+            "store": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/StoreConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Configuration for the built-in long-term memory store, including semantic search indexing.\n\nIf omitted, no vector index is set up (the object store will still be present, however).\n"
+            },
+            "ui": {
+              "anyOf": [
+                {
+                  "type": "object",
+                  "additionalProperties": {
+                    "type": "string"
+                  }
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Named definitions of UI components emitted by the agent, each pointing to a JS/TS file.\n"
+            }
+          },
+          "required": [
+            "node_version",
+            "graphs"
+          ]
+        }
+      ]
+    },
+    "AuthConfig": {
+      "title": "AuthConfig",
+      "description": "Configuration for custom authentication logic and how it integrates into the OpenAPI spec.",
+      "type": "object",
+      "properties": {
+        "disable_studio_auth": {
+          "type": "boolean",
+          "description": "Optional. Whether to disable LangSmith API-key authentication for requests originating the Studio.\n\nDefaults to False, meaning that if a particular header is set, the server will verify the `x-api-key` header\nvalue is a valid API key for the deployment's workspace. If True, all requests will go through your custom\nauthentication logic, regardless of origin of the request.\n"
+        },
+        "openapi": {
+          "$ref": "#/$defs/SecurityConfig",
+          "description": "Required. Detailed security configuration that merges into your deployment's OpenAPI spec.\n\n{\n}\n}\n}\n},\n]\n}\n"
+        },
+        "path": {
+          "type": "string",
+          "description": "Required. Path to an instance of the Auth() class that implements custom authentication.\n"
+        }
+      },
+      "required": []
+    },
+    "SecurityConfig": {
+      "title": "SecurityConfig",
+      "description": "Configuration for OpenAPI security definitions and requirements.\n\nUseful for specifying global or path-level authentication and authorization flows\n(e.g., OAuth2, API key headers, etc.).",
+      "type": "object",
+      "properties": {
+        "paths": {
+          "type": "object",
+          "additionalProperties": {
+            "type": "object",
+            "additionalProperties": {
+              "type": "array",
+              "items": {
+                "type": "object",
+                "additionalProperties": {
+                  "type": "array",
+                  "items": {
+                    "type": "string"
+                  }
+                }
+              }
+            }
+          },
+          "description": "Optional. Path-specific security overrides.\n\n- Keys that are HTTP methods (e.g., \"GET\", \"POST\"),\n- Values are lists of security definitions (just like `security`) for that method.\n"
+        },
+        "security": {
+          "type": "array",
+          "items": {
+            "type": "object",
+            "additionalProperties": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              }
+            }
+          },
+          "description": "Optional. Global security requirements across all endpoints.\n\nEach element in the list maps a security scheme (e.g. \"OAuth2\") to a list of scopes (e.g. [\"read\", \"write\"])."
+        },
+        "securitySchemes": {
+          "type": "object",
+          "additionalProperties": {
+            "type": "object"
+          },
+          "description": "Required. Dict describing each security scheme recognized by your OpenAPI spec.\n\nKeys are scheme names (e.g. \"OAuth2\", \"ApiKeyAuth\") and values are their definitions."
+        }
+      },
+      "required": []
+    },
+    "HttpConfig": {
+      "title": "HttpConfig",
+      "description": "Configuration for the built-in HTTP server that powers your deployment's routes and endpoints.",
+      "type": "object",
+      "properties": {
+        "app": {
+          "type": "string",
+          "description": "Optional. Import path to a custom Starlette/FastAPI application to mount.\n"
+        },
+        "cors": {
+          "anyOf": [
+            {
+              "$ref": "#/$defs/CorsConfig"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "description": "Optional. Defines CORS restrictions. If omitted, no special rules are set and\ncross-origin behavior depends on default server settings.\n"
+        },
+        "disable_assistants": {
+          "type": "boolean",
+          "description": "Optional. If True, /assistants routes are removed from the server.\n\nDefault is False (meaning /assistants is enabled).\n"
+        },
+        "disable_meta": {
+          "type": "boolean",
+          "description": "Optional. If True, all meta endpoints (/ok, /info, /metrics, /docs) are disabled.\n\nDefault is False.\n"
+        },
+        "disable_runs": {
+          "type": "boolean",
+          "description": "Optional. If True, /runs routes are removed.\n\nDefault is False.\n"
+        },
+        "disable_store": {
+          "type": "boolean",
+          "description": "Optional. If True, /store routes are removed, disabling direct store interactions via HTTP.\n\nDefault is False.\n"
+        },
+        "disable_threads": {
+          "type": "boolean",
+          "description": "Optional. If True, /threads routes are removed.\n\nDefault is False.\n"
+        }
+      },
+      "required": []
+    },
+    "CorsConfig": {
+      "title": "CorsConfig",
+      "description": "Specifies Cross-Origin Resource Sharing (CORS) rules for your server.\n\nIf omitted, defaults are typically very restrictive (often no cross-origin requests).\nConfigure carefully if you want to allow usage from browsers hosted on other domains.",
+      "type": "object",
+      "properties": {
+        "allow_credentials": {
+          "type": "boolean",
+          "description": "Optional. If True, cross-origin requests can include credentials (cookies, auth headers).\n\nDefault False to avoid accidentally exposing secured endpoints to untrusted sites.\n"
+        },
+        "allow_headers": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          },
+          "description": "Optional. HTTP headers that can be used in cross-origin requests (e.g. [\"Content-Type\", \"Authorization\"])."
+        },
+        "allow_methods": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          },
+          "description": "Optional. HTTP methods permitted for cross-origin requests (e.g. [\"GET\", \"POST\"]).\n\nDefault might be [\"GET\", \"POST\", \"OPTIONS\"] depending on your server framework.\n"
+        },
+        "allow_origin_regex": {
+          "type": "string",
+          "description": "Optional. A regex pattern for matching allowed origins, used if you have dynamic subdomains.\n"
+        },
+        "allow_origins": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          },
+          "description": "Optional. List of allowed origins (e.g., \"https://example.com\").\n\nDefault is often an empty list (no external origins).\nUse \"*\" only if you trust all origins, as that bypasses most restrictions.\n"
+        },
+        "expose_headers": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          },
+          "description": "Optional. List of headers that browsers are allowed to read from the response in cross-origin contexts."
+        },
+        "max_age": {
+          "type": "integer",
+          "description": "Optional. How many seconds the browser may cache preflight responses.\n\nDefault might be 600 (10 minutes). Larger values reduce preflight requests but can cause stale configurations.\n"
+        }
+      },
+      "required": []
+    },
+    "StoreConfig": {
+      "title": "StoreConfig",
+      "description": "Configuration for the built-in long-term memory store.\n\nThis store can optionally perform semantic search. If you omit `index`,\nthe store will just handle traditional (non-embedded) data without vector lookups.",
+      "type": "object",
+      "properties": {
+        "index": {
+          "anyOf": [
+            {
+              "$ref": "#/$defs/IndexConfig"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "description": "Optional. Defines the vector-based semantic search configuration.\n\n- Generate embeddings according to `index.embed`\n- Enforce the embedding dimension given by `index.dims`\n- Embed only specified JSON fields (if any) from `index.fields`\n\nIf omitted, no vector index is initialized.\n"
+        },
+        "ttl": {
+          "anyOf": [
+            {
+              "$ref": "#/$defs/TTLConfig"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "description": "Optional. Defines the TTL (time-to-live) behavior configuration.\n\nIf provided, the store will apply TTL settings according to the configuration.\nIf omitted, no TTL behavior is configured.\n"
+        }
+      },
+      "required": []
+    },
+    "IndexConfig": {
+      "title": "IndexConfig",
+      "description": "Configuration for indexing documents for semantic search in the store.\n\nThis governs how text is converted into embeddings and stored for vector-based lookups.",
+      "type": "object",
+      "properties": {
+        "dims": {
+          "type": "integer",
+          "description": "Required. Dimensionality of the embedding vectors you will store.\n\nMust match the output dimension of your selected embedding model or custom embed function.\nIf mismatched, you will likely encounter shape/size errors when inserting or querying vectors.\n\n"
+        },
+        "embed": {
+          "type": "string",
+          "description": "Required. Identifier or reference to the embedding model or a custom embedding function.\n\n- \"my_custom_embed\" if it's a known alias in your system\n"
+        },
+        "fields": {
+          "anyOf": [
+            {
+              "type": "array",
+              "items": {
+                "type": "string"
+              }
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "description": "Optional. List of JSON fields to extract before generating embeddings.\n\nDefaults to [\"$\"], which means the entire JSON object is embedded as one piece of text.\nIf you provide multiple fields (e.g. [\"title\", \"content\"]), each is extracted and embedded separately,\noften saving token usage if you only care about certain parts of the data.\n"
+        }
+      },
+      "required": []
+    },
+    "TTLConfig": {
+      "title": "TTLConfig",
+      "description": "Configuration for TTL (time-to-live) behavior in the store.",
+      "type": "object",
+      "properties": {
+        "default_ttl": {
+          "anyOf": [
+            {
+              "type": "number"
+            },
+            {
+              "type": "null"
+            }
+          ]
+        },
+        "refresh_on_read": {
+          "type": "boolean"
+        },
+        "sweep_interval_minutes": {
+          "anyOf": [
+            {
+              "type": "integer"
+            },
+            {
+              "type": "null"
+            }
+          ]
+        }
+      },
+      "required": []
+    }
+  },
+  "title": "LangGraph CLI Configuration",
+  "description": "Configuration schema for langgraph-cli",
+  "version": "v0"
+}

libs/cli/schemas/schema.v0.json

@@ -0,0 +1,480 @@
+{
+  "$ref": "#/$defs/Config",
+  "$defs": {
+    "Config": {
+      "title": "Config",
+      "description": "Top-level config for langgraph-cli or similar deployment tooling.",
+      "type": "object",
+      "required": [],
+      "oneOf": [
+        {
+          "type": "object",
+          "properties": {
+            "python_version": {
+              "type": "string",
+              "description": "Optional. Python version in 'major.minor' format (e.g. '3.11').\nMust be at least 3.11 or greater for this deployment to function properly.\n",
+              "enum": [
+                "3.11",
+                "3.12"
+              ]
+            },
+            "pip_config_file": {
+              "anyOf": [
+                {
+                  "type": "string"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Path to a pip config file (e.g., \"/etc/pip.conf\" or \"pip.ini\") for controlling\npackage installation (custom indices, credentials, etc.).\n\nOnly relevant if Python dependencies are installed via pip. If omitted, default pip settings are used.\n"
+            },
+            "auth": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/AuthConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Custom authentication config, including the path to your Python auth logic and\nthe OpenAPI security definitions it uses.\n"
+            },
+            "dependencies": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "List of Python dependencies to install, either from PyPI or local paths.\n"
+            },
+            "dockerfile_lines": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "Optional. Additional Docker instructions that will be appended to your base Dockerfile.\n\nUseful for installing OS packages, setting environment variables, etc."
+            },
+            "env": {
+              "anyOf": [
+                {
+                  "type": "object",
+                  "additionalProperties": {
+                    "type": "string"
+                  }
+                },
+                {
+                  "type": "string"
+                }
+              ],
+              "description": "Optional. Environment variables to set for your deployment.\n\n- If given as a dict, keys are variable names and values are their values.\n- If given as a string, it must be a path to a file containing lines in KEY=VALUE format.\n\nenv=\".env\n"
+            },
+            "graphs": {
+              "type": "object",
+              "additionalProperties": {
+                "type": "string"
+              },
+              "description": "Optional. Named definitions of graphs, each pointing to a Python object.\n\n\nGraphs can be StateGraph, @entrypoint, or any other Pregel object OR they can point to (async) context\nmanagers that accept a single configuration argument (of type RunnableConfig) and return a pregel object\n(instance of Stategraph, etc.).\n"
+            },
+            "http": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/HttpConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Configuration for the built-in HTTP server, controlling which custom routes are exposed\nand how cross-origin requests are handled.\n"
+            },
+            "store": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/StoreConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Configuration for the built-in long-term memory store, including semantic search indexing.\n\nIf omitted, no vector index is set up (the object store will still be present, however).\n"
+            },
+            "ui": {
+              "anyOf": [
+                {
+                  "type": "object",
+                  "additionalProperties": {
+                    "type": "string"
+                  }
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Named definitions of UI components emitted by the agent, each pointing to a JS/TS file.\n"
+            }
+          },
+          "required": [
+            "dependencies",
+            "graphs"
+          ]
+        },
+        {
+          "type": "object",
+          "properties": {
+            "node_version": {
+              "anyOf": [
+                {
+                  "type": "string",
+                  "enum": [
+                    "20"
+                  ]
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Node.js version as a major version (e.g. '20'), if your deployment needs Node.\nMust be >= 20 if provided.\n"
+            },
+            "auth": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/AuthConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Custom authentication config, including the path to your Python auth logic and\nthe OpenAPI security definitions it uses.\n"
+            },
+            "dependencies": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "List of Python dependencies to install, either from PyPI or local paths.\n"
+            },
+            "dockerfile_lines": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "Optional. Additional Docker instructions that will be appended to your base Dockerfile.\n\nUseful for installing OS packages, setting environment variables, etc."
+            },
+            "env": {
+              "anyOf": [
+                {
+                  "type": "object",
+                  "additionalProperties": {
+                    "type": "string"
+                  }
+                },
+                {
+                  "type": "string"
+                }
+              ],
+              "description": "Optional. Environment variables to set for your deployment.\n\n- If given as a dict, keys are variable names and values are their values.\n- If given as a string, it must be a path to a file containing lines in KEY=VALUE format.\n\nenv=\".env\n"
+            },
+            "graphs": {
+              "type": "object",
+              "additionalProperties": {
+                "type": "string"
+              },
+              "description": "Optional. Named definitions of graphs, each pointing to a Python object.\n\n\nGraphs can be StateGraph, @entrypoint, or any other Pregel object OR they can point to (async) context\nmanagers that accept a single configuration argument (of type RunnableConfig) and return a pregel object\n(instance of Stategraph, etc.).\n"
+            },
+            "http": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/HttpConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Configuration for the built-in HTTP server, controlling which custom routes are exposed\nand how cross-origin requests are handled.\n"
+            },
+            "store": {
+              "anyOf": [
+                {
+                  "$ref": "#/$defs/StoreConfig"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Configuration for the built-in long-term memory store, including semantic search indexing.\n\nIf omitted, no vector index is set up (the object store will still be present, however).\n"
+            },
+            "ui": {
+              "anyOf": [
+                {
+                  "type": "object",
+                  "additionalProperties": {
+                    "type": "string"
+                  }
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "description": "Optional. Named definitions of UI components emitted by the agent, each pointing to a JS/TS file.\n"
+            }
+          },
+          "required": [
+            "node_version",
+            "graphs"
+          ]
+        }
+      ]
+    },
+    "AuthConfig": {
+      "title": "AuthConfig",
+      "description": "Configuration for custom authentication logic and how it integrates into the OpenAPI spec.",
+      "type": "object",
+      "properties": {
+        "disable_studio_auth": {
+          "type": "boolean",
+          "description": "Optional. Whether to disable LangSmith API-key authentication for requests originating the Studio.\n\nDefaults to False, meaning that if a particular header is set, the server will verify the `x-api-key` header\nvalue is a valid API key for the deployment's workspace. If True, all requests will go through your custom\nauthentication logic, regardless of origin of the request.\n"
+        },
+        "openapi": {
+          "$ref": "#/$defs/SecurityConfig",
+          "description": "Required. Detailed security configuration that merges into your deployment's OpenAPI spec.\n\n{\n}\n}\n}\n},\n]\n}\n"
+        },
+        "path": {
+          "type": "string",
+          "description": "Required. Path to an instance of the Auth() class that implements custom authentication.\n"
+        }
+      },
+      "required": []
+    },
+    "SecurityConfig": {
+      "title": "SecurityConfig",
+      "description": "Configuration for OpenAPI security definitions and requirements.\n\nUseful for specifying global or path-level authentication and authorization flows\n(e.g., OAuth2, API key headers, etc.).",
+      "type": "object",
+      "properties": {
+        "paths": {
+          "type": "object",
+          "additionalProperties": {
+            "type": "object",
+            "additionalProperties": {
+              "type": "array",
+              "items": {
+                "type": "object",
+                "additionalProperties": {
+                  "type": "array",
+                  "items": {
+                    "type": "string"
+                  }
+                }
+              }
+            }
+          },
+          "description": "Optional. Path-specific security overrides.\n\n- Keys that are HTTP methods (e.g., \"GET\", \"POST\"),\n- Values are lists of security definitions (just like `security`) for that method.\n"
+        },
+        "security": {
+          "type": "array",
+          "items": {
+            "type": "object",
+            "additionalProperties": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              }
+            }
+          },
+          "description": "Optional. Global security requirements across all endpoints.\n\nEach element in the list maps a security scheme (e.g. \"OAuth2\") to a list of scopes (e.g. [\"read\", \"write\"])."
+        },
+        "securitySchemes": {
+          "type": "object",
+          "additionalProperties": {
+            "type": "object"
+          },
+          "description": "Required. Dict describing each security scheme recognized by your OpenAPI spec.\n\nKeys are scheme names (e.g. \"OAuth2\", \"ApiKeyAuth\") and values are their definitions."
+        }
+      },
+      "required": []
+    },
+    "HttpConfig": {
+      "title": "HttpConfig",
+      "description": "Configuration for the built-in HTTP server that powers your deployment's routes and endpoints.",
+      "type": "object",
+      "properties": {
+        "app": {
+          "type": "string",
+          "description": "Optional. Import path to a custom Starlette/FastAPI application to mount.\n"
+        },
+        "cors": {
+          "anyOf": [
+            {
+              "$ref": "#/$defs/CorsConfig"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "description": "Optional. Defines CORS restrictions. If omitted, no special rules are set and\ncross-origin behavior depends on default server settings.\n"
+        },
+        "disable_assistants": {
+          "type": "boolean",
+          "description": "Optional. If True, /assistants routes are removed from the server.\n\nDefault is False (meaning /assistants is enabled).\n"
+        },
+        "disable_meta": {
+          "type": "boolean",
+          "description": "Optional. If True, all meta endpoints (/ok, /info, /metrics, /docs) are disabled.\n\nDefault is False.\n"
+        },
+        "disable_runs": {
+          "type": "boolean",
+          "description": "Optional. If True, /runs routes are removed.\n\nDefault is False.\n"
+        },
+        "disable_store": {
+          "type": "boolean",
+          "description": "Optional. If True, /store routes are removed, disabling direct store interactions via HTTP.\n\nDefault is False.\n"
+        },
+        "disable_threads": {
+          "type": "boolean",
+          "description": "Optional. If True, /threads routes are removed.\n\nDefault is False.\n"
+        }
+      },
+      "required": []
+    },
+    "CorsConfig": {
+      "title": "CorsConfig",
+      "description": "Specifies Cross-Origin Resource Sharing (CORS) rules for your server.\n\nIf omitted, defaults are typically very restrictive (often no cross-origin requests).\nConfigure carefully if you want to allow usage from browsers hosted on other domains.",
+      "type": "object",
+      "properties": {
+        "allow_credentials": {
+          "type": "boolean",
+          "description": "Optional. If True, cross-origin requests can include credentials (cookies, auth headers).\n\nDefault False to avoid accidentally exposing secured endpoints to untrusted sites.\n"
+        },
+        "allow_headers": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          },
+          "description": "Optional. HTTP headers that can be used in cross-origin requests (e.g. [\"Content-Type\", \"Authorization\"])."
+        },
+        "allow_methods": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          },
+          "description": "Optional. HTTP methods permitted for cross-origin requests (e.g. [\"GET\", \"POST\"]).\n\nDefault might be [\"GET\", \"POST\", \"OPTIONS\"] depending on your server framework.\n"
+        },
+        "allow_origin_regex": {
+          "type": "string",
+          "description": "Optional. A regex pattern for matching allowed origins, used if you have dynamic subdomains.\n"
+        },
+        "allow_origins": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          },
+          "description": "Optional. List of allowed origins (e.g., \"https://example.com\").\n\nDefault is often an empty list (no external origins).\nUse \"*\" only if you trust all origins, as that bypasses most restrictions.\n"
+        },
+        "expose_headers": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          },
+          "description": "Optional. List of headers that browsers are allowed to read from the response in cross-origin contexts."
+        },
+        "max_age": {
+          "type": "integer",
+          "description": "Optional. How many seconds the browser may cache preflight responses.\n\nDefault might be 600 (10 minutes). Larger values reduce preflight requests but can cause stale configurations.\n"
+        }
+      },
+      "required": []
+    },
+    "StoreConfig": {
+      "title": "StoreConfig",
+      "description": "Configuration for the built-in long-term memory store.\n\nThis store can optionally perform semantic search. If you omit `index`,\nthe store will just handle traditional (non-embedded) data without vector lookups.",
+      "type": "object",
+      "properties": {
+        "index": {
+          "anyOf": [
+            {
+              "$ref": "#/$defs/IndexConfig"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "description": "Optional. Defines the vector-based semantic search configuration.\n\n- Generate embeddings according to `index.embed`\n- Enforce the embedding dimension given by `index.dims`\n- Embed only specified JSON fields (if any) from `index.fields`\n\nIf omitted, no vector index is initialized.\n"
+        },
+        "ttl": {
+          "anyOf": [
+            {
+              "$ref": "#/$defs/TTLConfig"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "description": "Optional. Defines the TTL (time-to-live) behavior configuration.\n\nIf provided, the store will apply TTL settings according to the configuration.\nIf omitted, no TTL behavior is configured.\n"
+        }
+      },
+      "required": []
+    },
+    "IndexConfig": {
+      "title": "IndexConfig",
+      "description": "Configuration for indexing documents for semantic search in the store.\n\nThis governs how text is converted into embeddings and stored for vector-based lookups.",
+      "type": "object",
+      "properties": {
+        "dims": {
+          "type": "integer",
+          "description": "Required. Dimensionality of the embedding vectors you will store.\n\nMust match the output dimension of your selected embedding model or custom embed function.\nIf mismatched, you will likely encounter shape/size errors when inserting or querying vectors.\n\n"
+        },
+        "embed": {
+          "type": "string",
+          "description": "Required. Identifier or reference to the embedding model or a custom embedding function.\n\n- \"my_custom_embed\" if it's a known alias in your system\n"
+        },
+        "fields": {
+          "anyOf": [
+            {
+              "type": "array",
+              "items": {
+                "type": "string"
+              }
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "description": "Optional. List of JSON fields to extract before generating embeddings.\n\nDefaults to [\"$\"], which means the entire JSON object is embedded as one piece of text.\nIf you provide multiple fields (e.g. [\"title\", \"content\"]), each is extracted and embedded separately,\noften saving token usage if you only care about certain parts of the data.\n"
+        }
+      },
+      "required": []
+    },
+    "TTLConfig": {
+      "title": "TTLConfig",
+      "description": "Configuration for TTL (time-to-live) behavior in the store.",
+      "type": "object",
+      "properties": {
+        "default_ttl": {
+          "anyOf": [
+            {
+              "type": "number"
+            },
+            {
+              "type": "null"
+            }
+          ]
+        },
+        "refresh_on_read": {
+          "type": "boolean"
+        },
+        "sweep_interval_minutes": {
+          "anyOf": [
+            {
+              "type": "integer"
+            },
+            {
+              "type": "null"
+            }
+          ]
+        }
+      },
+      "required": []
+    }
+  },
+  "title": "LangGraph CLI Configuration",
+  "description": "Configuration schema for langgraph-cli",
+  "version": "v0"
+}

libs/cli/tests/unit_tests/test_config.py

@@ -33,6 +33,7 @@ def test_validate_config():
         "store": None,
         "auth": None,
         "http": None,
+        "ui": None,
         **expected_config,
     }
     actual_config = validate_config(expected_config)
@@ -52,6 +53,7 @@ def test_validate_config():
         "store": None,
         "auth": None,
         "http": None,
+        "ui": None,
     }
     actual_config = validate_config(expected_config)
     assert actual_config == expected_config
@@ -467,6 +469,7 @@ def test_config_to_docker_nodejs():
                 "node_version": "20",
                 "graphs": graphs,
                 "dockerfile_lines": ["ARG meow", "ARG foo"],
+                "ui": {"agent": "./graphs/agent.ui.jsx"},
             }
         ),
         "langchain/langgraphjs-api",
@@ -477,6 +480,7 @@ ARG foo
 ADD . /deps/unit_tests
 RUN cd /deps/unit_tests && npm i
 ENV LANGSERVE_GRAPHS='{"agent": "./graphs/agent.js:graph"}'
+ENV LANGGRAPH_UI='{"agent": "./graphs/agent.ui.jsx"}'
 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"""
 

libs/langgraph/Makefile

@@ -58,9 +58,11 @@ WORKERS ?= auto
 XDIST_ARGS := $(if $(WORKERS),-n $(WORKERS) --dist worksteal,)
 MAXFAIL ?=
 MAXFAIL_ARGS := $(if $(MAXFAIL),--maxfail $(MAXFAIL),)
+# Add an '-x' if xdist is enabled
+XDIST_ARGS := $(if $(WORKERS),-x $(XDIST_ARGS),)
 
 test_watch:
-	make start-postgres && poetry run ptw . -- --ff -vv -x $(XDIST_ARGS) $(MAXFAIL_ARGS) --snapshot-update --tb short $(TEST); \
+	make start-postgres && poetry run ptw . -- --ff -vv $(XDIST_ARGS) $(MAXFAIL_ARGS) --snapshot-update --tb short $(TEST); \
 	EXIT_CODE=$$?; \
 	make stop-postgres; \
 	exit $$EXIT_CODE

libs/langgraph/README.md

@@ -1,339 +1,87 @@
-# 🦜🕸️LangGraph
+<picture class="github-only">
+  <source media="(prefers-color-scheme: light)" srcset="docs/docs/static/wordmark_dark.svg">
+  <source media="(prefers-color-scheme: dark)" srcset="docs/docs/static/wordmark_light.svg">
+  <img alt="LangGraph Logo" src="docs/docs/static/wordmark_dark.svg" width="80%">
+</picture>
 
-![Version](https://img.shields.io/pypi/v/langgraph)
+<div>
+<br>
+</div>
+
+[![Version](https://img.shields.io/pypi/v/langgraph.svg)](https://pypi.org/project/langgraph/)
 [![Downloads](https://static.pepy.tech/badge/langgraph/month)](https://pepy.tech/project/langgraph)
 [![Open Issues](https://img.shields.io/github/issues-raw/langchain-ai/langgraph)](https://github.com/langchain-ai/langgraph/issues)
 [![Docs](https://img.shields.io/badge/docs-latest-blue)](https://langchain-ai.github.io/langgraph/)
 
-⚡ Building language agents as graphs ⚡
-
 > [!NOTE]
 > Looking for the JS version? See the [JS repo](https://github.com/langchain-ai/langgraphjs) and the [JS docs](https://langchain-ai.github.io/langgraphjs/).
 
-## Overview
+LangGraph — used by Replit, Uber, LinkedIn, GitLab and more — is a low-level orchestration framework for building controllable agents. While langchain provides integrations and composable components to streamline LLM application development, the LangGraph library enables agent orchestration — offering customizable architectures, long-term memory, and human-in-the-loop to reliably handle complex tasks.
 
-[LangGraph](https://langchain-ai.github.io/langgraph/) is a library for building
-stateful, multi-actor applications with LLMs, used to create agent and multi-agent
-workflows. Check out an introductory tutorial [here](https://langchain-ai.github.io/langgraph/tutorials/introduction/).
-
-
-LangGraph is inspired by [Pregel](https://research.google/pubs/pub37252/) and [Apache Beam](https://beam.apache.org/). The public interface draws inspiration from [NetworkX](https://networkx.org/documentation/latest/). LangGraph is built by LangChain Inc, the creators of LangChain, but can be used without LangChain.
-
-### Why use LangGraph?
-
-LangGraph powers [production-grade agents](https://www.langchain.com/built-with-langgraph), trusted by Linkedin, Uber, Klarna, GitLab, and many more. LangGraph provides fine-grained control over both the flow and state of your agent applications. It implements a central [persistence layer](https://langchain-ai.github.io/langgraph/concepts/persistence/), enabling features that are common to most agent architectures:
-
-- **Memory**: LangGraph persists arbitrary aspects of your application's state,
-supporting memory of conversations and other updates within and across user
-interactions;
-- **Human-in-the-loop**: Because state is checkpointed, execution can be interrupted
-and resumed, allowing for decisions, validation, and corrections at key stages via
-human input.
-
-Standardizing these components allows individuals and teams to focus on the behavior
-of their agent, instead of its supporting infrastructure.
-
-Through [LangGraph Platform](#langgraph-platform), LangGraph also provides tooling for
-the development, deployment, debugging, and monitoring of your applications.
-
-LangGraph integrates seamlessly with
-[LangChain](https://python.langchain.com/docs/introduction/) and
-[LangSmith](https://docs.smith.langchain.com/) (but does not require them).
-
-To learn more about LangGraph, check out our first LangChain Academy
-course, *Introduction to LangGraph*, available for free
-[here](https://academy.langchain.com/courses/intro-to-langgraph).
-
-### LangGraph Platform
-
-[LangGraph Platform](https://langchain-ai.github.io/langgraph/concepts/langgraph_platform) is infrastructure for deploying LangGraph agents. It is a commercial solution for deploying agentic applications to production, built on the open-source LangGraph framework. The LangGraph Platform consists of several components that work together to support the development, deployment, debugging, and monitoring of LangGraph applications: [LangGraph Server](https://langchain-ai.github.io/langgraph/concepts/langgraph_server) (APIs), [LangGraph SDKs](https://langchain-ai.github.io/langgraph/concepts/sdk) (clients for the APIs), [LangGraph CLI](https://langchain-ai.github.io/langgraph/concepts/langgraph_cli) (command line tool for building the server), and [LangGraph Studio](https://langchain-ai.github.io/langgraph/concepts/langgraph_studio) (UI/debugger).
-
-See deployment options [here](https://langchain-ai.github.io/langgraph/concepts/deployment_options/)
-(includes a free tier).
-
-Here are some common issues that arise in complex deployments, which LangGraph Platform addresses:
-
-- **Streaming support**: LangGraph Server provides [multiple streaming modes](https://langchain-ai.github.io/langgraph/concepts/streaming) optimized for various application needs
-- **Background runs**: Runs agents asynchronously in the background
-- **Support for long running agents**: Infrastructure that can handle long running processes
-- **[Double texting](https://langchain-ai.github.io/langgraph/concepts/double_texting)**: Handle the case where you get two messages from the user before the agent can respond
-- **Handle burstiness**: Task queue for ensuring requests are handled consistently without loss, even under heavy loads
-
-## Installation
-
-```shell
+```bash
 pip install -U langgraph
 ```
 
-## Example
-
-Let's build a tool-calling [ReAct-style](https://langchain-ai.github.io/langgraph/concepts/agentic_concepts/#react-implementation) agent that uses a search tool!
-
-```shell
-pip install langchain-anthropic
-```
-
-```shell
-export ANTHROPIC_API_KEY=sk-...
-```
-
-Optionally, we can set up [LangSmith](https://docs.smith.langchain.com/) for best-in-class observability.
-
-```shell
-export LANGSMITH_TRACING=true
-export LANGSMITH_API_KEY=lsv2_sk_...
-```
-
-The simplest way to create a tool-calling agent in LangGraph is to use `create_react_agent`:
-
-<details open>
-  <summary>High-level implementation</summary>
+To learn more about how to use LangGraph, check out [the docs](https://langchain-ai.github.io/langgraph/). We show a simple example below of how to create a ReAct agent.
 
 ```python
+# This code depends on pip install langchain[anthropic]
 from langgraph.prebuilt import create_react_agent
-from langgraph.checkpoint.memory import MemorySaver
-from langchain_anthropic import ChatAnthropic
-from langchain_core.tools import tool
 
-# Define the tools for the agent to use
-@tool
 def search(query: str):
     """Call to surf the web."""
-    # This is a placeholder, but don't tell the LLM that...
     if "sf" in query.lower() or "san francisco" in query.lower():
         return "It's 60 degrees and foggy."
     return "It's 90 degrees and sunny."
 
-
-tools = [search]
-model = ChatAnthropic(model="claude-3-5-sonnet-latest", temperature=0)
-
-# Initialize memory to persist state between graph runs
-checkpointer = MemorySaver()
-
-app = create_react_agent(model, tools, checkpointer=checkpointer)
-
-# Use the agent
-final_state = app.invoke(
-    {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-    config={"configurable": {"thread_id": 42}}
+agent = create_react_agent("anthropic:claude-3-7-sonnet-latest", tools=[search])
+agent.invoke(
+    {"messages": [{"role": "user", "content": "what is the weather in sf"}]}
 )
-final_state["messages"][-1].content
-```
-```
-"Based on the search results, I can tell you that the current weather in San Francisco is:\n\nTemperature: 60 degrees Fahrenheit\nConditions: Foggy\n\nSan Francisco is known for its microclimates and frequent fog, especially during the summer months. The temperature of 60°F (about 15.5°C) is quite typical for the city, which tends to have mild temperatures year-round. The fog, often referred to as "Karl the Fog" by locals, is a characteristic feature of San Francisco\'s weather, particularly in the mornings and evenings.\n\nIs there anything else you\'d like to know about the weather in San Francisco or any other location?"
 ```
 
-Now when we pass the same <code>"thread_id"</code>, the conversation context is retained via the saved state (i.e. stored list of messages)
+## Why use LangGraph?
 
-```python
-final_state = app.invoke(
-    {"messages": [{"role": "user", "content": "what about ny"}]},
-    config={"configurable": {"thread_id": 42}}
-)
-final_state["messages"][-1].content
-```
+LangGraph is built for developers who want to build powerful, adaptable AI agents. Developers choose LangGraph for:
 
-```
-"Based on the search results, I can tell you that the current weather in New York City is:\n\nTemperature: 90 degrees Fahrenheit (approximately 32.2 degrees Celsius)\nConditions: Sunny\n\nThis weather is quite different from what we just saw in San Francisco. New York is experiencing much warmer temperatures right now. Here are a few points to note:\n\n1. The temperature of 90°F is quite hot, typical of summer weather in New York City.\n2. The sunny conditions suggest clear skies, which is great for outdoor activities but also means it might feel even hotter due to direct sunlight.\n3. This kind of weather in New York often comes with high humidity, which can make it feel even warmer than the actual temperature suggests.\n\nIt's interesting to see the stark contrast between San Francisco's mild, foggy weather and New York's hot, sunny conditions. This difference illustrates how varied weather can be across different parts of the United States, even on the same day.\n\nIs there anything else you'd like to know about the weather in New York or any other location?"
-```
-</details>
+- **Reliability and controllability.** Steer agent actions with moderation checks and human-in-the-loop approvals. LangGraph persists context for long-running workflows, keeping your agents on course.
+- **Low-level and extensible.** Build custom agents with fully descriptive, low-level primitives – free from rigid abstractions that limit customization. Design scalable multi-agent systems, with each agent serving a specific role tailored to your use case.
+- **First-class streaming support.** With token-by-token streaming and streaming of intermediate steps, LangGraph gives users clear visibility into agent reasoning and actions as they unfold in real time.
 
-> [!TIP]
-> LangGraph is a **low-level** framework that allows you to implement any custom agent
-architectures. Click on the low-level implementation below to see how to implement a
-tool-calling agent from scratch.
+LangGraph is trusted in production and powering agents for companies like:
 
-<details>
-<summary>Low-level implementation</summary>
+- [Klarna](https://blog.langchain.dev/customers-klarna/): Customer support bot for 85 million active users
+- [Elastic](https://www.elastic.co/blog/elastic-security-generative-ai-features): Security AI assistant for threat detection
+- [Uber](https://dpe.org/sessions/ty-smith-adam-huda/this-year-in-ubers-ai-driven-developer-productivity-revolution/): Automated unit test generation
+- [Replit](https://www.langchain.com/breakoutagents/replit): Code generation
+- And many more ([see list here](https://www.langchain.com/built-with-langgraph))
 
-```python
-from typing import Literal
+## LangGraph’s ecosystem
 
-from langchain_anthropic import ChatAnthropic
-from langchain_core.tools import tool
-from langgraph.checkpoint.memory import MemorySaver
-from langgraph.graph import END, START, StateGraph, MessagesState
-from langgraph.prebuilt import ToolNode
+While LangGraph can be used standalone, it also integrates seamlessly with any LangChain product, giving developers a full suite of tools for building agents. To improve your LLM application development, pair LangGraph with:
 
+- [LangSmith](http://www.langchain.com/langsmith) — Helpful for agent evals and observability. Debug poor-performing LLM app runs, evaluate agent trajectories, gain visibility in production, and improve performance over time.
+- [LangGraph Platform](https://langchain-ai.github.io/langgraph/concepts/#langgraph-platform) — Deploy and scale agents effortlessly with a purpose-built deployment platform for long running, stateful workflows. Discover, reuse, configure, and share agents across teams — and iterate quickly with visual prototyping in [LangGraph Studio](https://langchain-ai.github.io/langgraph/concepts/langgraph_studio/).
 
-# Define the tools for the agent to use
-@tool
-def search(query: str):
-    """Call to surf the web."""
-    # This is a placeholder, but don't tell the LLM that...
-    if "sf" in query.lower() or "san francisco" in query.lower():
-        return "It's 60 degrees and foggy."
-    return "It's 90 degrees and sunny."
+## Pairing with LangGraph Platform
 
+While LangGraph is our open-source agent orchestration framework, enterprises that need scalable agent deployment can benefit from [LangGraph Platform](https://langchain-ai.github.io/langgraph/concepts/langgraph_platform/).
 
-tools = [search]
+LangGraph Platform can help engineering teams:
 
-tool_node = ToolNode(tools)
+- **Accelerate agent development**: Quickly create agent UXs with configurable templates and [LangGraph Studio](https://langchain-ai.github.io/langgraph/concepts/langgraph_studio/) for visualizing and debugging agent interactions.
+- **Deploy seamlessly**: We handle the complexity of deploying your agent. LangGraph Platform includes robust APIs for memory, threads, and cron jobs plus auto-scaling task queues & servers.
+- **Centralize agent management & reusability**: Discover, reuse, and manage agents across the organization. Business users can also modify agents without coding.
 
-model = ChatAnthropic(model="claude-3-5-sonnet-latest", temperature=0).bind_tools(tools)
+## Additional resources
 
-# Define the function that determines whether to continue or not
-def should_continue(state: MessagesState) -> Literal["tools", END]:
-    messages = state['messages']
-    last_message = messages[-1]
-    # If the LLM makes a tool call, then we route to the "tools" node
-    if last_message.tool_calls:
-        return "tools"
-    # Otherwise, we stop (reply to the user)
-    return END
+- [LangChain Academy](https://academy.langchain.com/courses/intro-to-langgraph): Learn the basics of LangGraph in our free, structured course.
+- [Tutorials](https://langchain-ai.github.io/langgraph/tutorials/): Simple walkthroughs with guided examples on getting started with LangGraph.
+- [Templates](https://langchain-ai.github.io/langgraph/concepts/template_applications/): Pre-built reference apps for common agentic workflows (e.g. ReAct agent, memory, retrieval etc.) that can be cloned and adapted.
+- [How-to Guides](https://langchain-ai.github.io/langgraph/how-tos/): Quick, actionable code snippets for topics such as streaming, adding memory & persistence, and design patterns (e.g. branching, subgraphs, etc.).
+- [API Reference](https://langchain-ai.github.io/langgraph/reference/graphs/): Detailed reference on core classes, methods, how to use the graph and checkpointing APIs, and higher-level prebuilt components.
+- [Built with LangGraph](https://www.langchain.com/built-with-langgraph): Hear how industry leaders use LangGraph to ship powerful, production-ready AI applications.
 
+## Acknowledgements
 
-# Define the function that calls the model
-def call_model(state: MessagesState):
-    messages = state['messages']
-    response = model.invoke(messages)
-    # We return a list, because this will get added to the existing list
-    return {"messages": [response]}
-
-
-# Define a new graph
-workflow = StateGraph(MessagesState)
-
-# Define the two nodes we will cycle between
-workflow.add_node("agent", call_model)
-workflow.add_node("tools", tool_node)
-
-# Set the entrypoint as `agent`
-# This means that this node is the first one called
-workflow.add_edge(START, "agent")
-
-# We now add a conditional edge
-workflow.add_conditional_edges(
-    # First, we define the start node. We use `agent`.
-    # This means these are the edges taken after the `agent` node is called.
-    "agent",
-    # Next, we pass in the function that will determine which node is called next.
-    should_continue,
-)
-
-# We now add a normal edge from `tools` to `agent`.
-# This means that after `tools` is called, `agent` node is called next.
-workflow.add_edge("tools", 'agent')
-
-# Initialize memory to persist state between graph runs
-checkpointer = MemorySaver()
-
-# Finally, we compile it!
-# This compiles it into a LangChain Runnable,
-# meaning you can use it as you would any other runnable.
-# Note that we're (optionally) passing the memory when compiling the graph
-app = workflow.compile(checkpointer=checkpointer)
-
-# Use the agent
-final_state = app.invoke(
-    {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
-    config={"configurable": {"thread_id": 42}}
-)
-final_state["messages"][-1].content
-```
-
-<b>Step-by-step Breakdown</b>:
-
-<details>
-<summary>Initialize the model and tools.</summary>
-<ul>
-  <li>
-    We use <code>ChatAnthropic</code> as our LLM. <strong>NOTE:</strong> we need to make sure the model knows that it has these tools available to call. We can do this by converting the LangChain tools into the format for OpenAI tool calling using the <code>.bind_tools()</code> method.
-  </li>
-  <li>
-    We define the tools we want to use - a search tool in our case. It is really easy to create your own tools - see documentation here on how to do that <a href="https://python.langchain.com/docs/how_to/custom_tools/">here</a>.
-  </li>
-</ul>
-</details>
-
-<details>
-<summary>Initialize graph with state.</summary>
-
-<ul>
-    <li>We initialize graph (<code>StateGraph</code>) by passing state schema (in our case <code>MessagesState</code>)</li>
-    <li><code>MessagesState</code> is a prebuilt state schema that has one attribute -- a list of LangChain <code>Message</code> objects, as well as logic for merging the updates from each node into the state.</li>
-</ul>
-</details>
-
-<details>
-<summary>Define graph nodes.</summary>
-
-There are two main nodes we need:
-
-<ul>
-    <li>The <code>agent</code> node: responsible for deciding what (if any) actions to take.</li>
-    <li>The <code>tools</code> node that invokes tools: if the agent decides to take an action, this node will then execute that action.</li>
-</ul>
-</details>
-
-<details>
-<summary>Define entry point and graph edges.</summary>
-
-First, we need to set the entry point for graph execution - <code>agent</code> node.
-
-Then we define one normal and one conditional edge. Conditional edge means that the destination depends on the contents of the graph's state (<code>MessagesState</code>). In our case, the destination is not known until the agent (LLM) decides.
-
-<ul>
-  <li>Conditional edge: after the agent is called, we should either:
-    <ul>
-      <li>a. Run tools if the agent said to take an action, OR</li>
-      <li>b. Finish (respond to the user) if the agent did not ask to run tools</li>
-    </ul>
-  </li>
-  <li>Normal edge: after the tools are invoked, the graph should always return to the agent to decide what to do next</li>
-</ul>
-</details>
-
-<details>
-<summary>Compile the graph.</summary>
-
-<ul>
-  <li>
-    When we compile the graph, we turn it into a LangChain 
-    <a href="https://python.langchain.com/docs/concepts/runnables/">Runnable</a>, 
-    which automatically enables calling <code>.invoke()</code>, <code>.stream()</code> and <code>.batch()</code> 
-    with your inputs
-  </li>
-  <li>
-    We can also optionally pass checkpointer object for persisting state between graph runs, and enabling memory, 
-    human-in-the-loop workflows, time travel and more. In our case we use <code>MemorySaver</code> - 
-    a simple in-memory checkpointer
-  </li>
-</ul>
-</details>
-
-<details>
-<summary>Execute the graph.</summary>
-
-<ol>
-  <li>LangGraph adds the input message to the internal state, then passes the state to the entrypoint node, <code>"agent"</code>.</li>
-  <li>The <code>"agent"</code> node executes, invoking the chat model.</li>
-  <li>The chat model returns an <code>AIMessage</code>. LangGraph adds this to the state.</li>
-  <li>Graph cycles the following steps until there are no more <code>tool_calls</code> on <code>AIMessage</code>:
-    <ul>
-      <li>If <code>AIMessage</code> has <code>tool_calls</code>, <code>"tools"</code> node executes</li>
-      <li>The <code>"agent"</code> node executes again and returns <code>AIMessage</code></li>
-    </ul>
-  </li>
-  <li>Execution progresses to the special <code>END</code> value and outputs the final state. And as a result, we get a list of all our chat messages as output.</li>
-</ol>
-</details>
-
-</details>
-
-## Documentation
-
-* [Tutorials](https://langchain-ai.github.io/langgraph/tutorials/): Learn to build with LangGraph through guided examples.
-* [How-to Guides](https://langchain-ai.github.io/langgraph/how-tos/): Accomplish specific things within LangGraph, from streaming, to adding memory & persistence, to common design patterns (branching, subgraphs, etc.), these are the place to go if you want to copy and run a specific code snippet.
-* [Conceptual Guides](https://langchain-ai.github.io/langgraph/concepts/high_level/): In-depth explanations of the key concepts and principles behind LangGraph, such as nodes, edges, state and more.
-* [API Reference](https://langchain-ai.github.io/langgraph/reference/graphs/): Review important classes and methods, simple examples of how to use the graph and checkpointing APIs, higher-level prebuilt components and more.
-* [LangGraph Platform](https://langchain-ai.github.io/langgraph/concepts/#langgraph-platform): LangGraph Platform is a commercial solution for deploying agentic applications in production, built on the open-source LangGraph framework.
-
-## Resources
-
-* [Built with LangGraph](https://www.langchain.com/built-with-langgraph): Hear how industry leaders use LangGraph to ship powerful, production-ready AI applications.
-
-## Contributing
-
-For more information on how to contribute, see [here](https://github.com/langchain-ai/langgraph/blob/main/CONTRIBUTING.md).
+LangGraph is inspired by [Pregel](https://research.google/pubs/pub37252/) and [Apache Beam](https://beam.apache.org/). The public interface draws inspiration from [NetworkX](https://networkx.org/documentation/latest/). LangGraph is built by LangChain Inc, the creators of LangChain, but can be used without LangChain.

libs/langgraph/bench/__main__.py

@@ -6,7 +6,9 @@ from pyperf._runner import Runner
 from uvloop import new_event_loop
 
 from bench.fanout_to_subgraph import fanout_to_subgraph, fanout_to_subgraph_sync
+from bench.pydantic_state import pydantic_state
 from bench.react_agent import react_agent
+from bench.sequential import create_sequential
 from bench.wide_state import wide_state
 from langgraph.checkpoint.memory import MemorySaver
 from langgraph.pregel import Pregel
@@ -203,6 +205,126 @@ benchmarks = (
             ]
         },
     ),
+    (
+        "sequential_20",
+        create_sequential(20).compile(),
+        create_sequential(20).compile(),
+        {"messages": []},  # Empty list of messages
+    ),
+    (
+        "sequential_50",
+        create_sequential(50).compile(),
+        create_sequential(50).compile(),
+        {"messages": []},  # Empty list of messages
+    ),
+    (
+        "sequential_100",
+        create_sequential(100).compile(),
+        create_sequential(100).compile(),
+        {"messages": []},  # Empty list of messages
+    ),
+    (
+        "sequential_200",
+        create_sequential(200).compile(),
+        create_sequential(200).compile(),
+        {"messages": []},  # Empty list of messages
+    ),
+    (
+        "pydantic_state_25x300",
+        pydantic_state(300).compile(checkpointer=None),
+        pydantic_state(300).compile(checkpointer=None),
+        {
+            "messages": [
+                {
+                    str(i) * 10: {
+                        str(j) * 10: ["hi?" * 10, True, 1, 6327816386138, None] * 5
+                        for j in range(5)
+                    }
+                    for i in range(5)
+                }
+            ]
+        },
+    ),
+    (
+        "pydantic_state_25x300_checkpoint",
+        pydantic_state(300).compile(checkpointer=MemorySaver()),
+        pydantic_state(300).compile(checkpointer=MemorySaver()),
+        {
+            "messages": [
+                {
+                    str(i) * 10: {
+                        str(j) * 10: ["hi?" * 10, True, 1, 6327816386138, None] * 5
+                        for j in range(5)
+                    }
+                    for i in range(5)
+                }
+            ]
+        },
+    ),
+    (
+        "pydantic_state_15x600",
+        pydantic_state(600).compile(checkpointer=None),
+        pydantic_state(600).compile(checkpointer=None),
+        {
+            "messages": [
+                {
+                    str(i) * 10: {
+                        str(j) * 10: ["hi?" * 10, True, 1, 6327816386138, None] * 5
+                        for j in range(5)
+                    }
+                    for i in range(3)
+                }
+            ]
+        },
+    ),
+    (
+        "pydantic_state_15x600_checkpoint",
+        pydantic_state(600).compile(checkpointer=MemorySaver()),
+        pydantic_state(600).compile(checkpointer=MemorySaver()),
+        {
+            "messages": [
+                {
+                    str(i) * 10: {
+                        str(j) * 10: ["hi?" * 10, True, 1, 6327816386138, None] * 5
+                        for j in range(5)
+                    }
+                    for i in range(3)
+                }
+            ]
+        },
+    ),
+    (
+        "pydantic_state_9x1200",
+        pydantic_state(1200).compile(checkpointer=None),
+        pydantic_state(1200).compile(checkpointer=None),
+        {
+            "messages": [
+                {
+                    str(i) * 10: {
+                        str(j) * 10: ["hi?" * 10, True, 1, 6327816386138, None] * 5
+                        for j in range(3)
+                    }
+                    for i in range(3)
+                }
+            ]
+        },
+    ),
+    (
+        "pydantic_state_9x1200_checkpoint",
+        pydantic_state(1200).compile(checkpointer=MemorySaver()),
+        pydantic_state(1200).compile(checkpointer=MemorySaver()),
+        {
+            "messages": [
+                {
+                    str(i) * 10: {
+                        str(j) * 10: ["hi?" * 10, True, 1, 6327816386138, None] * 5
+                        for j in range(3)
+                    }
+                    for i in range(3)
+                }
+            ]
+        },
+    ),
 )
 
 

libs/langgraph/bench/pydantic_state.py

@@ -0,0 +1,327 @@
+import operator
+from functools import partial
+from random import choice
+from typing import Annotated, Optional, Sequence
+
+from pydantic import BaseModel, Field, field_validator
+
+from langgraph.constants import END, START
+from langgraph.graph.state import StateGraph
+
+
+def pydantic_state(n: int) -> StateGraph:
+    class State(BaseModel):
+        messages: Annotated[list, operator.add] = Field(default_factory=list)
+
+        @field_validator("messages", mode="after")
+        @classmethod
+        def validate_messages(cls, v):
+            if not isinstance(v, list):
+                raise TypeError("messages must be a list")
+            for msg in v:
+                if not isinstance(msg, dict):
+                    raise TypeError("messages must be a list of dicts")
+                if not all(isinstance(k, str) for k in msg.keys()):
+                    raise TypeError("messages must be a list of dicts with str keys")
+            return v
+
+        trigger_events: Annotated[list, operator.add] = Field(default_factory=list)
+        """The external events that are converted by the graph."""
+
+        @field_validator("trigger_events", mode="after")
+        @classmethod
+        def validate_trigger_events(cls, v):
+            if not isinstance(v, list):
+                raise TypeError("trigger_events must be a list")
+            for event in v:
+                if not isinstance(event, dict):
+                    raise TypeError("trigger_events must be a list of dicts")
+                if not all(isinstance(k, str) for k in event.keys()):
+                    raise TypeError(
+                        "trigger_events must be a list of dicts with str keys"
+                    )
+            return v
+
+        primary_issue_medium: Annotated[str, lambda x, y: y or x] = Field(
+            default="email"
+        )
+        """The primary issue medium for the current conversation."""
+
+        @field_validator("primary_issue_medium", mode="after")
+        @classmethod
+        def validate_primary_issue_medium(cls, v):
+            if not isinstance(v, str):
+                raise TypeError("primary_issue_medium must be a string")
+            return v
+
+        autoresponse: Annotated[Optional[dict], lambda _, y: y] = Field(
+            default=None
+        )  # Always overwrite
+
+        @field_validator("autoresponse", mode="after")
+        @classmethod
+        def validate_autoresponse(cls, v):
+            if v is not None and not isinstance(v, dict):
+                raise TypeError("autoresponse must be a dict or None")
+            return v
+
+        issue: Annotated[dict | None, lambda x, y: y if y else x] = Field(default=None)
+
+        @field_validator("issue", mode="after")
+        @classmethod
+        def validate_issue(cls, v):
+            if v is not None and not isinstance(v, dict):
+                raise TypeError("issue must be a dict or None")
+            return v
+
+        relevant_rules: Optional[list[dict]] = Field(default=None)
+        """SOPs fetched from the rulebook that are relevant to the current conversation."""
+
+        @field_validator("relevant_rules", mode="after")
+        @classmethod
+        def validate_relevant_rules(cls, v):
+            if v is None:
+                return v
+            if not isinstance(v, list):
+                raise TypeError("relevant_rules must be a list or None")
+            for rule in v:
+                if not isinstance(rule, dict):
+                    raise TypeError("relevant_rules must be a list of dicts")
+                if not all(isinstance(k, str) for k in rule.keys()):
+                    raise TypeError(
+                        "relevant_rules must be a list of dicts with str keys"
+                    )
+            return v
+
+        memory_docs: Optional[list[dict]] = Field(default=None)
+        """Memory docs fetched from the memory service that are relevant to the current conversation."""
+
+        @field_validator("memory_docs", mode="after")
+        @classmethod
+        def validate_memory_docs(cls, v):
+            if v is None:
+                return v
+            if not isinstance(v, list):
+                raise TypeError("memory_docs must be a list or None")
+            for doc in v:
+                if not isinstance(doc, dict):
+                    raise TypeError("memory_docs must be a list of dicts")
+                if not all(isinstance(k, str) for k in doc.keys()):
+                    raise TypeError("memory_docs must be a list of dicts with str keys")
+            return v
+
+        categorizations: Annotated[list[dict], operator.add] = Field(
+            default_factory=list
+        )
+        """The issue categorizations auto-generated by the AI."""
+
+        @field_validator("categorizations", mode="after")
+        @classmethod
+        def validate_categorizations(cls, v):
+            if not isinstance(v, list):
+                raise TypeError("categorizations must be a list")
+            for categorization in v:
+                if not isinstance(categorization, dict):
+                    raise TypeError("categorizations must be a list of dicts")
+                if not all(isinstance(k, str) for k in categorization.keys()):
+                    raise TypeError(
+                        "categorizations must be a list of dicts with str keys"
+                    )
+            return v
+
+        responses: Annotated[list[dict], operator.add] = Field(default_factory=list)
+        """The draft responses recommended by the AI."""
+
+        @field_validator("responses", mode="after")
+        @classmethod
+        def validate_responses(cls, v):
+            if not isinstance(v, list):
+                raise TypeError("responses must be a list")
+            for response in v:
+                if not isinstance(response, dict):
+                    raise TypeError("responses must be a list of dicts")
+                if not all(isinstance(k, str) for k in response.keys()):
+                    raise TypeError("responses must be a list of dicts with str keys")
+            return v
+
+        user_info: Annotated[Optional[dict], lambda x, y: y if y is not None else x] = (
+            Field(default=None)
+        )
+        """The current user state (by email)."""
+
+        @field_validator("user_info", mode="after")
+        @classmethod
+        def validate_user_info(cls, v):
+            if v is not None and not isinstance(v, dict):
+                raise TypeError("user_info must be a dict or None")
+            return v
+
+        crm_info: Annotated[Optional[dict], lambda x, y: y if y is not None else x] = (
+            Field(default=None)
+        )
+        """The CRM information for organization the current user is from."""
+
+        @field_validator("crm_info", mode="after")
+        @classmethod
+        def validate_crm_info(cls, v):
+            if v is not None and not isinstance(v, dict):
+                raise TypeError("crm_info must be a dict or None")
+            return v
+
+        email_thread_id: Annotated[
+            Optional[str], lambda x, y: y if y is not None else x
+        ] = Field(default=None)
+        """The current email thread ID."""
+
+        @field_validator("email_thread_id", mode="after")
+        @classmethod
+        def validate_email_thread_id(cls, v):
+            if v is not None and not isinstance(v, str):
+                raise TypeError("email_thread_id must be a string or None")
+            return v
+
+        slack_participants: Annotated[dict, operator.or_] = Field(default_factory=dict)
+        """The growing list of current slack participants."""
+
+        @field_validator("slack_participants", mode="after")
+        @classmethod
+        def validate_slack_participants(cls, v):
+            if not isinstance(v, dict):
+                raise TypeError("slack_participants must be a dict")
+            for participant in v:
+                if not isinstance(participant, str):
+                    raise TypeError("slack_participants must be a dict with str keys")
+            return v
+
+        bot_id: Optional[str] = Field(default=None)
+        """The ID of the bot user in the slack channel."""
+
+        @field_validator("bot_id", mode="after")
+        @classmethod
+        def validate_bot_id(cls, v):
+            if v is not None and not isinstance(v, str):
+                raise TypeError("bot_id must be a string or None")
+            return v
+
+        notified_assignees: Annotated[dict, operator.or_] = Field(default_factory=dict)
+
+        @field_validator("notified_assignees", mode="after")
+        def validate_notified_assignees(cls, v):
+            if not isinstance(v, dict):
+                raise TypeError("notified_assignees must be a dict")
+            for assignee in v:
+                if not isinstance(assignee, str):
+                    raise TypeError("notified_assignees must be a dict with str keys")
+            return v
+
+    list_fields = {
+        "messages",
+        "trigger_events",
+        "categorizations",
+        "responses",
+        "memory_docs",
+        "relevant_rules",
+    }
+    dict_fields = {
+        "user_info",
+        "crm_info",
+        "slack_participants",
+        "notified_assignees",
+        "autoresponse",
+        "issue",
+    }
+
+    def read_write(read: str, write: Sequence[str], input: State) -> dict:
+        val = getattr(input, read)
+        val = {val: val} if isinstance(val, str) else val
+        val_single = val[-1] if isinstance(val, list) else val
+        val_list = val if isinstance(val, list) else [val]
+        return {
+            k: val_list
+            if k in list_fields
+            else val_single
+            if k in dict_fields
+            else "".join(choice("abcdefghijklmnopqrstuvwxyz") for _ in range(n))
+            for k in write
+        }
+
+    builder = StateGraph(State)
+    builder.add_edge(START, "one")
+    builder.add_node(
+        "one",
+        partial(read_write, "messages", ["trigger_events", "primary_issue_medium"]),
+    )
+    builder.add_edge("one", "two")
+    builder.add_node(
+        "two",
+        partial(read_write, "trigger_events", ["autoresponse", "issue"]),
+    )
+    builder.add_edge("two", "three")
+    builder.add_edge("two", "four")
+    builder.add_node(
+        "three",
+        partial(read_write, "autoresponse", ["relevant_rules"]),
+    )
+    builder.add_node(
+        "four",
+        partial(
+            read_write,
+            "trigger_events",
+            ["categorizations", "responses", "memory_docs"],
+        ),
+    )
+    builder.add_node(
+        "five",
+        partial(
+            read_write,
+            "categorizations",
+            [
+                "user_info",
+                "crm_info",
+                "email_thread_id",
+                "slack_participants",
+                "bot_id",
+                "notified_assignees",
+            ],
+        ),
+    )
+    builder.add_edge(["three", "four"], "five")
+    builder.add_edge("five", "six")
+    builder.add_node(
+        "six",
+        partial(read_write, "responses", ["messages"]),
+    )
+    builder.add_conditional_edges(
+        "six", lambda state: END if len(state.messages) > n else "one"
+    )
+
+    return builder
+
+
+if __name__ == "__main__":
+    import asyncio
+
+    import uvloop
+
+    from langgraph.checkpoint.memory import MemorySaver
+
+    graph = pydantic_state(1000).compile(checkpointer=MemorySaver())
+    input = {
+        "messages": [
+            {
+                str(i) * 10: {
+                    str(j) * 10: ["hi?" * 10, True, 1, 6327816386138, None] * 5
+                    for j in range(5)
+                }
+                for i in range(5)
+            }
+        ]
+    }
+    config = {"configurable": {"thread_id": "1"}, "recursion_limit": 20000000000}
+
+    async def run():
+        async for c in graph.astream(input, config=config):
+            print(c.keys())
+
+    uvloop.install()
+    asyncio.run(run())

libs/langgraph/bench/sequential.py

@@ -0,0 +1,48 @@
+"""Create a sequential no-op graph consisting of a few hundred nodes."""
+
+from langgraph.graph import MessagesState, StateGraph
+from langgraph.utils.runnable import RunnableCallable
+
+
+def create_sequential(number_nodes) -> StateGraph:
+    """Create a sequential no-op graph consisting of a few hundred nodes."""
+    builder = StateGraph(MessagesState)
+
+    def noop(state: MessagesState) -> None:
+        """No-op function."""
+        pass
+
+    async def anoop(state: MessagesState) -> None:
+        """No-op function."""
+        pass
+
+    prev_node = "__start__"
+
+    for i in range(number_nodes):
+        name = f"node_{i}"
+        builder.add_node(name, RunnableCallable(noop, anoop))
+        builder.add_edge(prev_node, name)
+        prev_node = name
+
+    builder.add_edge(prev_node, "__end__")
+    return builder
+
+
+if __name__ == "__main__":
+    import asyncio
+    import time
+
+    import uvloop
+
+    graph = create_sequential(2000).compile()
+    input = {"messages": []}  # Empty list of messages
+    config = {"recursion_limit": 20000000000}
+
+    async def run():
+        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/channels/any_value.py

@@ -1,8 +1,9 @@
-from typing import Generic, Optional, Sequence, Type
+from typing import Any, Generic, Optional, Sequence, Type
 
 from typing_extensions import Self
 
 from langgraph.channels.base import BaseChannel, Value
+from langgraph.constants import MISSING
 from langgraph.errors import EmptyChannelError
 
 
@@ -12,6 +13,10 @@ class AnyValue(Generic[Value], BaseChannel[Value, Value, Value]):
 
     __slots__ = ("typ", "value")
 
+    def __init__(self, typ: Any, key: str = "") -> None:
+        super().__init__(typ, key)
+        self.value = MISSING
+
     def __eq__(self, value: object) -> bool:
         return isinstance(value, AnyValue)
 
@@ -34,17 +39,19 @@ class AnyValue(Generic[Value], BaseChannel[Value, Value, Value]):
 
     def update(self, values: Sequence[Value]) -> bool:
         if len(values) == 0:
-            try:
-                del self.value
-                return True
-            except AttributeError:
+            if self.value is MISSING:
                 return False
+            else:
+                self.value = MISSING
+                return True
 
         self.value = values[-1]
         return True
 
     def get(self) -> Value:
-        try:
-            return self.value
-        except AttributeError:
+        if self.value is MISSING:
             raise EmptyChannelError()
+        return self.value
+
+    def is_available(self) -> bool:
+        return self.value is not MISSING

libs/langgraph/langgraph/channels/base.py

@@ -64,6 +64,17 @@ class BaseChannel(Generic[Value, Update, C], ABC):
         """
         return False
 
+    def is_available(self) -> bool:
+        """Return True if the channel is available (not empty), False otherwise.
+        Subclasses should override this method to provide a more efficient
+        implementation than calling get() and catching EmptyChannelError.
+        """
+        try:
+            self.get()
+            return True
+        except EmptyChannelError:
+            return False
+
 
 __all__ = [
     "BaseChannel",

libs/langgraph/langgraph/channels/binop.py

@@ -10,6 +10,7 @@ from typing import (
 from typing_extensions import NotRequired, Required, Self
 
 from langgraph.channels.base import BaseChannel, Value
+from langgraph.constants import MISSING
 from langgraph.errors import EmptyChannelError
 
 
@@ -51,7 +52,7 @@ class BinaryOperatorAggregate(Generic[Value], BaseChannel[Value, Value, Value]):
         try:
             self.value = typ()
         except Exception:
-            pass
+            self.value = MISSING
 
     def __eq__(self, value: object) -> bool:
         return isinstance(value, BinaryOperatorAggregate) and (
@@ -81,7 +82,7 @@ class BinaryOperatorAggregate(Generic[Value], BaseChannel[Value, Value, Value]):
     def update(self, values: Sequence[Value]) -> bool:
         if not values:
             return False
-        if not hasattr(self, "value"):
+        if self.value is MISSING:
             self.value = values[0]
             values = values[1:]
         for value in values:
@@ -89,7 +90,9 @@ class BinaryOperatorAggregate(Generic[Value], BaseChannel[Value, Value, Value]):
         return True
 
     def get(self) -> Value:
-        try:
-            return self.value
-        except AttributeError:
+        if self.value is MISSING:
             raise EmptyChannelError()
+        return self.value
+
+    def is_available(self) -> bool:
+        return self.value is not MISSING

libs/langgraph/langgraph/channels/dynamic_barrier_value.py

@@ -85,6 +85,9 @@ class DynamicBarrierValue(
             raise EmptyChannelError()
         return None
 
+    def is_available(self) -> bool:
+        return self.seen == self.names
+
     def consume(self) -> bool:
         if self.seen == self.names:
             self.seen = set()

libs/langgraph/langgraph/channels/ephemeral_value.py

@@ -3,6 +3,7 @@ from typing import Any, Generic, Optional, Sequence, Type
 from typing_extensions import Self
 
 from langgraph.channels.base import BaseChannel, Value
+from langgraph.constants import MISSING
 from langgraph.errors import EmptyChannelError, InvalidUpdateError
 
 
@@ -14,6 +15,7 @@ class EphemeralValue(Generic[Value], BaseChannel[Value, Value, Value]):
     def __init__(self, typ: Any, guard: bool = True) -> None:
         super().__init__(typ)
         self.guard = guard
+        self.value = MISSING
 
     def __eq__(self, value: object) -> bool:
         return isinstance(value, EphemeralValue) and value.guard == self.guard
@@ -37,10 +39,10 @@ class EphemeralValue(Generic[Value], BaseChannel[Value, Value, Value]):
 
     def update(self, values: Sequence[Value]) -> bool:
         if len(values) == 0:
-            try:
-                del self.value
+            if self.value is not MISSING:
+                self.value = MISSING
                 return True
-            except AttributeError:
+            else:
                 return False
         if len(values) != 1 and self.guard:
             raise InvalidUpdateError(
@@ -51,7 +53,9 @@ class EphemeralValue(Generic[Value], BaseChannel[Value, Value, Value]):
         return True
 
     def get(self) -> Value:
-        try:
-            return self.value
-        except AttributeError:
+        if self.value is MISSING:
             raise EmptyChannelError()
+        return self.value
+
+    def is_available(self) -> bool:
+        return self.value is not MISSING

libs/langgraph/langgraph/channels/last_value.py

@@ -1,8 +1,9 @@
-from typing import Generic, Optional, Sequence, Type
+from typing import Any, Generic, Optional, Sequence, Type
 
 from typing_extensions import Self
 
 from langgraph.channels.base import BaseChannel, Value
+from langgraph.constants import MISSING
 from langgraph.errors import (
     EmptyChannelError,
     ErrorCode,
@@ -16,6 +17,10 @@ class LastValue(Generic[Value], BaseChannel[Value, Value, Value]):
 
     __slots__ = ("value",)
 
+    def __init__(self, typ: Any, key: str = "") -> None:
+        super().__init__(typ, key)
+        self.value = MISSING
+
     def __eq__(self, value: object) -> bool:
         return isinstance(value, LastValue)
 
@@ -50,7 +55,9 @@ class LastValue(Generic[Value], BaseChannel[Value, Value, Value]):
         return True
 
     def get(self) -> Value:
-        try:
-            return self.value
-        except AttributeError:
+        if self.value is MISSING:
             raise EmptyChannelError()
+        return self.value
+
+    def is_available(self) -> bool:
+        return self.value is not MISSING

libs/langgraph/langgraph/channels/named_barrier_value.py

@@ -60,6 +60,9 @@ class NamedBarrierValue(Generic[Value], BaseChannel[Value, Value, set[Value]]):
             raise EmptyChannelError()
         return None
 
+    def is_available(self) -> bool:
+        return self.seen == self.names
+
     def consume(self) -> bool:
         if self.seen == self.names:
             self.seen = set()

libs/langgraph/langgraph/channels/topic.py

@@ -75,3 +75,6 @@ class Topic(
             return list(self.values)
         else:
             raise EmptyChannelError
+
+    def is_available(self) -> bool:
+        return bool(self.values)

libs/langgraph/langgraph/channels/untracked_value.py

@@ -3,6 +3,7 @@ from typing import Generic, Optional, Sequence, Type
 from typing_extensions import Self
 
 from langgraph.channels.base import BaseChannel, Value
+from langgraph.constants import MISSING
 from langgraph.errors import EmptyChannelError, InvalidUpdateError
 
 
@@ -14,6 +15,7 @@ class UntrackedValue(Generic[Value], BaseChannel[Value, Value, Value]):
     def __init__(self, typ: Type[Value], guard: bool = True) -> None:
         super().__init__(typ)
         self.guard = guard
+        self.value = MISSING
 
     def __eq__(self, value: object) -> bool:
         return isinstance(value, UntrackedValue) and value.guard == self.guard
@@ -48,7 +50,9 @@ class UntrackedValue(Generic[Value], BaseChannel[Value, Value, Value]):
         return True
 
     def get(self) -> Value:
-        try:
-            return self.value
-        except AttributeError:
+        if self.value is MISSING:
             raise EmptyChannelError()
+        return self.value
+
+    def is_available(self) -> bool:
+        return self.value is not MISSING

libs/langgraph/langgraph/func/__init__.py

@@ -36,23 +36,31 @@ from langgraph.types import _DC_KWARGS, RetryPolicy, StreamMode
 
 @overload
 def task(
-    *, name: Optional[str] = None, retry: Optional[RetryPolicy] = None
-) -> Callable[[Callable[P, T]], Callable[P, SyncAsyncFuture[T]]]: ...
+    *,
+    name: Optional[str] = None,
+    retry: Optional[RetryPolicy] = None,
+) -> Callable[
+    [Union[Callable[P, Awaitable[T]], Callable[P, T]]],
+    Callable[P, SyncAsyncFuture[T]],
+]: ...
 
 
 @overload
 def task(
-    __func_or_none__: Callable[P, T],
+    __func_or_none__: Union[Callable[P, Awaitable[T]], Callable[P, T]],
 ) -> Callable[P, SyncAsyncFuture[T]]: ...
 
 
 def task(
-    __func_or_none__: Optional[Union[Callable[P, T], Callable[P, Awaitable[T]]]] = None,
+    __func_or_none__: Optional[Union[Callable[P, Awaitable[T]], Callable[P, T]]] = None,
     *,
     name: Optional[str] = None,
     retry: Optional[RetryPolicy] = None,
 ) -> Union[
-    Callable[[Callable[P, T]], Callable[P, SyncAsyncFuture[T]]],
+    Callable[
+        [Union[Callable[P, Awaitable[T]], Callable[P, T]]],
+        Callable[P, SyncAsyncFuture[T]],
+    ],
     Callable[P, SyncAsyncFuture[T]],
 ]:
     """Define a LangGraph task using the `task` decorator.
@@ -345,7 +353,7 @@ class entrypoint:
         value: R
         """Value to return. A value will always be returned even if it is None."""
         save: S
-        """The value for the state for the next checkpoint. 
+        """The value for the state for the next checkpoint.
 
         A value will always be saved even if it is None.
         """

libs/langgraph/langgraph/graph/branch.py

@@ -0,0 +1,215 @@
+import asyncio
+from inspect import (
+    isfunction,
+    ismethod,
+    signature,
+)
+from types import FunctionType
+from typing import (
+    Any,
+    Awaitable,
+    Callable,
+    Hashable,
+    Literal,
+    NamedTuple,
+    Optional,
+    Sequence,
+    Type,
+    Union,
+    cast,
+    get_args,
+    get_origin,
+    get_type_hints,
+)
+
+from langchain_core.runnables import (
+    Runnable,
+    RunnableConfig,
+    RunnableLambda,
+)
+
+from langgraph.constants import END, START
+from langgraph.errors import InvalidUpdateError
+from langgraph.pregel.write import ChannelWrite
+from langgraph.types import Send
+from langgraph.utils.runnable import (
+    RunnableCallable,
+)
+
+
+def _get_branch_path_input_schema(
+    path: Union[
+        Callable[..., Union[Hashable, list[Hashable]]],
+        Callable[..., Awaitable[Union[Hashable, list[Hashable]]]],
+        Runnable[Any, Union[Hashable, list[Hashable]]],
+    ],
+) -> Optional[Type[Any]]:
+    input = None
+    # detect input schema annotation in the branch callable
+    try:
+        callable_: Optional[
+            Union[
+                Callable[..., Union[Hashable, list[Hashable]]],
+                Callable[..., Awaitable[Union[Hashable, list[Hashable]]]],
+            ]
+        ] = None
+        if isinstance(path, (RunnableCallable, RunnableLambda)):
+            if isfunction(path.func) or ismethod(path.func):
+                callable_ = path.func
+            elif (callable_method := getattr(path.func, "__call__", None)) and ismethod(
+                callable_method
+            ):
+                callable_ = callable_method
+            elif isfunction(path.afunc) or ismethod(path.afunc):
+                callable_ = path.afunc
+            elif (
+                callable_method := getattr(path.afunc, "__call__", None)
+            ) and ismethod(callable_method):
+                callable_ = callable_method
+        elif callable(path):
+            callable_ = path
+
+        if callable_ is not None and (hints := get_type_hints(callable_)):
+            first_parameter_name = next(
+                iter(signature(cast(FunctionType, callable_)).parameters.keys())
+            )
+            if input_hint := hints.get(first_parameter_name):
+                if isinstance(input_hint, type) and get_type_hints(input_hint):
+                    input = input_hint
+    except (TypeError, StopIteration):
+        pass
+
+    return input
+
+
+class Branch(NamedTuple):
+    path: Runnable[Any, Union[Hashable, list[Hashable]]]
+    ends: Optional[dict[Hashable, str]]
+    then: Optional[str] = None
+    input_schema: Optional[Type[Any]] = None
+
+    @classmethod
+    def from_path(
+        cls,
+        path: Runnable[Any, Union[Hashable, list[Hashable]]],
+        path_map: Optional[Union[dict[Hashable, str], list[str]]],
+        then: Optional[str] = None,
+        infer_schema: bool = False,
+    ) -> "Branch":
+        # coerce path_map to a dictionary
+        path_map_: Optional[dict[Hashable, str]] = None
+        try:
+            if isinstance(path_map, dict):
+                path_map_ = path_map.copy()
+            elif isinstance(path_map, list):
+                path_map_ = {name: name for name in path_map}
+            else:
+                # find func
+                func: Optional[Callable] = None
+                if isinstance(path, (RunnableCallable, RunnableLambda)):
+                    func = path.func or path.afunc
+                if func is not None:
+                    # find callable method
+                    if (cal := getattr(path, "__call__", None)) and ismethod(cal):
+                        func = cal
+                    # get the return type
+                    if rtn_type := get_type_hints(func).get("return"):
+                        if get_origin(rtn_type) is Literal:
+                            path_map_ = {name: name for name in get_args(rtn_type)}
+        except Exception:
+            pass
+        # infer input schema
+        input_schema = _get_branch_path_input_schema(path) if infer_schema else None
+        # create branch
+        return cls(path=path, ends=path_map_, then=then, input_schema=input_schema)
+
+    def run(
+        self,
+        writer: Callable[
+            [Sequence[Union[str, Send]], RunnableConfig], Optional[ChannelWrite]
+        ],
+        reader: Optional[Callable[[RunnableConfig], Any]] = None,
+    ) -> RunnableCallable:
+        return ChannelWrite.register_writer(
+            RunnableCallable(
+                func=self._route,
+                afunc=self._aroute,
+                writer=writer,
+                reader=reader,
+                name=None,
+                trace=False,
+            )
+        )
+
+    def _route(
+        self,
+        input: Any,
+        config: RunnableConfig,
+        *,
+        reader: Optional[Callable[[RunnableConfig], Any]],
+        writer: Callable[
+            [Sequence[Union[str, Send]], RunnableConfig], Optional[ChannelWrite]
+        ],
+    ) -> Runnable:
+        if reader:
+            value = reader(config)
+            # passthrough additional keys from node to branch
+            # only doable when using dict states
+            if (
+                isinstance(value, dict)
+                and isinstance(input, dict)
+                and self.input_schema is None
+            ):
+                value = {**input, **value}
+        else:
+            value = input
+        result = self.path.invoke(value, config)
+        return self._finish(writer, input, result, config)
+
+    async def _aroute(
+        self,
+        input: Any,
+        config: RunnableConfig,
+        *,
+        reader: Optional[Callable[[RunnableConfig], Any]],
+        writer: Callable[
+            [Sequence[Union[str, Send]], RunnableConfig], Optional[ChannelWrite]
+        ],
+    ) -> Runnable:
+        if reader:
+            value = await asyncio.to_thread(reader, config)
+            # passthrough additional keys from node to branch
+            # only doable when using dict states
+            if (
+                isinstance(value, dict)
+                and isinstance(input, dict)
+                and self.input_schema is None
+            ):
+                value = {**input, **value}
+        else:
+            value = input
+        result = await self.path.ainvoke(value, config)
+        return self._finish(writer, input, result, config)
+
+    def _finish(
+        self,
+        writer: Callable[
+            [Sequence[Union[str, Send]], RunnableConfig], Optional[ChannelWrite]
+        ],
+        input: Any,
+        result: Any,
+        config: RunnableConfig,
+    ) -> Union[Runnable, Any]:
+        if not isinstance(result, (list, tuple)):
+            result = [result]
+        if self.ends:
+            destinations: Sequence[Union[Send, str]] = [
+                r if isinstance(r, Send) else self.ends[r] for r in result
+            ]
+        else:
+            destinations = cast(Sequence[Union[Send, str]], result)
+        if any(dest is None or dest == START for dest in destinations):
+            raise ValueError("Branch did not return a valid destination")
+        if any(p.node == END for p in destinations if isinstance(p, Send)):
+            raise InvalidUpdateError("Cannot send a packet to the END node")
+        return writer(destinations, config) or input

libs/langgraph/langgraph/graph/graph.py

@@ -6,15 +6,11 @@ from typing import (
     Awaitable,
     Callable,
     Hashable,
-    Literal,
     NamedTuple,
     Optional,
     Sequence,
     Union,
     cast,
-    get_args,
-    get_origin,
-    get_type_hints,
     overload,
 )
 
@@ -34,12 +30,13 @@ from langgraph.constants import (
     TAG_HIDDEN,
     Send,
 )
-from langgraph.errors import InvalidUpdateError
+from langgraph.graph.branch import Branch
 from langgraph.pregel import Channel, Pregel
+from langgraph.pregel.protocol import PregelProtocol
 from langgraph.pregel.read import PregelNode
 from langgraph.pregel.write import ChannelWrite, ChannelWriteEntry
 from langgraph.types import All, Checkpointer
-from langgraph.utils.runnable import RunnableCallable, RunnableLike, coerce_to_runnable
+from langgraph.utils.runnable import RunnableLike, coerce_to_runnable
 
 logger = logging.getLogger(__name__)
 
@@ -50,95 +47,6 @@ class NodeSpec(NamedTuple):
     ends: Optional[Union[tuple[str, ...], dict[str, str]]] = EMPTY_SEQ
 
 
-class Branch(NamedTuple):
-    path: Runnable[Any, Union[Hashable, list[Hashable]]]
-    ends: Optional[dict[Hashable, str]]
-    then: Optional[str] = None
-
-    def run(
-        self,
-        writer: Callable[
-            [Sequence[Union[str, Send]], RunnableConfig], Optional[ChannelWrite]
-        ],
-        reader: Optional[Callable[[RunnableConfig], Any]] = None,
-    ) -> RunnableCallable:
-        return ChannelWrite.register_writer(
-            RunnableCallable(
-                func=self._route,
-                afunc=self._aroute,
-                writer=writer,
-                reader=reader,
-                name=None,
-                trace=False,
-            )
-        )
-
-    def _route(
-        self,
-        input: Any,
-        config: RunnableConfig,
-        *,
-        reader: Optional[Callable[[RunnableConfig], Any]],
-        writer: Callable[
-            [Sequence[Union[str, Send]], RunnableConfig], Optional[ChannelWrite]
-        ],
-    ) -> Runnable:
-        if reader:
-            value = reader(config)
-            # passthrough additional keys from node to branch
-            # only doable when using dict states
-            if isinstance(value, dict) and isinstance(input, dict):
-                value = {**input, **value}
-        else:
-            value = input
-        result = self.path.invoke(value, config)
-        return self._finish(writer, input, result, config)
-
-    async def _aroute(
-        self,
-        input: Any,
-        config: RunnableConfig,
-        *,
-        reader: Optional[Callable[[RunnableConfig], Any]],
-        writer: Callable[
-            [Sequence[Union[str, Send]], RunnableConfig], Optional[ChannelWrite]
-        ],
-    ) -> Runnable:
-        if reader:
-            value = await asyncio.to_thread(reader, config)
-            # passthrough additional keys from node to branch
-            # only doable when using dict states
-            if isinstance(value, dict) and isinstance(input, dict):
-                value = {**input, **value}
-        else:
-            value = input
-        result = await self.path.ainvoke(value, config)
-        return self._finish(writer, input, result, config)
-
-    def _finish(
-        self,
-        writer: Callable[
-            [Sequence[Union[str, Send]], RunnableConfig], Optional[ChannelWrite]
-        ],
-        input: Any,
-        result: Any,
-        config: RunnableConfig,
-    ) -> Union[Runnable, Any]:
-        if not isinstance(result, (list, tuple)):
-            result = [result]
-        if self.ends:
-            destinations: Sequence[Union[Send, str]] = [
-                r if isinstance(r, Send) else self.ends[r] for r in result
-            ]
-        else:
-            destinations = cast(Sequence[Union[Send, str]], result)
-        if any(dest is None or dest == START for dest in destinations):
-            raise ValueError("Branch did not return a valid destination")
-        if any(p.node == END for p in destinations if isinstance(p, Send)):
-            raise InvalidUpdateError("Cannot send a packet to the END node")
-        return writer(destinations, config) or input
-
-
 class Graph:
     def __init__(self) -> None:
         self.nodes: dict[str, NodeSpec] = {}
@@ -267,25 +175,7 @@ class Graph:
                 "Adding an edge to a graph that has already been compiled. This will "
                 "not be reflected in the compiled graph."
             )
-        # coerce path_map to a dictionary
-        try:
-            if isinstance(path_map, dict):
-                path_map_ = path_map.copy()
-            elif isinstance(path_map, list):
-                path_map_ = {name: name for name in path_map}
-            elif isinstance(path, Runnable):
-                path_map_ = None
-            elif rtn_type := get_type_hints(path.__call__).get(  # type: ignore[operator]
-                "return"
-            ) or get_type_hints(path).get("return"):
-                if get_origin(rtn_type) is Literal:
-                    path_map_ = {name: name for name in get_args(rtn_type)}
-                else:
-                    path_map_ = None
-            else:
-                path_map_ = None
-        except Exception:
-            path_map_ = None
+
         # find a name for the condition
         path = coerce_to_runnable(path, name=None, trace=True)
         name = path.name or "condition"
@@ -295,7 +185,7 @@ class Graph:
                 f"Branch with name `{path.name}` already exists for node " f"`{source}`"
             )
         # save it
-        self.branches[source][name] = Branch(path, path_map_, then)
+        self.branches[source][name] = Branch.from_path(path, path_map, then, False)
         return self
 
     def set_entry_point(self, key: str) -> Self:
@@ -530,7 +420,38 @@ class CompiledGraph(Pregel):
         *,
         xray: Union[int, bool] = False,
     ) -> DrawableGraph:
-        return self.get_graph(config, xray=xray)
+        """Returns a drawable representation of the computation graph."""
+        from langgraph.pregel.remote import RemoteGraph
+
+        # gather subgraphs
+        if xray:
+            subpregels: dict[str, PregelProtocol] = {
+                k: v
+                async for k, v in self.aget_subgraphs()
+                if isinstance(v, (CompiledGraph, RemoteGraph))
+            }
+            subgraphs = {
+                k: v
+                for k, v in zip(
+                    subpregels,
+                    await asyncio.gather(
+                        *(
+                            p.aget_graph(
+                                config,
+                                xray=xray
+                                if isinstance(xray, bool) or xray <= 0
+                                else xray - 1,
+                            )
+                            for p in subpregels.values()
+                        )
+                    ),
+                )
+            }
+        else:
+            subgraphs = {}
+
+        # draw the graph
+        return self._draw_graph(config, subgraphs=subgraphs)
 
     def get_graph(
         self,
@@ -539,17 +460,36 @@ class CompiledGraph(Pregel):
         xray: Union[int, bool] = False,
     ) -> DrawableGraph:
         """Returns a drawable representation of the computation graph."""
+        from langgraph.pregel.remote import RemoteGraph
+
+        # gather subgraphs
+        if xray:
+            subgraphs = {
+                k: v.get_graph(
+                    config,
+                    xray=xray if isinstance(xray, bool) or xray <= 0 else xray - 1,
+                )
+                for k, v in self.get_subgraphs()
+                if isinstance(v, (CompiledGraph, RemoteGraph))
+            }
+        else:
+            subgraphs = {}
+
+        # draw the graph
+        return self._draw_graph(config, subgraphs=subgraphs)
+
+    def _draw_graph(
+        self,
+        config: Optional[RunnableConfig] = None,
+        *,
+        subgraphs: dict[str, DrawableGraph] = {},
+    ) -> DrawableGraph:
+        # create the graph
         graph = DrawableGraph()
         start_nodes: dict[str, DrawableNode] = {
             START: graph.add_node(self.get_input_schema(config), START)
         }
         end_nodes: dict[str, DrawableNode] = {}
-        if xray:
-            subgraphs = {
-                k: v for k, v in self.get_subgraphs() if isinstance(v, CompiledGraph)
-            }
-        else:
-            subgraphs = {}
 
         def add_edge(
             start: str,
@@ -575,16 +515,11 @@ class CompiledGraph(Pregel):
                 metadata["__interrupt"] = "before"
             elif key in self.interrupt_after_nodes:
                 metadata["__interrupt"] = "after"
-            if xray and key in subgraphs:
-                subgraph = subgraphs[key].get_graph(
-                    config=config,
-                    xray=xray - 1
-                    if isinstance(xray, int) and not isinstance(xray, bool) and xray > 0
-                    else xray,
-                )
+            if key in subgraphs:
+                subgraph = subgraphs[key]
                 subgraph.trim_first_node()
                 subgraph.trim_last_node()
-                if len(subgraph.nodes) > 1:
+                if len(subgraph.nodes) >= 1:
                     e, s = graph.extend(subgraph, prefix=key)
                     if e is None:
                         raise ValueError(

libs/langgraph/langgraph/graph/schema_utils.py

@@ -0,0 +1,210 @@
+import logging
+import weakref
+from inspect import isclass
+from typing import (
+    Any,
+    Callable,
+    Optional,
+    Type,
+    Union,
+    get_args,
+    get_origin,
+    get_type_hints,
+)
+
+from pydantic import BaseModel
+from pydantic.v1 import BaseModel as BaseModelV1
+from typing_extensions import Annotated
+
+logger = logging.getLogger(__name__)
+
+
+_cache: weakref.WeakKeyDictionary[Type[Any], dict[int, "SchemaCoercionMapper"]] = (
+    weakref.WeakKeyDictionary()
+)
+
+
+class SchemaCoercionMapper:
+    def __new__(
+        cls,
+        schema: Type[Any],
+        type_hints: Optional[dict[str, Any]] = None,
+        max_depth: int = 12,
+    ) -> "SchemaCoercionMapper":
+        if schema not in _cache:
+            _cache[schema] = {}
+        if max_depth in _cache[schema]:
+            return _cache[schema][max_depth]
+
+        inst = super().__new__(cls)
+        _cache[schema][max_depth] = inst
+        return inst
+
+    def __init__(
+        self,
+        schema: Type[Any],
+        type_hints: Optional[dict[str, Any]] = None,
+        max_depth: int = 12,
+    ):
+        if hasattr(self, "_inited"):
+            return
+        self._inited = True
+        self.schema = schema
+        self.type_hints = (
+            type_hints
+            if type_hints is not None
+            else get_type_hints(schema, localns={schema.__name__: schema})
+        )
+        self.max_depth = max_depth
+
+        if issubclass(schema, BaseModel):
+            self._fields = {
+                n: self.type_hints.get(n, f.annotation)
+                for n, f in schema.model_fields.items()
+            }
+            self._construct: Callable[..., Any] = schema.model_construct
+
+        elif issubclass(schema, BaseModelV1):
+            self._fields = {
+                n: self.type_hints.get(n, f.annotation)
+                for n, f in schema.__fields__.items()
+            }
+            self._construct = schema.construct
+        else:
+            raise TypeError("Schema is neither valid Pydantic v1 nor v2 model.")
+        self._field_coercers: Optional[dict[str, Callable[[Any, Any], Any]]] = None
+
+    def __call__(self, input_data: Any, depth: Optional[int] = None) -> Any:
+        return self.coerce(input_data, depth)
+
+    def coerce(self, input_data: Any, depth: Optional[int] = None) -> Any:
+        if depth is None:
+            depth = self.max_depth
+        if not isinstance(input_data, dict) or depth <= 0:
+            return input_data
+        processed = {}
+        if self._field_coercers is None:
+            self._field_coercers = {
+                n: self._build_coercer(t, depth - 1) for n, t in self._fields.items()
+            }
+        for k, v in input_data.items():
+            fn = self._field_coercers.get(k)
+            processed[k] = fn(v, depth - 1) if fn else v
+        return self._construct(**processed)
+
+    def _build_coercer(
+        self, field_type: Any, depth: int, throw: bool = False
+    ) -> Callable[[Any, Any], Any]:
+        if depth == 0:
+            return self._passthrough
+        origin = get_origin(field_type)
+
+        if origin is Annotated:
+            real_type, *_ = get_args(field_type)
+            sub = self._build_coercer(real_type, depth - 1)
+            return lambda v, d: sub(v, d)
+        if isclass(field_type):
+            is_class_ = True
+            try:
+                is_base_model = issubclass(field_type, BaseModel)
+            except TypeError:
+                is_class_ = False
+                is_base_model = False
+
+            if is_base_model:
+                mapper = SchemaCoercionMapper(field_type, max_depth=depth - 1)
+                return lambda v, d: mapper.coerce(v, d) if isinstance(v, dict) else v
+            if is_class_ and issubclass(field_type, BaseModelV1):
+                mapper = SchemaCoercionMapper(field_type, max_depth=depth - 1)
+                return lambda v, d: mapper.coerce(v, d) if isinstance(v, dict) else v
+        if origin is list or field_type is list:
+            args = get_args(field_type)
+            if len(args) != 1:
+                return lambda v, d: v
+            sub = self._build_coercer(args[0], depth - 1)
+
+            def list_coercer(v: Any, d: Any) -> Any:
+                if not isinstance(v, (list, tuple)):
+                    return v
+                return [sub(x, d - 1) for x in v]
+
+            return list_coercer
+        if origin is set or field_type is set:
+            args = get_args(field_type)
+            if len(args) != 1:
+                return lambda v, d: v
+            sub = self._build_coercer(args[0], depth - 1)
+
+            def set_coercer(v: Any, d: Any) -> Any:
+                if not isinstance(v, (list, tuple, set)):
+                    return v
+                return {sub(x, d - 1) for x in v}
+
+            return set_coercer
+        if origin is dict or field_type is dict:
+            args = get_args(field_type)
+            if len(args) != 2:
+
+                def dict_coercer(v: Any, d: Any) -> Any:
+                    if not isinstance(v, dict):
+                        if throw:
+                            raise TypeError("Expected dict, got %s" % type(v))
+                    return v
+
+                return dict_coercer
+            k_sub = self._build_coercer(args[0], depth - 1)
+            v_sub = self._build_coercer(args[1], depth - 1)
+
+            def dict_coercer(v: Any, d: Any) -> Any:
+                if not isinstance(v, dict):
+                    if throw:
+                        raise TypeError("Expected dict, got %s" % type(v))
+                    return v
+                return {k_sub(k, d - 1): v_sub(val, d - 1) for k, val in v.items()}
+
+            return dict_coercer
+
+        if origin is tuple:
+            targs = get_args(field_type)
+            if not targs:
+                return lambda v, d: v
+            subs = [self._build_coercer(a, depth - 1) for a in targs]
+
+            def tuple_coercer(v: Any, d: Any) -> Any:
+                if not isinstance(v, (list, tuple)):
+                    return v
+                out = []
+                for i, sp in enumerate(subs):
+                    out.append(sp(v[i] if i < len(v) else None, d - 1))
+                return tuple(out)
+
+            return tuple_coercer
+        if origin is Union:
+            uargs = get_args(field_type)
+            subs, none_in_union = [], False
+            for ix, arg in enumerate(uargs):
+                if arg is type(None):
+                    none_in_union = True
+                else:
+                    subs.append(
+                        self._build_coercer(arg, depth - 1, throw=ix < len(uargs) - 1)
+                    )
+
+            def union_coercer(v: Any, d: Any) -> Any:
+                if v is None and none_in_union:
+                    return None
+                err = None
+                for sp in subs:
+                    try:
+                        return sp(v, d - 1)
+                    except TypeError as e:
+                        err = e
+                if err:
+                    raise err
+                return v
+
+            return union_coercer
+        return self._passthrough
+
+    def _passthrough(self, v: Any, d: Any) -> Any:
+        return v

libs/langgraph/langgraph/graph/state.py

@@ -7,7 +7,9 @@ from inspect import isclass, isfunction, ismethod, signature
 from types import FunctionType
 from typing import (
     Any,
+    Awaitable,
     Callable,
+    Hashable,
     Literal,
     NamedTuple,
     Optional,
@@ -40,7 +42,15 @@ from langgraph.errors import (
     ParentCommand,
     create_error_message,
 )
-from langgraph.graph.graph import END, START, Branch, CompiledGraph, Graph, Send
+from langgraph.graph.branch import Branch
+from langgraph.graph.graph import (
+    END,
+    START,
+    CompiledGraph,
+    Graph,
+    Send,
+)
+from langgraph.graph.schema_utils import SchemaCoercionMapper
 from langgraph.managed.base import (
     ChannelKeyPlaceholder,
     ChannelTypePlaceholder,
@@ -175,6 +185,7 @@ class StateGraph(Graph):
         self.schemas = {}
         self.channels = {}
         self.managed = {}
+        self.type_hints: dict[Type[Any], dict[str, Any]] = {}
         self.schema = state_schema
         self.input = input
         self.output = output
@@ -193,7 +204,7 @@ class StateGraph(Graph):
     def _add_schema(self, schema: Type[Any], /, allow_managed: bool = True) -> None:
         if schema not in self.schemas:
             _warn_invalid_state_schema(schema)
-            channels, managed = _get_channels(schema)
+            channels, managed, type_hints = _get_channels(schema)
             if managed and not allow_managed:
                 names = ", ".join(managed)
                 schema_name = getattr(schema, "__name__", "")
@@ -202,6 +213,7 @@ class StateGraph(Graph):
                     " Managed channels are not permitted in Input/Output schema."
                 )
             self.schemas[schema] = {**channels, **managed}
+            self.type_hints[schema] = type_hints
             for key, channel in channels.items():
                 if key in self.channels:
                     if self.channels[key] != channel:
@@ -406,7 +418,7 @@ class StateGraph(Graph):
                         and (vals := get_args(rargs[0]))
                     ):
                         ends = vals
-        except (TypeError, StopIteration):
+        except (NameError, TypeError, StopIteration):
             pass
 
         if destinations is not None:
@@ -461,6 +473,57 @@ class StateGraph(Graph):
         self.waiting_edges.add((tuple(start_key), end_key))
         return self
 
+    def add_conditional_edges(
+        self,
+        source: str,
+        path: Union[
+            Callable[..., Union[Hashable, list[Hashable]]],
+            Callable[..., Awaitable[Union[Hashable, list[Hashable]]]],
+            Runnable[Any, Union[Hashable, list[Hashable]]],
+        ],
+        path_map: Optional[Union[dict[Hashable, str], list[str]]] = None,
+        then: Optional[str] = None,
+    ) -> Self:
+        """Add a conditional edge from the starting node to any number of destination nodes.
+
+        Args:
+            source (str): The starting node. This conditional edge will run when
+                exiting this node.
+            path (Union[Callable, Runnable]): The callable that determines the next
+                node or nodes. If not specifying `path_map` it should return one or
+                more nodes. If it returns END, the graph will stop execution.
+            path_map (Optional[dict[Hashable, str]]): Optional mapping of paths to node
+                names. If omitted the paths returned by `path` should be node names.
+            then (Optional[str]): The name of a node to execute after the nodes
+                selected by `path`.
+
+        Returns:
+            Self: The instance of the graph, allowing for method chaining.
+
+        Note: Without typehints on the `path` function's return value (e.g., `-> Literal["foo", "__end__"]:`)
+            or a path_map, the graph visualization assumes the edge could transition to any node in the graph.
+
+        """  # noqa: E501
+        if self.compiled:
+            logger.warning(
+                "Adding an edge to a graph that has already been compiled. This will "
+                "not be reflected in the compiled graph."
+            )
+
+        # find a name for the condition
+        path = coerce_to_runnable(path, name=None, trace=True)
+        name = path.name or "condition"
+        # validate the condition
+        if name in self.branches[source]:
+            raise ValueError(
+                f"Branch with name `{path.name}` already exists for node " f"`{source}`"
+            )
+        # save it
+        self.branches[source][name] = Branch.from_path(path, path_map, then, True)
+        if schema := self.branches[source][name].input_schema:
+            self._add_schema(schema)
+        return self
+
     def add_sequence(
         self,
         nodes: Sequence[Union[RunnableLike, tuple[str, RunnableLike]]],
@@ -566,6 +629,13 @@ class StateGraph(Graph):
         compiled = CompiledStateGraph(
             builder=self,
             config_type=self.config_schema,
+            input_model=(
+                self.input
+                if len(self.channels) > 1
+                and isclass(self.input)
+                and issubclass(self.input, (BaseModel, BaseModelV1))
+                else None
+            ),
             nodes={},
             channels={
                 **self.channels,
@@ -694,24 +764,32 @@ class CompiledStateGraph(CompiledGraph):
                     else:
                         updates.extend(_get_updates(i) or ())
                 return updates
-            elif get_type_hints(type(input)):
-                # if input is a Pydantic model, only update values
-                # for the keys that have been explicitly set by the users
-                # (this is needed to avoid sending updates for fields with None defaults)
-                output_keys_ = output_keys
+            elif (t := type(input)) and get_type_hints(t):
                 # Pydantic v2
-                if hasattr(input, "model_fields_set"):
-                    output_keys_ = [
-                        k for k in output_keys if k in input.model_fields_set
-                    ]
+                if isinstance(input, BaseModel):
+                    keep: Optional[set[str]] = input.model_fields_set
+                    defaults = {k: v.default for k, v in input.model_fields.items()}
                 # Pydantic v1
-                elif hasattr(input, "__fields_set__"):
-                    output_keys_ = [k for k in output_keys if k in input.__fields_set__]
+                elif isinstance(input, BaseModelV1):
+                    keep = input.__fields_set__
+                    defaults = {k: v.default for k, v in t.__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, getattr(input, k))
-                    for k in output_keys_
-                    if getattr(input, k, MISSING) is not MISSING
+                    (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)
+                    )
                 ]
             else:
                 msg = create_error_message(
@@ -737,7 +815,6 @@ class CompiledStateGraph(CompiledGraph):
                     ChannelWrite(
                         write_entries,
                         tags=[TAG_HIDDEN],
-                        require_at_least_one_of=output_keys,
                     ),
                 ],
             )
@@ -746,16 +823,18 @@ class CompiledStateGraph(CompiledGraph):
             input_values = {k: k for k in self.builder.schemas[input_schema]}
             is_single_input = len(input_values) == 1 and "__root__" in input_values
 
+            branch_channel = f"branch:to:{key}"
             self.channels[key] = EphemeralValue(Any, guard=False)
+            self.channels[branch_channel] = EphemeralValue(Any, guard=False)
             self.nodes[key] = PregelNode(
-                triggers=[],
+                triggers=[branch_channel],
                 # 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=(
-                    None
-                    if is_single_input or issubclass(input_schema, dict)
-                    else partial(_coerce_state, input_schema)
+                mapper=_pick_mapper(
+                    list(input_values),
+                    input_schema,
+                    self.builder.type_hints[input_schema],
                 ),
                 writers=[
                     # publish to this channel and state keys
@@ -807,7 +886,7 @@ class CompiledStateGraph(CompiledGraph):
             if filtered := [p for p in packets if p != END]:
                 writes = [
                     (
-                        ChannelWriteEntry(f"branch:{start}:{name}:{p}", start)
+                        ChannelWriteEntry(f"branch:to:{p}", start)
                         if not isinstance(p, Send)
                         else p
                     )
@@ -826,12 +905,12 @@ class CompiledStateGraph(CompiledGraph):
                     config, cast(Sequence[Union[Send, ChannelWriteEntry]], writes)
                 )
 
-        # attach branch publisher
-        schema = (
+        schema = branch.input_schema or (
             self.builder.nodes[start].input
             if start in self.builder.nodes
             else self.builder.schema
         )
+        # attach branch publisher
         self.nodes[start] |= branch.run(
             branch_writer,
             _get_state_reader(self.builder, schema) if with_reader else None,
@@ -843,11 +922,6 @@ class CompiledStateGraph(CompiledGraph):
             if branch.ends
             else [node for node in self.builder.nodes if node != branch.then]
         )
-        for end in ends:
-            if end != END:
-                channel_name = f"branch:{start}:{name}:{end}"
-                self.channels[channel_name] = EphemeralValue(Any, guard=False)
-                self.nodes[end].triggers.append(channel_name)
 
         # attach then subscriber
         if branch.then and branch.then != END:
@@ -871,14 +945,23 @@ def _get_state_reader(
         select=select[0] if select == ["__root__"] else select,
         fresh=True,
         # coerce state dict to schema class (eg. pydantic model)
-        mapper=(
-            None
-            if state_keys == ["__root__"] or issubclass(schema, dict)
-            else partial(_coerce_state, schema)
-        ),
+        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]]:
+    if state_keys == ["__root__"]:
+        return None
+    if isclass(schema):
+        if issubclass(schema, dict):
+            return None
+        if issubclass(schema, (BaseModel, BaseModelV1)):
+            return SchemaCoercionMapper(schema, type_hints)
+    return partial(_coerce_state, schema)
+
+
 def _coerce_state(schema: Type[Any], input: dict[str, Any]) -> dict[str, Any]:
     return schema(**input)
 
@@ -937,18 +1020,24 @@ CONTROL_BRANCH = Branch(CONTROL_BRANCH_PATH, None)
 
 def _get_channels(
     schema: Type[dict],
-) -> tuple[dict[str, BaseChannel], dict[str, ManagedValueSpec]]:
+) -> tuple[dict[str, BaseChannel], dict[str, ManagedValueSpec], dict[str, Any]]:
     if not hasattr(schema, "__annotations__"):
-        return {"__root__": _get_channel("__root__", schema, allow_managed=False)}, {}
+        return (
+            {"__root__": _get_channel("__root__", schema, allow_managed=False)},
+            {},
+            {},
+        )
 
+    type_hints = get_type_hints(schema, include_extras=True)
     all_keys = {
         name: _get_channel(name, typ)
-        for name, typ in get_type_hints(schema, include_extras=True).items()
+        for name, typ in type_hints.items()
         if name != "__slots__"
     }
     return (
         {k: v for k, v in all_keys.items() if isinstance(v, BaseChannel)},
         {k: v for k, v in all_keys.items() if is_managed_value(v)},
+        type_hints,
     )
 
 

libs/langgraph/langgraph/pregel/algo.py

@@ -1,4 +1,4 @@
-import functools
+import binascii
 import itertools
 import sys
 from collections import defaultdict, deque
@@ -19,7 +19,6 @@ from typing import (
     cast,
     overload,
 )
-from uuid import UUID
 
 from langchain_core.callbacks import Callbacks
 from langchain_core.callbacks.manager import AsyncParentRunManager, ParentRunManager
@@ -28,6 +27,7 @@ from langchain_core.runnables.config import RunnableConfig
 from langgraph.channels.base import BaseChannel
 from langgraph.checkpoint.base import (
     BaseCheckpointSaver,
+    ChannelVersions,
     Checkpoint,
     PendingWrite,
     V,
@@ -233,10 +233,21 @@ def apply_writes(
     channels: Mapping[str, BaseChannel],
     tasks: Iterable[WritesProtocol],
     get_next_version: Optional[GetNextVersion],
-) -> dict[str, list[Any]]:
+) -> tuple[dict[str, list[Any]], set[str]]:
     """Apply writes from a set of tasks (usually the tasks from a Pregel step)
     to the checkpoint and channels, and return managed values writes to be applied
-    externally."""
+    externally.
+
+    Args:
+        checkpoint: The checkpoint to update.
+        channels: The channels to update.
+        tasks: The tasks to apply writes from.
+        get_next_version: Optional function to determine the next version of a channel.
+
+    Returns:
+        A tuple containing the managed values writes to be applied externally, and
+        the set of channels that were updated in this step.
+    """
     # sort tasks on path, to ensure deterministic order for update application
     # any path parts after the 3rd are ignored for sorting
     # (we use them for eg. task ids which aren't good for sorting)
@@ -318,9 +329,8 @@ def apply_writes(
                         max_version,
                         channels[chan],
                     )
-
     # Return managed values writes to be applied externally
-    return pending_writes_by_managed
+    return pending_writes_by_managed, updated_channels
 
 
 @overload
@@ -337,6 +347,8 @@ def prepare_next_tasks(
     store: Literal[None] = None,
     checkpointer: Literal[None] = None,
     manager: Literal[None] = None,
+    trigger_to_nodes: Optional[Mapping[str, list[str]]] = None,
+    updated_channels: Optional[set[str]] = None,
 ) -> dict[str, PregelTask]: ...
 
 
@@ -354,6 +366,8 @@ def prepare_next_tasks(
     store: Optional[BaseStore],
     checkpointer: Optional[BaseCheckpointSaver],
     manager: Union[None, ParentRunManager, AsyncParentRunManager],
+    trigger_to_nodes: Optional[Mapping[str, list[str]]] = None,
+    updated_channels: Optional[set[str]] = None,
 ) -> dict[str, PregelExecutableTask]: ...
 
 
@@ -370,10 +384,37 @@ def prepare_next_tasks(
     store: Optional[BaseStore] = None,
     checkpointer: Optional[BaseCheckpointSaver] = None,
     manager: Union[None, ParentRunManager, AsyncParentRunManager] = None,
+    trigger_to_nodes: Optional[Mapping[str, list[str]]] = None,
+    updated_channels: Optional[set[str]] = None,
 ) -> Union[dict[str, PregelTask], dict[str, PregelExecutableTask]]:
     """Prepare the set of tasks that will make up the next Pregel step.
-    This is the union of all PUSH tasks (Sends) and PULL tasks (nodes triggered
-    by edges)."""
+
+    Args:
+        checkpoint: The current checkpoint.
+        pending_writes: The list of pending writes.
+        processes: The mapping of process names to PregelNode instances.
+        channels: The mapping of channel names to BaseChannel instances.
+        managed: The mapping of managed value names to functions.
+        config: The runnable configuration.
+        step: The current step.
+        for_execution: Whether the tasks are being prepared for execution.
+        store: An instance of BaseStore to make it available for usage within tasks.
+        checkpointer: Checkpointer instance used for saving checkpoints.
+        manager: The parent run manager to use for the tasks.
+        trigger_to_nodes: Optional: Mapping of channel names to the set of nodes
+            that are can be triggered by that channel.
+        updated_channels: Optional. Set of channel names that have been updated during
+            the previous step. Using in conjunction with trigger_to_nodes to speed
+            up the process of determining which nodes should be triggered in the next
+            step.
+
+    Returns:
+        A dictionary of tasks to be executed. The keys are the task ids and the values
+        are the tasks themselves. This is the union of all PUSH tasks (Sends)
+        and PULL tasks (nodes triggered by edges).
+    """
+    checkpoint_id_bytes = binascii.unhexlify(checkpoint["id"].replace("-", ""))
+    null_version = checkpoint_null_version(checkpoint)
     tasks: list[Union[PregelTask, PregelExecutableTask]] = []
     # Consume pending_sends from previous step
     for idx, _ in enumerate(checkpoint["pending_sends"]):
@@ -381,6 +422,8 @@ def prepare_next_tasks(
             (PUSH, idx),
             None,
             checkpoint=checkpoint,
+            checkpoint_id_bytes=checkpoint_id_bytes,
+            checkpoint_null_version=null_version,
             pending_writes=pending_writes,
             processes=processes,
             channels=channels,
@@ -393,13 +436,34 @@ def prepare_next_tasks(
             manager=manager,
         ):
             tasks.append(task)
+
+    # This section is an optimization that allows which nodes will be active
+    # during the next step.
+    # When there's information about:
+    # 1. Which channels were updated in the previous step
+    # 2. Which nodes are triggered by which channels
+    # Then we can determine which nodes should be triggered in the next step
+    # without having to cycle through all nodes.
+    if updated_channels and trigger_to_nodes:
+        triggered_nodes: set[str] = set()
+        # Get all nodes that have triggers associated with an updated channel
+        for channel in updated_channels:
+            if node_ids := trigger_to_nodes.get(channel):
+                triggered_nodes.update(node_ids)
+        # Sort the nodes to ensure deterministic order
+        candidate_nodes: Iterable[str] = sorted(triggered_nodes)
+    else:
+        candidate_nodes = processes.keys()
+
     # Check if any processes should be run in next step
     # If so, prepare the values to be passed to them
-    for name in processes:
+    for name in candidate_nodes:
         if task := prepare_single_task(
             (PULL, name),
             None,
             checkpoint=checkpoint,
+            checkpoint_id_bytes=checkpoint_id_bytes,
+            checkpoint_null_version=null_version,
             pending_writes=pending_writes,
             processes=processes,
             channels=channels,
@@ -415,11 +479,16 @@ def prepare_next_tasks(
     return {t.id: t for t in tasks}
 
 
+PUSH_TRIGGER = (PUSH,)
+
+
 def prepare_single_task(
     task_path: tuple[Any, ...],
     task_id_checksum: Optional[str],
     *,
     checkpoint: Checkpoint,
+    checkpoint_id_bytes: bytes,
+    checkpoint_null_version: Optional[V],
     pending_writes: list[PendingWrite],
     processes: Mapping[str, PregelNode],
     channels: Mapping[str, BaseChannel],
@@ -433,7 +502,6 @@ def prepare_single_task(
 ) -> 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."""
-    checkpoint_id = UUID(checkpoint["id"]).bytes
     configurable = config.get(CONF, {})
     parent_ns = configurable.get(CONFIG_KEY_CHECKPOINT_NS, "")
 
@@ -446,10 +514,10 @@ def prepare_single_task(
         if name is None:
             raise ValueError("`call` functions must have a `__name__` attribute")
         # create task id
-        triggers = [PUSH]
+        triggers: Sequence[str] = PUSH_TRIGGER
         checkpoint_ns = f"{parent_ns}{NS_SEP}{name}" if parent_ns else name
         task_id = _uuid5_str(
-            checkpoint_id,
+            checkpoint_id_bytes,
             checkpoint_ns,
             str(step),
             name,
@@ -507,6 +575,7 @@ def prepare_single_task(
                         CONFIG_KEY_CHECKPOINT_ID: None,
                         CONFIG_KEY_CHECKPOINT_NS: task_checkpoint_ns,
                         CONFIG_KEY_SCRATCHPAD: _scratchpad(
+                            config[CONF].get(CONFIG_KEY_SCRATCHPAD),
                             pending_writes,
                             task_id,
                         ),
@@ -539,12 +608,12 @@ def prepare_single_task(
                 )
                 return
             # create task id
-            triggers = [PUSH]
+            triggers = PUSH_TRIGGER
             checkpoint_ns = (
                 f"{parent_ns}{NS_SEP}{packet.node}" if parent_ns else packet.node
             )
             task_id = _uuid5_str(
-                checkpoint_id,
+                checkpoint_id_bytes,
                 checkpoint_ns,
                 str(step),
                 packet.node,
@@ -616,6 +685,7 @@ def prepare_single_task(
                             CONFIG_KEY_CHECKPOINT_ID: None,
                             CONFIG_KEY_CHECKPOINT_NS: task_checkpoint_ns,
                             CONFIG_KEY_SCRATCHPAD: _scratchpad(
+                                config[CONF].get(CONFIG_KEY_SCRATCHPAD),
                                 pending_writes,
                                 task_id,
                             ),
@@ -640,21 +710,17 @@ def prepare_single_task(
         if name not in processes:
             return
         proc = processes[name]
-        version_type = type(next(iter(checkpoint["channel_versions"].values()), None))
-        null_version = version_type()  # type: ignore[misc]
-        if null_version is None:
+        if checkpoint_null_version is None:
             return
-        seen = checkpoint["versions_seen"].get(name, {})
         # If any of the channels read by this process were updated
-        if triggers := sorted(
-            chan
-            for chan in proc.triggers
-            if not isinstance(
-                read_channel(channels, chan, return_exception=True), EmptyChannelError
-            )
-            and checkpoint["channel_versions"].get(chan, null_version)  # type: ignore[operator]
-            > seen.get(chan, null_version)
+        if _triggers(
+            channels,
+            checkpoint["channel_versions"],
+            checkpoint["versions_seen"].get(name),
+            checkpoint_null_version,
+            proc,
         ):
+            triggers = tuple(sorted(proc.triggers))
             try:
                 val = next(
                     _proc_input(proc, managed, channels, for_execution=for_execution)
@@ -671,7 +737,7 @@ def prepare_single_task(
             # create task id
             checkpoint_ns = f"{parent_ns}{NS_SEP}{name}" if parent_ns else name
             task_id = _uuid5_str(
-                checkpoint_id,
+                checkpoint_id_bytes,
                 checkpoint_ns,
                 str(step),
                 name,
@@ -714,7 +780,7 @@ def prepare_single_task(
                                 CONFIG_KEY_SEND: partial(
                                     local_write,
                                     writes.extend,
-                                    processes.keys(),
+                                    tuple(processes.keys()),
                                 ),
                                 CONFIG_KEY_READ: partial(
                                     local_read,
@@ -723,7 +789,10 @@ def prepare_single_task(
                                     channels,
                                     managed,
                                     PregelTaskWrites(
-                                        task_path[:3], name, writes, triggers
+                                        task_path[:3],
+                                        name,
+                                        writes,
+                                        triggers,
                                     ),
                                     config,
                                 ),
@@ -741,6 +810,7 @@ def prepare_single_task(
                                 CONFIG_KEY_CHECKPOINT_ID: None,
                                 CONFIG_KEY_CHECKPOINT_NS: task_checkpoint_ns,
                                 CONFIG_KEY_SCRATCHPAD: _scratchpad(
+                                    config[CONF].get(CONFIG_KEY_SCRATCHPAD),
                                     pending_writes,
                                     task_id,
                                 ),
@@ -761,13 +831,59 @@ def prepare_single_task(
                 return PregelTask(task_id, name, task_path[:3])
 
 
+def checkpoint_null_version(
+    checkpoint: Checkpoint,
+) -> Optional[V]:
+    """Get the null version for the checkpoint, if available."""
+    for version in checkpoint["channel_versions"].values():
+        return type(version)()
+    return None
+
+
+def _triggers(
+    channels: Mapping[str, BaseChannel],
+    versions: ChannelVersions,
+    seen: Optional[ChannelVersions],
+    null_version: V,
+    proc: PregelNode,
+) -> Sequence[str]:
+    if seen is None:
+        for chan in proc.triggers:
+            if channels[chan].is_available():
+                return (chan,)
+    else:
+        for chan in proc.triggers:
+            if channels[chan].is_available() and versions.get(  # type: ignore[operator]
+                chan, null_version
+            ) > seen.get(chan, null_version):
+                return (chan,)
+    return EMPTY_SEQ
+
+
 def _scratchpad(
+    parent_scratchpad: Optional[PregelScratchpad],
     pending_writes: list[PendingWrite],
     task_id: str,
 ) -> PregelScratchpad:
+    # None cannot be used as a resume value, because it would be difficult to
+    # distinguish from missing when used over http
     null_resume_write = next(
         (w for w in pending_writes if w[0] == NULL_TASK_ID and w[1] == RESUME), None
     )
+
+    def get_null_resume(consume: bool = False) -> Any:
+        if null_resume_write is None:
+            if parent_scratchpad is not None:
+                return parent_scratchpad.get_null_resume(consume)
+            return None
+        if consume:
+            try:
+                pending_writes.remove(null_resume_write)
+                return null_resume_write[2]
+            except ValueError:
+                return None
+        return null_resume_write[2]
+
     # using itertools.count as an atomic counter (+= 1 is not thread-safe)
     return PregelScratchpad(
         # call
@@ -777,10 +893,7 @@ def _scratchpad(
         resume=next(
             (w[2] for w in pending_writes if w[0] == task_id and w[1] == RESUME), []
         ),
-        null_resume=null_resume_write[2] if null_resume_write is not None else None,
-        _consume_null_resume=functools.partial(pending_writes.remove, null_resume_write)
-        if null_resume_write is not None
-        else lambda: None,
+        get_null_resume=get_null_resume,
         # subgraph
         subgraph_counter=itertools.count(0).__next__,
     )

libs/langgraph/langgraph/pregel/debug.py

@@ -25,8 +25,10 @@ from langgraph.constants import (
     CONFIG_KEY_CHECKPOINT_NS,
     ERROR,
     INTERRUPT,
+    MISSING,
     NS_END,
     NS_SEP,
+    RETURN,
     TAG_HIDDEN,
 )
 from langgraph.pregel.io import read_channels
@@ -132,8 +134,15 @@ def map_debug_task_results(
             "id": task.id,
             "name": task.name,
             "error": next((w[1] for w in writes if w[0] == ERROR), None),
-            "result": [w for w in writes if w[0] in stream_channels_list],
-            "interrupts": [asdict(w[1]) for w in writes if w[0] == INTERRUPT],
+            "result": [
+                w for w in writes if w[0] in stream_channels_list or w[0] == RETURN
+            ],
+            "interrupts": [
+                asdict(v)
+                for w in writes
+                if w[0] == INTERRUPT
+                for v in (w[1] if isinstance(w[1], Sequence) else [w[1]])
+            ],
         },
     }
 
@@ -264,49 +273,64 @@ def tasks_w_writes(
 ) -> tuple[PregelTask, ...]:
     """Apply writes / subgraph states to tasks to be returned in a StateSnapshot."""
     pending_writes = pending_writes or []
-    return tuple(
-        PregelTask(
-            task.id,
-            task.name,
-            task.path,
-            next(
-                (
-                    exc
-                    for tid, n, exc in pending_writes
-                    if tid == task.id and n == ERROR
-                ),
-                None,
-            ),
-            tuple(
-                v for tid, n, v in pending_writes if tid == task.id and n == INTERRUPT
-            ),
-            states.get(task.id) if states else None,
+    out: list[PregelTask] = []
+    for task in tasks:
+        rtn = next(
             (
+                val
+                for tid, chan, val in pending_writes
+                if tid == task.id and chan == RETURN
+            ),
+            MISSING,
+        )
+        out.append(
+            PregelTask(
+                task.id,
+                task.name,
+                task.path,
                 next(
                     (
-                        val
-                        for tid, chan, val in pending_writes
-                        if tid == task.id and chan == output_keys
+                        exc
+                        for tid, n, exc in pending_writes
+                        if tid == task.id and n == ERROR
                     ),
                     None,
-                )
-                if isinstance(output_keys, str)
-                else {
-                    chan: val
-                    for tid, chan, val in pending_writes
-                    if tid == task.id
-                    and (
-                        chan == output_keys
-                        if isinstance(output_keys, str)
-                        else chan in output_keys
+                ),
+                tuple(
+                    v
+                    for tid, n, vv in pending_writes
+                    if tid == task.id and n == INTERRUPT
+                    for v in (vv if isinstance(vv, Sequence) else [vv])
+                ),
+                states.get(task.id) if states else None,
+                (
+                    rtn
+                    if rtn is not MISSING
+                    else next(
+                        (
+                            val
+                            for tid, chan, val in pending_writes
+                            if tid == task.id and chan == output_keys
+                        ),
+                        None,
                     )
-                }
+                    if isinstance(output_keys, str)
+                    else {
+                        chan: val
+                        for tid, chan, val in pending_writes
+                        if tid == task.id
+                        and (
+                            chan == output_keys
+                            if isinstance(output_keys, str)
+                            else chan in output_keys
+                        )
+                    }
+                )
+                if any(
+                    w[0] == task.id and w[1] not in (ERROR, INTERRUPT)
+                    for w in pending_writes
+                )
+                else None,
             )
-            if any(
-                w[0] == task.id and w[1] not in (ERROR, INTERRUPT)
-                for w in pending_writes
-            )
-            else None,
         )
-        for task in tasks
-    )
+    return tuple(out)

libs/langgraph/langgraph/pregel/io.py

@@ -14,7 +14,6 @@ from langgraph.constants import (
     NULL_TASK_ID,
     RESUME,
     RETURN,
-    SELF,
     START,
     TAG_HIDDEN,
     TASKS,
@@ -28,7 +27,7 @@ def is_task_id(task_id: str) -> bool:
     """Check if a string is a valid task id."""
     try:
         UUID(task_id)
-    except ValueError:
+    except Exception:
         return False
     return True
 
@@ -38,14 +37,11 @@ def read_channel(
     chan: str,
     *,
     catch: bool = True,
-    return_exception: bool = False,
 ) -> Any:
     try:
         return channels[chan].get()
-    except EmptyChannelError as exc:
-        if return_exception:
-            return exc
-        elif catch:
+    except EmptyChannelError:
+        if catch:
             return None
         else:
             raise
@@ -84,7 +80,7 @@ def map_command(
             if isinstance(send, Send):
                 yield (NULL_TASK_ID, TASKS, send)
             elif isinstance(send, str):
-                yield (NULL_TASK_ID, f"branch:{START}:{SELF}:{send}", START)
+                yield (NULL_TASK_ID, f"branch:to:{send}", START)
             else:
                 raise TypeError(
                     f"In Command.goto, expected Send/str, got {type(send).__name__}"

libs/langgraph/langgraph/pregel/loop.py

@@ -1,8 +1,9 @@
 import asyncio
+import binascii
 import concurrent.futures
+import dataclasses
 from collections import defaultdict, deque
 from contextlib import AsyncExitStack, ExitStack
-from dataclasses import replace
 from inspect import signature
 from types import TracebackType
 from typing import (
@@ -24,6 +25,7 @@ from typing import (
 
 from langchain_core.callbacks import AsyncParentRunManager, ParentRunManager
 from langchain_core.runnables import RunnableConfig
+from pydantic import BaseModel
 from typing_extensions import ParamSpec, Self
 
 from langgraph.channels.base import BaseChannel
@@ -55,6 +57,7 @@ from langgraph.constants import (
     ERROR,
     INPUT,
     INTERRUPT,
+    MISSING,
     NS_SEP,
     NULL_TASK_ID,
     PUSH,
@@ -67,7 +70,6 @@ from langgraph.errors import (
     EmptyInputError,
     GraphDelegate,
     GraphInterrupt,
-    ParentCommand,
 )
 from langgraph.managed.base import (
     ManagedValueMapping,
@@ -79,6 +81,7 @@ from langgraph.pregel.algo import (
     GetNextVersion,
     PregelTaskWrites,
     apply_writes,
+    checkpoint_null_version,
     increment,
     prepare_next_tasks,
     prepare_single_task,
@@ -126,6 +129,7 @@ P = ParamSpec("P")
 
 INPUT_DONE = object()
 INPUT_RESUMING = object()
+INPUT_SHOULD_VALIDATE = object()
 SPECIAL_CHANNELS = (ERROR, INTERRUPT, SCHEDULED)
 
 
@@ -140,6 +144,7 @@ def DuplexStream(*streams: StreamProtocol) -> StreamProtocol:
 
 class PregelLoop(LoopProtocol):
     input: Optional[Any]
+    input_model: Optional[Type[BaseModel]]
     checkpointer: Optional[BaseCheckpointSaver]
     nodes: Mapping[str, PregelNode]
     specs: Mapping[str, Union[BaseChannel, ManagedValueSpec]]
@@ -203,7 +208,9 @@ class PregelLoop(LoopProtocol):
         interrupt_after: Union[All, Sequence[str]] = EMPTY_SEQ,
         interrupt_before: Union[All, Sequence[str]] = EMPTY_SEQ,
         manager: Union[None, AsyncParentRunManager, ParentRunManager] = None,
+        input_model: Optional[Type[BaseModel]] = None,
         debug: bool = False,
+        trigger_to_nodes: Optional[Mapping[str, list[str]]] = None,
     ) -> None:
         super().__init__(
             step=0,
@@ -213,6 +220,7 @@ class PregelLoop(LoopProtocol):
             store=store,
         )
         self.input = input
+        self.input_model = input_model
         self.checkpointer = checkpointer
         self.nodes = nodes
         self.specs = specs
@@ -226,6 +234,7 @@ class PregelLoop(LoopProtocol):
             CONFIG_KEY_CHECKPOINT_ID not in config[CONF]
             or CONFIG_KEY_DEDUPE_TASKS in config[CONF]
         )
+        self.trigger_to_nodes = trigger_to_nodes
         self.debug = debug
         if self.stream is not None and CONFIG_KEY_STREAM in config[CONF]:
             self.stream = DuplexStream(self.stream, config[CONF][CONFIG_KEY_STREAM])
@@ -343,12 +352,16 @@ class PregelLoop(LoopProtocol):
         ):
             self.to_interrupt.append(task)
             return
+        checkpoint_id_bytes = binascii.unhexlify(self.checkpoint["id"].replace("-", ""))
+        null_version = checkpoint_null_version(self.checkpoint)
         if pushed := cast(
             Optional[PregelExecutableTask],
             prepare_single_task(
                 (PUSH, task.path, write_idx, task.id, call),
                 None,
                 checkpoint=self.checkpoint,
+                checkpoint_id_bytes=checkpoint_id_bytes,
+                checkpoint_null_version=null_version,
                 pending_writes=self.checkpoint_pending_writes,
                 processes=self.nodes,
                 channels=self.channels,
@@ -396,13 +409,16 @@ class PregelLoop(LoopProtocol):
         if self.status != "pending":
             raise RuntimeError("Cannot tick when status is no longer 'pending'")
 
-        if self.input not in (INPUT_DONE, INPUT_RESUMING):
+        updated_channels: set[str] | None = None
+
+        if self.input not in (INPUT_DONE, INPUT_RESUMING, INPUT_SHOULD_VALIDATE):
             self._first(input_keys=input_keys)
         elif self.to_interrupt:
             # if we need to interrupt, do so
             self.status = "interrupt_before"
             raise GraphInterrupt()
         elif all(task.writes for task in self.tasks.values()):
+            # finish superstep
             writes = [w for t in self.tasks.values() for w in t.writes]
             # debug flag
             if self.debug:
@@ -416,7 +432,7 @@ class PregelLoop(LoopProtocol):
                     ),
                 )
             # all tasks have finished
-            mv_writes = apply_writes(
+            mv_writes, updated_channels = apply_writes(
                 self.checkpoint,
                 self.channels,
                 self.tasks.values(),
@@ -425,6 +441,13 @@ class PregelLoop(LoopProtocol):
             # apply writes to managed values
             for key, values in mv_writes.items():
                 self._update_mv(key, values)
+            # validate input if requested
+            if self.input is INPUT_SHOULD_VALIDATE:
+                self.input = INPUT_DONE
+                # validate
+                cast(Type[BaseModel], self.input_model)(
+                    **read_channels(self.channels, self.stream_keys)
+                )
             # produce values output
             self._emit(
                 "values", map_output_values, self.output_keys, writes, self.channels
@@ -451,6 +474,9 @@ class PregelLoop(LoopProtocol):
             ):
                 self.status = "interrupt_after"
                 raise GraphInterrupt()
+
+            # unset resuming flag
+            self.config[CONF].pop(CONFIG_KEY_RESUMING, None)
         else:
             return False
 
@@ -472,6 +498,8 @@ class PregelLoop(LoopProtocol):
             manager=self.manager,
             store=self.store,
             checkpointer=self.checkpointer,
+            trigger_to_nodes=self.trigger_to_nodes,
+            updated_channels=updated_channels,
         )
         self.to_interrupt = []
 
@@ -550,7 +578,7 @@ class PregelLoop(LoopProtocol):
                         self.checkpoint["versions_seen"].get(INTERRUPT, {}).values(),
                         default=None,
                     ):
-                        self.tasks[tid] = task._replace(scheduled=True)
+                        self.tasks[tid] = dataclasses.replace(task, scheduled=True)
                 else:
                     task.writes.append((k, v))
 
@@ -562,19 +590,16 @@ class PregelLoop(LoopProtocol):
         is_resuming = bool(self.checkpoint["channel_versions"]) and bool(
             configurable.get(
                 CONFIG_KEY_RESUMING,
-                self.input is None or isinstance(self.input, Command),
+                self.input is None
+                or isinstance(self.input, Command)
+                or (
+                    not self.is_nested
+                    and self.config.get("metadata", {}).get("run_id")
+                    == self.checkpoint_metadata.get("run_id", MISSING)
+                ),
             )
         )
 
-        # take resume value from parent
-        if scratchpad := cast(
-            Optional[PregelScratchpad], configurable.get(CONFIG_KEY_SCRATCHPAD)
-        ):
-            if (
-                isinstance(scratchpad, PregelScratchpad)
-                and scratchpad.null_resume is not None
-            ):
-                self.put_writes(NULL_TASK_ID, [(RESUME, scratchpad.null_resume)])
         # map command to writes
         if isinstance(self.input, Command):
             if self.input.resume is not None and not self.checkpointer:
@@ -594,7 +619,7 @@ class PregelLoop(LoopProtocol):
         if null_writes := [
             w[1:] for w in self.checkpoint_pending_writes if w[0] == NULL_TASK_ID
         ]:
-            mv_writes = apply_writes(
+            mv_writes, _ = apply_writes(
                 self.checkpoint,
                 self.channels,
                 [PregelTaskWrites((), INPUT, null_writes, [])],
@@ -613,6 +638,8 @@ class PregelLoop(LoopProtocol):
             self._emit(
                 "values", map_output_values, self.output_keys, True, self.channels
             )
+            # set flag
+            self.input = INPUT_RESUMING
         # map inputs to channel updates
         elif input_writes := deque(map_input(input_keys, self.input)):
             # TODO shouldn't these writes be passed to put_writes too?
@@ -641,7 +668,7 @@ class PregelLoop(LoopProtocol):
                 manager=None,
             )
             # apply input writes
-            mv_writes = apply_writes(
+            mv_writes, _ = apply_writes(
                 self.checkpoint,
                 self.channels,
                 [
@@ -653,10 +680,19 @@ class PregelLoop(LoopProtocol):
             assert not mv_writes, "Can't write to SharedValues in graph input"
             # save input checkpoint
             self._put_checkpoint({"source": "input", "writes": dict(input_writes)})
+            # set flag
+            if (
+                self.input_model is not None
+                and not isinstance(self.input, self.input_model)
+                and not isinstance(self.stream_keys, str)
+            ):
+                self.input = INPUT_SHOULD_VALIDATE
+            else:
+                self.input = INPUT_DONE
         elif CONFIG_KEY_RESUMING not in configurable:
             raise EmptyInputError(f"Received no input for {input_keys}")
-        # done with input
-        self.input = INPUT_RESUMING if is_resuming else INPUT_DONE
+        else:
+            self.input = INPUT_DONE
         # update config
         if not self.is_nested:
             self.config = patch_configurable(
@@ -738,15 +774,6 @@ class PregelLoop(LoopProtocol):
         exc_value: Optional[BaseException],
         traceback: Optional[TracebackType],
     ) -> Optional[bool]:
-        # add current state to parent command
-        if isinstance(exc_value, ParentCommand):
-            cmd = exc_value.args[0]
-            state = (
-                [(self.output_keys, read_channels(self.channels, self.output_keys))]
-                if isinstance(self.output_keys, str)
-                else list(read_channels(self.channels, self.output_keys).items())
-            )
-            exc_value.args = (replace(cmd, update=[*state, *cmd._update_as_tuples()]),)
         # suppress interrupt
         suppress = isinstance(exc_value, GraphInterrupt) and not self.is_nested
         if suppress:
@@ -756,7 +783,7 @@ class PregelLoop(LoopProtocol):
                 and self.checkpoint_pending_writes
                 and any(task.writes for task in self.tasks.values())
             ):
-                mv_writes = apply_writes(
+                mv_writes, _ = apply_writes(
                     self.checkpoint,
                     self.channels,
                     self.tasks.values(),
@@ -771,11 +798,14 @@ class PregelLoop(LoopProtocol):
                     [w for t in self.tasks.values() for w in t.writes],
                     self.channels,
                 )
-            # emit INTERRUPT event
-            self._emit(
-                "updates",
-                lambda: iter([{INTERRUPT: cast(GraphInterrupt, exc_value).args[0]}]),
-            )
+            # emit INTERRUPT if exception is empty (otherwise emitted by put_writes)
+            if exc_value is not None and (not exc_value.args or not exc_value.args[0]):
+                self._emit(
+                    "updates",
+                    lambda: iter(
+                        [{INTERRUPT: cast(GraphInterrupt, exc_value).args[0]}]
+                    ),
+                )
             # save final output
             self.output = read_channels(self.channels, self.output_keys)
             # suppress interrupt
@@ -806,7 +836,25 @@ class PregelLoop(LoopProtocol):
                 "tags", EMPTY_SEQ
             ):
                 return
-            if writes[0][0] != ERROR and writes[0][0] != INTERRUPT:
+            if writes[0][0] == INTERRUPT:
+                self._emit(
+                    "updates",
+                    lambda: iter(
+                        [
+                            {
+                                INTERRUPT: tuple(
+                                    v
+                                    for w in writes
+                                    if w[0] == INTERRUPT
+                                    for v in (
+                                        w[1] if isinstance(w[1], Sequence) else (w[1],)
+                                    )
+                                )
+                            }
+                        ]
+                    ),
+                )
+            elif writes[0][0] != ERROR:
                 self._emit(
                     "updates",
                     map_output_updates,
@@ -840,10 +888,13 @@ class SyncPregelLoop(PregelLoop, ContextManager):
         interrupt_before: Union[All, Sequence[str]] = EMPTY_SEQ,
         output_keys: Union[str, Sequence[str]] = EMPTY_SEQ,
         stream_keys: Union[str, Sequence[str]] = EMPTY_SEQ,
+        input_model: Optional[Type[BaseModel]] = None,
         debug: bool = False,
+        trigger_to_nodes: Optional[Mapping[str, list[str]]] = None,
     ) -> None:
         super().__init__(
             input,
+            input_model=input_model,
             stream=stream,
             config=config,
             checkpointer=checkpointer,
@@ -856,6 +907,7 @@ class SyncPregelLoop(PregelLoop, ContextManager):
             interrupt_before=interrupt_before,
             manager=manager,
             debug=debug,
+            trigger_to_nodes=trigger_to_nodes,
         )
         self.stack = ExitStack()
         if checkpointer:
@@ -979,10 +1031,13 @@ class AsyncPregelLoop(PregelLoop, AsyncContextManager):
         manager: Union[None, AsyncParentRunManager, ParentRunManager] = None,
         output_keys: Union[str, Sequence[str]] = EMPTY_SEQ,
         stream_keys: Union[str, Sequence[str]] = EMPTY_SEQ,
+        input_model: Optional[Type[BaseModel]] = None,
         debug: bool = False,
+        trigger_to_nodes: Optional[Mapping[str, list[str]]] = None,
     ) -> None:
         super().__init__(
             input,
+            input_model=input_model,
             stream=stream,
             config=config,
             checkpointer=checkpointer,
@@ -995,6 +1050,7 @@ class AsyncPregelLoop(PregelLoop, AsyncContextManager):
             interrupt_before=interrupt_before,
             manager=manager,
             debug=debug,
+            trigger_to_nodes=trigger_to_nodes,
         )
         self.stack = AsyncExitStack()
         if checkpointer:

libs/langgraph/langgraph/pregel/protocol.py

@@ -12,7 +12,7 @@ from langchain_core.runnables import Runnable, RunnableConfig
 from langchain_core.runnables.graph import Graph as DrawableGraph
 from typing_extensions import Self
 
-from langgraph.pregel.types import All, StateSnapshot, StreamMode
+from langgraph.pregel.types import All, StateSnapshot, StateUpdate, StreamMode
 
 
 class PregelProtocol(
@@ -69,6 +69,20 @@ class PregelProtocol(
         limit: Optional[int] = None,
     ) -> AsyncIterator[StateSnapshot]: ...
 
+    @abstractmethod
+    def bulk_update_state(
+        self,
+        config: RunnableConfig,
+        updates: Sequence[Sequence[StateUpdate]],
+    ) -> RunnableConfig: ...
+
+    @abstractmethod
+    async def abulk_update_state(
+        self,
+        config: RunnableConfig,
+        updates: Sequence[Sequence[StateUpdate]],
+    ) -> RunnableConfig: ...
+
     @abstractmethod
     def update_state(
         self,

libs/langgraph/langgraph/pregel/read.py

@@ -201,7 +201,6 @@ class PregelNode(Runnable):
             writers[-2] = ChannelWrite(
                 writes=writers[-2].writes + writers[-1].writes,
                 tags=writers[-2].tags,
-                require_at_least_one_of=writers[-2].require_at_least_one_of,
             )
             writers.pop()
         return writers

libs/langgraph/langgraph/pregel/remote.py

@@ -457,6 +457,20 @@ class RemoteGraph(PregelProtocol):
         for state in states:
             yield self._create_state_snapshot(state)
 
+    def bulk_update_state(
+        self,
+        config: RunnableConfig,
+        updates: list[tuple[Optional[dict[str, Any]], Optional[str]]],
+    ) -> RunnableConfig:
+        raise NotImplementedError
+
+    async def abulk_update_state(
+        self,
+        config: RunnableConfig,
+        updates: list[tuple[Optional[dict[str, Any]], Optional[str]]],
+    ) -> RunnableConfig:
+        raise NotImplementedError
+
     def update_state(
         self,
         config: RunnableConfig,

libs/langgraph/langgraph/pregel/runner.py

@@ -2,6 +2,7 @@ import asyncio
 import concurrent.futures
 import threading
 import time
+import weakref
 from functools import partial
 from typing import (
     Any,
@@ -25,12 +26,10 @@ from langgraph.constants import (
     CONF,
     CONFIG_KEY_CALL,
     CONFIG_KEY_SCRATCHPAD,
-    CONFIG_KEY_SEND,
     ERROR,
     INTERRUPT,
     MISSING,
     NO_WRITES,
-    PUSH,
     RESUME,
     RETURN,
     TAG_HIDDEN,
@@ -48,7 +47,9 @@ E = TypeVar("E", threading.Event, asyncio.Event)
 
 class FuturesDict(Generic[F, E], dict[F, Optional[PregelExecutableTask]]):
     event: E
-    callback: Callable[[PregelExecutableTask, Optional[BaseException]], None]
+    callback: weakref.ref[
+        Callable[[PregelExecutableTask, Optional[BaseException]], None]
+    ]
     counter: int
     done: set[F]
     lock: threading.Lock
@@ -56,7 +57,9 @@ class FuturesDict(Generic[F, E], dict[F, Optional[PregelExecutableTask]]):
     def __init__(
         self,
         event: E,
-        callback: Callable[[PregelExecutableTask, Optional[BaseException]], None],
+        callback: weakref.ref[
+            Callable[[PregelExecutableTask, Optional[BaseException]], None]
+        ],
         future_type: Type[F],
         # used for generic typing, newer py supports FutureDict[...](...)
     ) -> None:
@@ -85,7 +88,7 @@ class FuturesDict(Generic[F, E], dict[F, Optional[PregelExecutableTask]]):
         fut: F,
     ) -> None:
         try:
-            self.callback(task, _exception(fut))
+            self.callback()(task, _exception(fut))  # type: ignore[misc]
         finally:
             with self.lock:
                 self.done.add(fut)
@@ -102,10 +105,13 @@ class PregelRunner:
     def __init__(
         self,
         *,
-        submit: Submit,
-        put_writes: Callable[[str, Sequence[tuple[str, Any]]], None],
-        schedule_task: Callable[
-            [PregelExecutableTask, int, Optional[Call]], Optional[PregelExecutableTask]
+        submit: weakref.ref[Submit],
+        put_writes: weakref.ref[Callable[[str, Sequence[tuple[str, Any]]], None]],
+        schedule_task: weakref.ref[
+            Callable[
+                [PregelExecutableTask, int, Optional[Call]],
+                Optional[PregelExecutableTask],
+            ]
         ],
         use_astream: bool = False,
         node_finished: Optional[Callable[[str], None]] = None,
@@ -125,99 +131,9 @@ class PregelRunner:
         retry_policy: Optional[RetryPolicy] = None,
         get_waiter: Optional[Callable[[], concurrent.futures.Future[None]]] = None,
     ) -> Iterator[None]:
-        def writer(
-            task: PregelExecutableTask,
-            writes: Sequence[tuple[str, Any]],
-            *,
-            calls: Optional[Sequence[Call]] = None,
-        ) -> Sequence[Optional[concurrent.futures.Future]]:
-            if all(w[0] != PUSH for w in writes):
-                return task.config[CONF][CONFIG_KEY_SEND](writes)
-
-            # schedule PUSH tasks, collect futures
-            scratchpad: PregelScratchpad = task.config[CONF][CONFIG_KEY_SCRATCHPAD]
-            rtn: dict[int, Optional[concurrent.futures.Future]] = {}
-            for idx, w in enumerate(writes):
-                # bail if not a PUSH write
-                if w[0] != PUSH:
-                    continue
-                # schedule the next task, if the callback returns one
-                wcall = calls[idx] if calls else None
-                if next_task := self.schedule_task(
-                    task, scratchpad.call_counter(), wcall
-                ):
-                    if fut := next(
-                        (
-                            f
-                            for f, t in futures.items()
-                            if t is not None and t == next_task.id
-                        ),
-                        None,
-                    ):
-                        # if the parent task was retried,
-                        # the next task might already be running
-                        rtn[idx] = fut
-                    elif next_task.writes:
-                        # if it already ran, return the result
-                        fut = concurrent.futures.Future()
-                        ret = next(
-                            (v for c, v in next_task.writes if c == RETURN), MISSING
-                        )
-                        if ret is not MISSING:
-                            fut.set_result(ret)
-                        elif exc := next(
-                            (v for c, v in next_task.writes if c == ERROR), None
-                        ):
-                            fut.set_exception(
-                                exc
-                                if isinstance(exc, BaseException)
-                                else Exception(exc)
-                            )
-                        else:
-                            fut.set_result(None)
-                        rtn[idx] = fut
-                    else:
-                        # schedule the next task
-                        fut = self.submit(
-                            run_with_retry,
-                            next_task,
-                            retry_policy,
-                            configurable={
-                                CONFIG_KEY_SEND: partial(writer, next_task),
-                                CONFIG_KEY_CALL: partial(call, next_task),
-                            },
-                            __reraise_on_exit__=reraise,
-                            # starting a new task in the next tick ensures
-                            # updates from this tick are committed/streamed first
-                            __next_tick__=True,
-                        )
-                        futures[fut] = next_task
-                        rtn[idx] = fut
-            return [rtn.get(i) for i in range(len(writes))]
-
-        def call(
-            task: PregelExecutableTask,
-            func: Callable[[Any], Union[Awaitable[Any], Any]],
-            input: Any,
-            *,
-            retry: Optional[RetryPolicy] = None,
-            callbacks: Callbacks = None,
-        ) -> concurrent.futures.Future[Any]:
-            if asyncio.iscoroutinefunction(func):
-                raise RuntimeError("In an sync context async tasks cannot be called")
-            (fut,) = writer(
-                task,
-                [(PUSH, None)],
-                calls=[Call(func, input, retry=retry, callbacks=callbacks)],
-            )
-            assert fut is not None, "writer did not return a future for call"
-            # return a chained future to ensure commit() callback is called
-            # before the returned future is resolved, to ensure stream order etc
-            return chain_future(fut, concurrent.futures.Future())
-
         tasks = tuple(tasks)
         futures = FuturesDict(
-            callback=self.commit,
+            callback=weakref.WeakMethod(self.commit),
             event=threading.Event(),
             future_type=concurrent.futures.Future,
         )
@@ -231,8 +147,15 @@ class PregelRunner:
                     t,
                     retry_policy,
                     configurable={
-                        CONFIG_KEY_SEND: partial(writer, t),
-                        CONFIG_KEY_CALL: partial(call, t),
+                        CONFIG_KEY_CALL: partial(
+                            _call,
+                            weakref.ref(t),
+                            retry=retry_policy,
+                            futures=weakref.ref(futures),
+                            schedule_task=self.schedule_task,
+                            submit=self.submit,
+                            reraise=reraise,
+                        ),
                     },
                 )
                 self.commit(t, None)
@@ -255,13 +178,20 @@ class PregelRunner:
         # schedule tasks
         for t in tasks:
             if not t.writes:
-                fut = self.submit(
+                fut = self.submit()(  # type: ignore[misc]
                     run_with_retry,
                     t,
                     retry_policy,
                     configurable={
-                        CONFIG_KEY_SEND: partial(writer, t),
-                        CONFIG_KEY_CALL: partial(call, t),
+                        CONFIG_KEY_CALL: partial(
+                            _call,
+                            weakref.ref(t),
+                            retry=retry_policy,
+                            futures=weakref.ref(futures),
+                            schedule_task=self.schedule_task,
+                            submit=self.submit,
+                            reraise=reraise,
+                        ),
                     },
                     __reraise_on_exit__=reraise,
                 )
@@ -313,125 +243,10 @@ class PregelRunner:
         retry_policy: Optional[RetryPolicy] = None,
         get_waiter: Optional[Callable[[], asyncio.Future[None]]] = None,
     ) -> AsyncIterator[None]:
-        def writer(
-            task: PregelExecutableTask,
-            writes: Sequence[tuple[str, Any]],
-            *,
-            calls: Optional[Sequence[Call]] = None,
-        ) -> Sequence[Optional[asyncio.Future]]:
-            if all(w[0] != PUSH for w in writes):
-                return task.config[CONF][CONFIG_KEY_SEND](writes)
-
-            # schedule PUSH tasks, collect futures
-            scratchpad: PregelScratchpad = task.config[CONF][CONFIG_KEY_SCRATCHPAD]
-            rtn: dict[int, Optional[asyncio.Future]] = {}
-            for idx, w in enumerate(writes):
-                # bail if not a PUSH write
-                if w[0] != PUSH:
-                    continue
-                # schedule the next task, if the callback returns one
-                wcall = calls[idx] if calls is not None else None
-                if next_task := self.schedule_task(
-                    task, scratchpad.call_counter(), wcall
-                ):
-                    # if the parent task was retried,
-                    # the next task might already be running
-                    if fut := next(
-                        (
-                            f
-                            for f, t in futures.items()
-                            if t is not None and t == next_task.id
-                        ),
-                        None,
-                    ):
-                        # if the parent task was retried,
-                        # the next task might already be running
-                        rtn[idx] = fut
-                    elif next_task.writes:
-                        # if it already ran, return the result
-                        fut = asyncio.Future(loop=loop)
-                        ret = next(
-                            (v for c, v in next_task.writes if c == RETURN), MISSING
-                        )
-                        if ret is not MISSING:
-                            fut.set_result(ret)
-                        elif exc := next(
-                            (v for c, v in next_task.writes if c == ERROR), None
-                        ):
-                            fut.set_exception(
-                                exc
-                                if isinstance(exc, BaseException)
-                                else Exception(exc)
-                            )
-                        else:
-                            fut.set_result(None)
-                        rtn[idx] = fut
-                    else:
-                        # schedule the next task
-                        fut = cast(
-                            asyncio.Future,
-                            self.submit(
-                                arun_with_retry,
-                                next_task,
-                                retry_policy,
-                                stream=self.use_astream,
-                                configurable={
-                                    CONFIG_KEY_SEND: partial(writer, next_task),
-                                    CONFIG_KEY_CALL: partial(call, next_task),
-                                },
-                                __name__=t.name,
-                                __cancel_on_exit__=True,
-                                __reraise_on_exit__=reraise,
-                                # starting a new task in the next tick ensures
-                                # updates from this tick are committed/streamed first
-                                __next_tick__=True,
-                            ),
-                        )
-                        futures[fut] = next_task
-                        rtn[idx] = fut
-            return [rtn.get(i) for i in range(len(writes))]
-
-        def call(
-            task: PregelExecutableTask,
-            func: Callable[[Any], Union[Awaitable[Any], Any]],
-            input: Any,
-            *,
-            retry: Optional[RetryPolicy] = None,
-            callbacks: Callbacks = None,
-        ) -> Union[asyncio.Future[Any], concurrent.futures.Future[Any]]:
-            (fut,) = writer(
-                task,
-                [(PUSH, None)],
-                calls=[Call(func, input, retry=retry, callbacks=callbacks)],
-            )
-            assert fut is not None, "writer did not return a future for call"
-            # return a chained future to ensure commit() callback is called
-            # before the returned future is resolved, to ensure stream order etc
-            try:
-                in_async = asyncio.current_task() is not None
-            except RuntimeError:
-                in_async = False
-            # if in async context return an async future
-            # otherwise return a chained sync future
-            if in_async:
-                if isinstance(fut, asyncio.Task):
-                    sfut: Union[asyncio.Future[Any], concurrent.futures.Future[Any]] = (
-                        asyncio.Future(loop=loop)
-                    )
-                    loop.call_soon_threadsafe(chain_future, fut, sfut)
-                    return sfut
-                else:
-                    # already wrapped in a future
-                    return fut
-            else:
-                sfut = concurrent.futures.Future()
-                loop.call_soon_threadsafe(chain_future, fut, sfut)
-                return sfut
-
         loop = asyncio.get_event_loop()
         tasks = tuple(tasks)
         futures = FuturesDict(
-            callback=self.commit,
+            callback=weakref.WeakMethod(self.commit),
             event=asyncio.Event(),
             future_type=asyncio.Future,
         )
@@ -446,8 +261,17 @@ class PregelRunner:
                     retry_policy,
                     stream=self.use_astream,
                     configurable={
-                        CONFIG_KEY_SEND: partial(writer, t),
-                        CONFIG_KEY_CALL: partial(call, t),
+                        CONFIG_KEY_CALL: partial(
+                            _acall,
+                            weakref.ref(t),
+                            stream=self.use_astream,
+                            retry=retry_policy,
+                            futures=weakref.ref(futures),
+                            schedule_task=self.schedule_task,
+                            submit=self.submit,
+                            reraise=reraise,
+                            loop=loop,
+                        ),
                     },
                 )
                 self.commit(t, None)
@@ -472,14 +296,23 @@ class PregelRunner:
             if not t.writes:
                 fut = cast(
                     asyncio.Future,
-                    self.submit(
+                    self.submit()(  # type: ignore[misc]
                         arun_with_retry,
                         t,
                         retry_policy,
                         stream=self.use_astream,
                         configurable={
-                            CONFIG_KEY_SEND: partial(writer, t),
-                            CONFIG_KEY_CALL: partial(call, t),
+                            CONFIG_KEY_CALL: partial(
+                                _acall,
+                                weakref.ref(t),
+                                retry=retry_policy,
+                                stream=self.use_astream,
+                                futures=weakref.ref(futures),
+                                schedule_task=self.schedule_task,
+                                submit=self.submit,
+                                reraise=reraise,
+                                loop=loop,
+                            ),
                         },
                         __name__=t.name,
                         __cancel_on_exit__=True,
@@ -539,19 +372,20 @@ class PregelRunner:
             # for cancelled tasks, also save error in task,
             # so loop can finish super-step
             task.writes.append((ERROR, exception))
-            self.put_writes(task.id, task.writes)
+            self.put_writes()(task.id, task.writes)  # type: ignore[misc]
         elif exception:
             if isinstance(exception, GraphInterrupt):
                 # save interrupt to checkpointer
-                if interrupts := [(INTERRUPT, i) for i in exception.args[0]]:
+                if exception.args[0]:
+                    writes = [(INTERRUPT, exception.args[0])]
                     if resumes := [w for w in task.writes if w[0] == RESUME]:
-                        interrupts.extend(resumes)
-                    self.put_writes(task.id, interrupts)
+                        writes.extend(resumes)
+                    self.put_writes()(task.id, writes)  # type: ignore[misc]
             elif isinstance(exception, GraphBubbleUp):
                 raise exception
             else:
                 # save error to checkpointer
-                self.put_writes(task.id, [(ERROR, exception)])
+                self.put_writes()(task.id, [(ERROR, exception)])  # type: ignore[misc]
         else:
             if self.node_finished and (
                 task.config is None or TAG_HIDDEN not in task.config.get("tags", [])
@@ -561,7 +395,7 @@ class PregelRunner:
                 # add no writes marker
                 task.writes.append((NO_WRITES, None))
             # save task writes to checkpointer
-            self.put_writes(task.id, task.writes)
+            self.put_writes()(task.id, task.writes)  # type: ignore[misc]
 
 
 def _should_stop_others(
@@ -608,6 +442,7 @@ def _panic_or_proceed(
             done.add(fut)
         else:
             inflight.add(fut)
+    interrupts: list[GraphInterrupt] = []
     while done:
         # if any task failed
         if exc := _exception(done.pop()):
@@ -616,7 +451,14 @@ def _panic_or_proceed(
                 inflight.pop().cancel()
             # raise the exception
             if panic:
-                raise exc
+                if isinstance(exc, GraphInterrupt):
+                    # collect interrupts
+                    interrupts.append(exc)
+                else:
+                    raise exc
+    # raise combined interrupts
+    if interrupts:
+        raise GraphInterrupt(tuple(i for exc in interrupts for i in exc.args[0]))
     if inflight:
         # if we got here means we timed out
         while inflight:
@@ -624,3 +466,193 @@ def _panic_or_proceed(
             inflight.pop().cancel()
         # raise timeout error
         raise timeout_exc_cls("Timed out")
+
+
+def _call(
+    task: weakref.ref[PregelExecutableTask],
+    func: Callable[[Any], Union[Awaitable[Any], Any]],
+    input: Any,
+    *,
+    retry: Optional[RetryPolicy] = None,
+    callbacks: Callbacks = None,
+    futures: weakref.ref[FuturesDict],
+    schedule_task: weakref.ref[
+        Callable[
+            [PregelExecutableTask, int, Optional[Call]], Optional[PregelExecutableTask]
+        ]
+    ],
+    submit: weakref.ref[Submit],
+    reraise: bool,
+) -> concurrent.futures.Future[Any]:
+    if asyncio.iscoroutinefunction(func):
+        raise RuntimeError("In an sync context async tasks cannot be called")
+
+    fut: Optional[concurrent.futures.Future] = None
+    # schedule PUSH tasks, collect futures
+    scratchpad: PregelScratchpad = task().config[CONF][CONFIG_KEY_SCRATCHPAD]  # type: ignore[union-attr]
+    # schedule the next task, if the callback returns one
+    if next_task := schedule_task()(  # type: ignore[misc]
+        task(),  # type: ignore[arg-type]
+        scratchpad.call_counter(),
+        Call(func, input, retry=retry, callbacks=callbacks),
+    ):
+        if fut := next(
+            (
+                f
+                for f, t in futures().items()  # type: ignore[union-attr]
+                if t is not None and t == next_task.id
+            ),
+            None,
+        ):
+            # if the parent task was retried,
+            # the next task might already be running
+            pass
+        elif next_task.writes:
+            # if it already ran, return the result
+            fut = concurrent.futures.Future()
+            ret = next((v for c, v in next_task.writes if c == RETURN), MISSING)
+            if ret is not MISSING:
+                fut.set_result(ret)
+            elif exc := next((v for c, v in next_task.writes if c == ERROR), None):
+                fut.set_exception(
+                    exc if isinstance(exc, BaseException) else Exception(exc)
+                )
+            else:
+                fut.set_result(None)
+        else:
+            # schedule the next task
+            fut = submit()(  # type: ignore[misc]
+                run_with_retry,
+                next_task,
+                retry,
+                configurable={
+                    CONFIG_KEY_CALL: partial(
+                        _call,
+                        weakref.ref(next_task),
+                        futures=futures,
+                        retry=retry,
+                        callbacks=callbacks,
+                        schedule_task=schedule_task,
+                        submit=submit,
+                        reraise=reraise,
+                    ),
+                },
+                __reraise_on_exit__=reraise,
+                # starting a new task in the next tick ensures
+                # updates from this tick are committed/streamed first
+                __next_tick__=True,
+            )
+            futures()[fut] = next_task  # type: ignore[index]
+    fut = cast(Union[asyncio.Future, concurrent.futures.Future], fut)
+    # return a chained future to ensure commit() callback is called
+    # before the returned future is resolved, to ensure stream order etc
+    return chain_future(fut, concurrent.futures.Future())
+
+
+def _acall(
+    task: weakref.ref[PregelExecutableTask],
+    func: Callable[[Any], Union[Awaitable[Any], Any]],
+    input: Any,
+    *,
+    retry: Optional[RetryPolicy] = None,
+    callbacks: Callbacks = None,
+    # injected dependencies
+    futures: weakref.ref[FuturesDict],
+    schedule_task: weakref.ref[
+        Callable[
+            [PregelExecutableTask, int, Optional[Call]], Optional[PregelExecutableTask]
+        ]
+    ],
+    submit: weakref.ref[Submit],
+    loop: asyncio.AbstractEventLoop,
+    reraise: bool = False,
+    stream: bool = False,
+) -> Union[asyncio.Future[Any], concurrent.futures.Future[Any]]:
+    fut: Optional[asyncio.Future] = None
+    # schedule PUSH tasks, collect futures
+    scratchpad: PregelScratchpad = task().config[CONF][CONFIG_KEY_SCRATCHPAD]  # type: ignore[union-attr]
+    # schedule the next task, if the callback returns one
+    if next_task := schedule_task()(  # type: ignore[misc]
+        task(),  # type: ignore[arg-type]
+        scratchpad.call_counter(),
+        Call(func, input, retry=retry, callbacks=callbacks),
+    ):
+        if fut := next(
+            (
+                f
+                for f, t in futures().items()  # type: ignore[union-attr]
+                if t is not None and t == next_task.id
+            ),
+            None,
+        ):
+            # if the parent task was retried,
+            # the next task might already be running
+
+            pass
+        elif next_task.writes:
+            # if it already ran, return the result
+            fut = asyncio.Future(loop=loop)
+            ret = next((v for c, v in next_task.writes if c == RETURN), MISSING)
+            if ret is not MISSING:
+                fut.set_result(ret)
+            elif exc := next((v for c, v in next_task.writes if c == ERROR), None):
+                fut.set_exception(
+                    exc if isinstance(exc, BaseException) else Exception(exc)
+                )
+            else:
+                fut.set_result(None)
+            futures()[fut] = next_task  # type: ignore[index]
+        else:
+            # schedule the next task
+            fut = cast(
+                asyncio.Future,
+                submit()(  # type: ignore[misc]
+                    arun_with_retry,
+                    next_task,
+                    retry,
+                    stream=stream,
+                    configurable={
+                        CONFIG_KEY_CALL: partial(
+                            _acall,
+                            weakref.ref(next_task),
+                            stream=stream,
+                            futures=futures,
+                            schedule_task=schedule_task,
+                            submit=submit,
+                            loop=loop,
+                            reraise=reraise,
+                        ),
+                    },
+                    __name__=task().name,  # type: ignore[union-attr]
+                    __cancel_on_exit__=True,
+                    __reraise_on_exit__=reraise,
+                    # starting a new task in the next tick ensures
+                    # updates from this tick are committed/streamed first
+                    __next_tick__=True,
+                ),
+            )
+            futures()[fut] = next_task  # type: ignore[index]
+
+    fut = cast(Union[asyncio.Future, concurrent.futures.Future], fut)
+    # return a chained future to ensure commit() callback is called
+    # before the returned future is resolved, to ensure stream order etc
+    try:
+        in_async = asyncio.current_task() is not None
+    except RuntimeError:
+        in_async = False
+    # if in async context return an async future
+    # otherwise return a chained sync future
+    if in_async:
+        if isinstance(fut, asyncio.Task):
+            sfut: Union[asyncio.Future[Any], concurrent.futures.Future[Any]] = (
+                asyncio.Future(loop=loop)
+            )
+            loop.call_soon_threadsafe(chain_future, fut, sfut)
+            return sfut
+        else:
+            # already wrapped in a future
+            return fut
+    else:
+        sfut = concurrent.futures.Future()
+        loop.call_soon_threadsafe(chain_future, fut, sfut)
+        return sfut

libs/langgraph/langgraph/pregel/types.py

@@ -7,6 +7,7 @@ from langgraph.types import (
     PregelTask,
     RetryPolicy,
     StateSnapshot,
+    StateUpdate,
     StreamMode,
     StreamWriter,
     default_retry_on,
@@ -14,6 +15,7 @@ from langgraph.types import (
 
 __all__ = [
     "All",
+    "StateUpdate",
     "CachePolicy",
     "PregelExecutableTask",
     "PregelTask",

libs/langgraph/langgraph/pregel/utils.py

@@ -1,7 +1,10 @@
-from typing import Optional
+import ast
+import inspect
+import textwrap
+from typing import Any, Callable, Optional
 
 from langchain_core.runnables import RunnableLambda, RunnableSequence
-from langchain_core.runnables.utils import get_function_nonlocals
+from typing_extensions import override
 
 from langgraph.checkpoint.base import ChannelVersions
 from langgraph.pregel.protocol import PregelProtocol
@@ -55,3 +58,152 @@ def find_subgraph_pregel(candidate: Runnable) -> Optional[PregelProtocol]:
                 )
 
     return None
+
+
+def get_function_nonlocals(func: Callable) -> list[Any]:
+    """Get the nonlocal variables accessed by a function.
+
+    Args:
+        func: The function to check.
+
+    Returns:
+        List[Any]: The nonlocal variables accessed by the function.
+    """
+    try:
+        code = inspect.getsource(func)
+        tree = ast.parse(textwrap.dedent(code))
+        visitor = FunctionNonLocals()
+        visitor.visit(tree)
+        values: list[Any] = []
+        closure = (
+            inspect.getclosurevars(func.__wrapped__)
+            if hasattr(func, "__wrapped__") and callable(func.__wrapped__)
+            else inspect.getclosurevars(func)
+        )
+        candidates = {**closure.globals, **closure.nonlocals}
+        for k, v in candidates.items():
+            if k in visitor.nonlocals:
+                values.append(v)
+            for kk in visitor.nonlocals:
+                if "." in kk and kk.startswith(k):
+                    vv = v
+                    for part in kk.split(".")[1:]:
+                        if vv is None:
+                            break
+                        else:
+                            try:
+                                vv = getattr(vv, part)
+                            except AttributeError:
+                                break
+                    else:
+                        values.append(vv)
+    except (SyntaxError, TypeError, OSError, SystemError):
+        return []
+
+    return values
+
+
+class FunctionNonLocals(ast.NodeVisitor):
+    """Get the nonlocal variables accessed of a function."""
+
+    def __init__(self) -> None:
+        self.nonlocals: set[str] = set()
+
+    @override
+    def visit_FunctionDef(self, node: ast.FunctionDef) -> Any:
+        """Visit a function definition.
+
+        Args:
+            node: The node to visit.
+
+        Returns:
+            Any: The result of the visit.
+        """
+        visitor = NonLocals()
+        visitor.visit(node)
+        self.nonlocals.update(visitor.loads - visitor.stores)
+
+    @override
+    def visit_AsyncFunctionDef(self, node: ast.AsyncFunctionDef) -> Any:
+        """Visit an async function definition.
+
+        Args:
+            node: The node to visit.
+
+        Returns:
+            Any: The result of the visit.
+        """
+        visitor = NonLocals()
+        visitor.visit(node)
+        self.nonlocals.update(visitor.loads - visitor.stores)
+
+    @override
+    def visit_Lambda(self, node: ast.Lambda) -> Any:
+        """Visit a lambda function.
+
+        Args:
+            node: The node to visit.
+
+        Returns:
+            Any: The result of the visit.
+        """
+        visitor = NonLocals()
+        visitor.visit(node)
+        self.nonlocals.update(visitor.loads - visitor.stores)
+
+
+class NonLocals(ast.NodeVisitor):
+    """Get nonlocal variables accessed."""
+
+    def __init__(self) -> None:
+        self.loads: set[str] = set()
+        self.stores: set[str] = set()
+
+    @override
+    def visit_Name(self, node: ast.Name) -> Any:
+        """Visit a name node.
+
+        Args:
+            node: The node to visit.
+
+        Returns:
+            Any: The result of the visit.
+        """
+        if isinstance(node.ctx, ast.Load):
+            self.loads.add(node.id)
+        elif isinstance(node.ctx, ast.Store):
+            self.stores.add(node.id)
+
+    @override
+    def visit_Attribute(self, node: ast.Attribute) -> Any:
+        """Visit an attribute node.
+
+        Args:
+            node: The node to visit.
+
+        Returns:
+            Any: The result of the visit.
+        """
+        if isinstance(node.ctx, ast.Load):
+            parent = node.value
+            attr_expr = node.attr
+            while isinstance(parent, ast.Attribute):
+                attr_expr = parent.attr + "." + attr_expr
+                parent = parent.value
+            if isinstance(parent, ast.Name):
+                self.loads.add(parent.id + "." + attr_expr)
+                self.loads.discard(parent.id)
+            elif isinstance(parent, ast.Call):
+                if isinstance(parent.func, ast.Name):
+                    self.loads.add(parent.func.id)
+                else:
+                    parent = parent.func
+                    attr_expr = ""
+                    while isinstance(parent, ast.Attribute):
+                        if attr_expr:
+                            attr_expr = parent.attr + "." + attr_expr
+                        else:
+                            attr_expr = parent.attr
+                        parent = parent.value
+                    if isinstance(parent, ast.Name):
+                        self.loads.add(parent.id + "." + attr_expr)

libs/langgraph/langgraph/pregel/write.py

@@ -49,21 +49,18 @@ class ChannelWrite(RunnableCallable):
 
     writes: list[Union[ChannelWriteEntry, ChannelWriteTupleEntry, Send]]
     """Sequence of write entries or Send objects to write."""
-    require_at_least_one_of: Optional[Sequence[str]]
-    """If defined, at least one of these channels must be written to."""
 
     def __init__(
         self,
         writes: Sequence[Union[ChannelWriteEntry, ChannelWriteTupleEntry, Send]],
         *,
         tags: Optional[Sequence[str]] = None,
-        require_at_least_one_of: Optional[Sequence[str]] = None,
+        require_at_least_one_of: Optional[Sequence[str]] = None,  # ignored
     ):
         super().__init__(func=self._write, afunc=self._awrite, name=None, tags=tags)
         self.writes = cast(
             list[Union[ChannelWriteEntry, ChannelWriteTupleEntry, Send]], writes
         )
-        self.require_at_least_one_of = require_at_least_one_of
 
     def get_name(
         self, suffix: Optional[str] = None, *, name: Optional[str] = None
@@ -96,7 +93,6 @@ class ChannelWrite(RunnableCallable):
         self.do_write(
             config,
             writes,
-            self.require_at_least_one_of if input is not None else None,
         )
         return input
 
@@ -112,7 +108,6 @@ class ChannelWrite(RunnableCallable):
         self.do_write(
             config,
             writes,
-            self.require_at_least_one_of if input is not None else None,
         )
         return input
 
@@ -120,7 +115,7 @@ class ChannelWrite(RunnableCallable):
     def do_write(
         config: RunnableConfig,
         writes: Sequence[Union[ChannelWriteEntry, ChannelWriteTupleEntry, Send]],
-        require_at_least_one_of: Optional[Sequence[str]] = None,
+        require_at_least_one_of: Optional[Sequence[str]] = None,  # ignored
     ) -> None:
         # validate
         for w in writes:
@@ -151,12 +146,6 @@ class ChannelWrite(RunnableCallable):
                 tuples.append((w.channel, value))
             else:
                 raise ValueError(f"Invalid write entry: {w}")
-        # assert required channels
-        if require_at_least_one_of is not None:
-            if not {chan for chan, _ in tuples} & set(require_at_least_one_of):
-                raise InvalidUpdateError(
-                    f"Must write to at least one of {require_at_least_one_of}"
-                )
         write: TYPE_SEND = config[CONF][CONFIG_KEY_SEND]
         write(tuples)
 

libs/langgraph/langgraph/types.py

@@ -130,7 +130,12 @@ class Interrupt:
     value: Any
     resumable: bool = False
     ns: Optional[Sequence[str]] = None
-    when: Literal["during"] = "during"
+    when: Literal["during"] = dataclasses.field(default="during", repr=False)
+
+
+class StateUpdate(NamedTuple):
+    values: Optional[dict[str, Any]]
+    as_node: Optional[str] = None
 
 
 class PregelTask(NamedTuple):
@@ -140,16 +145,23 @@ class PregelTask(NamedTuple):
     error: Optional[Exception] = None
     interrupts: tuple[Interrupt, ...] = ()
     state: Union[None, RunnableConfig, "StateSnapshot"] = None
-    result: Optional[dict[str, Any]] = None
+    result: Optional[Any] = None
 
 
-class PregelExecutableTask(NamedTuple):
+if sys.version_info > (3, 11):
+    _T_DC_KWARGS = {"weakref_slot": True, "slots": True, "frozen": True}
+else:
+    _T_DC_KWARGS = {"frozen": True}
+
+
+@dataclasses.dataclass(**_T_DC_KWARGS)
+class PregelExecutableTask:
     name: str
     input: Any
     proc: Runnable
     writes: deque[tuple[str, Any]]
     config: RunnableConfig
-    triggers: list[str]
+    triggers: Sequence[str]
     retry_policy: Optional[RetryPolicy]
     cache_policy: Optional[CachePolicy]
     id: str
@@ -351,20 +363,11 @@ class PregelScratchpad:
     call_counter: Callable[[], int]
     # interrupt
     interrupt_counter: Callable[[], int]
+    get_null_resume: Callable[[bool], Any]
     resume: list[Any]
-    null_resume: Optional[Any]
-    _consume_null_resume: Callable[[], None]
     # subgraph
     subgraph_counter: Callable[[], int]
 
-    def consume_null_resume(self) -> Any:
-        if self.null_resume is not None:
-            value = self.null_resume
-            self._consume_null_resume()
-            self.null_resume = None
-            return value
-        raise ValueError("No null resume to consume")
-
 
 def interrupt(value: Any) -> Any:
     """Interrupt the graph with a resumable exception from within a node.
@@ -480,9 +483,9 @@ def interrupt(value: Any) -> Any:
         if idx < len(scratchpad.resume):
             return scratchpad.resume[idx]
     # find current resume value
-    if scratchpad.null_resume is not None:
+    v = scratchpad.get_null_resume(True)
+    if v is not None:
         assert len(scratchpad.resume) == idx, (scratchpad.resume, idx)
-        v = scratchpad.consume_null_resume()
         scratchpad.resume.append(v)
         conf[CONFIG_KEY_SEND]([(RESUME, scratchpad.resume)])
         return v

libs/langgraph/langgraph/utils/config.py

@@ -1,4 +1,5 @@
 from collections import ChainMap
+from os import getenv
 from typing import Any, Optional, Sequence, cast
 
 from langchain_core.callbacks import (
@@ -11,7 +12,6 @@ from langchain_core.runnables import RunnableConfig
 from langchain_core.runnables.config import (
     CONFIG_KEYS,
     COPIABLE_KEYS,
-    DEFAULT_RECURSION_LIMIT,
     var_child_runnable_config,
 )
 
@@ -26,6 +26,8 @@ from langgraph.constants import (
     NS_SEP,
 )
 
+DEFAULT_RECURSION_LIMIT = int(getenv("LANGGRAPH_DEFAULT_RECURSION_LIMIT", "25"))
+
 
 def recast_checkpoint_ns(ns: str) -> str:
     """Remove task IDs from checkpoint namespace.

libs/langgraph/langgraph/utils/pydantic.py

@@ -1,5 +1,9 @@
+import sys
+import typing
+from dataclasses import is_dataclass
 from typing import Any, Dict, Optional, Union
 
+import typing_extensions
 from pydantic import BaseModel
 from pydantic.v1 import BaseModel as BaseModelV1
 
@@ -35,3 +39,31 @@ def create_model(
             v1_kwargs["__root__"] = root
 
         return create_model(model_name, **v1_kwargs, **(field_definitions or {}))
+
+
+def is_supported_by_pydantic(type_: Any) -> bool:
+    """Check if a given "complex" type is supported by pydantic.
+
+    This will return False for primitive types like int, str, etc.
+
+    The check is meant for container types like dataclasses, TypedDicts, etc.
+    """
+    if is_dataclass(type_):
+        return True
+
+    # Pydantic does not support mixing .v1 and root namespaces, so
+    # we only check for BaseModel (not pydantic.v1.BaseModel).
+    if isinstance(type_, type) and issubclass(type_, BaseModel):
+        return True
+
+    if hasattr(type_, "__orig_bases__"):
+        for base in type_.__orig_bases__:
+            if base is typing_extensions.TypedDict:
+                return True
+            elif base is typing.TypedDict:  # noqa: TID251
+                # ignoring TID251 since it's OK to use typing.TypedDict in this case.
+                # Pydantic supports typing.TypedDict from Python 3.12
+                # For older versions, only typing_extensions.TypedDict is supported.
+                if sys.version_info >= (3, 12):
+                    return True
+    return False

libs/langgraph/langgraph/utils/runnable.py

@@ -2,8 +2,8 @@ import asyncio
 import enum
 import inspect
 import sys
-from contextlib import AsyncExitStack
-from contextvars import copy_context
+from contextlib import AsyncExitStack, contextmanager
+from contextvars import Context, Token, copy_context
 from functools import partial, wraps
 from typing import (
     Any,
@@ -11,6 +11,7 @@ from typing import (
     Awaitable,
     Callable,
     Coroutine,
+    Generator,
     Iterator,
     Optional,
     Protocol,
@@ -53,13 +54,69 @@ from langgraph.utils.config import (
     patch_config,
 )
 
-try:
-    from langchain_core.runnables.config import _set_config_context
-except ImportError:
-    # For forwards compatibility
-    def _set_config_context(context: RunnableConfig) -> None:  # type: ignore
-        """Set the context for the current thread."""
-        var_child_runnable_config.set(context)
+
+def _set_config_context(
+    config: RunnableConfig,
+) -> tuple[Token[Optional[RunnableConfig]], Optional[dict[str, Any]]]:
+    """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
+
+        current_context = get_tracing_context()
+        _set_tracing_context({"parent": run})
+    return config_token, current_context
+
+
+@contextmanager
+def set_config_context(config: RunnableConfig) -> Generator[Context, None, None]:
+    """Set the child Runnable config + tracing context.
+
+    Args:
+        config (RunnableConfig): The config to set.
+    """
+    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,
+            {
+                "parent": None,
+                "project_name": None,
+                "tags": None,
+                "metadata": None,
+                "enabled": None,
+                "client": None,
+            },
+        )
 
 
 # Before Python 3.11 native StrEnum is not available
@@ -286,7 +343,6 @@ class RunnableCallable(Runnable):
 
             kwargs[kw] = _conf.get(config_key, default_value)
 
-        context = copy_context()
         if self.trace:
             callback_manager = get_callback_manager_for_config(config, self.tags)
             run_manager = callback_manager.on_chain_start(
@@ -297,17 +353,16 @@ class RunnableCallable(Runnable):
             )
             try:
                 child_config = patch_config(config, callbacks=run_manager.get_child())
-                context = copy_context()
-                context.run(_set_config_context, child_config)
-                ret = context.run(self.func, *args, **kwargs)
+                with set_config_context(child_config) as context:
+                    ret = context.run(self.func, *args, **kwargs)
             except BaseException as e:
                 run_manager.on_chain_error(e)
                 raise
             else:
                 run_manager.on_chain_end(ret)
         else:
-            context.run(_set_config_context, config)
-            ret = context.run(self.func, *args, **kwargs)
+            with set_config_context(config) as context:
+                ret = context.run(self.func, *args, **kwargs)
         if isinstance(ret, Runnable) and self.recurse:
             return ret.invoke(input, config)
         return ret
@@ -342,7 +397,6 @@ class RunnableCallable(Runnable):
                     f"Missing required config key '{config_key}' for '{self.name}'."
                 )
             kwargs[kw] = _conf.get(config_key, default_value)
-        context = copy_context()
         if self.trace:
             callback_manager = get_async_callback_manager_for_config(config, self.tags)
             run_manager = await callback_manager.on_chain_start(
@@ -353,24 +407,24 @@ class RunnableCallable(Runnable):
             )
             try:
                 child_config = patch_config(config, callbacks=run_manager.get_child())
-                context.run(_set_config_context, child_config)
-                coro = cast(Coroutine[None, None, Any], self.afunc(*args, **kwargs))
-                if ASYNCIO_ACCEPTS_CONTEXT:
-                    ret = await asyncio.create_task(coro, context=context)
-                else:
-                    ret = await coro
+                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)
+                    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:
-            context.run(_set_config_context, config)
-            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)
+            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:
             return await ret.ainvoke(input, config)
         return ret