How to clear state in langgraph?

[yhygta]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

How to clear state in langgraph?

Description

How to clear state in langgraph?

System Info

python 3.10

[dosubot[bot]]

Hey @yhygta, back for more LangChain adventures, I see! 🚀

Based on the context provided, it seems like you're looking to clear the state in LangGraph. However, the term "LangGraph" is not explicitly defined in the LangChain framework. It's inferred that you might be referring to the Graph class or the langgraph module.

If you're referring to the Graph class, it doesn't have a direct method to clear its state. The Graph class manages nodes and edges, where nodes can represent various entities and edges represent connections between these nodes. The class provides methods for adding and removing nodes and edges, but not for clearing the entire state.

If you're referring to the langgraph module, it's designed to facilitate the creation of graph indexes from text. It doesn't have a direct method to clear its state either.

However, if you're looking to clear the state of a conversation, you can do so by calling the clear() method on the chat_memory attribute of a ConversationBufferMemory instance. Here's how you can do it:

memory.chat_memory.clear()

This line of code effectively clears the state by removing all messages from the chat_memory.

If you're looking to clear the state of the SingleStoreDBChatMessageHistory class, you can do so by calling the clear() method on an instance of the class. Here's how you can do it:

message_history = SingleStoreDBChatMessageHistory(session_id="my-session", host="127.0.0.1", port=3306, user="user", password="password", database="db")
message_history.clear()

This will clear all messages associated with the session_id from the SingleStoreDB database.

Please note that clearing the state can have several potential side effects or consequences such as loss of cached data, performance impact, cost implications, concurrency considerations, and keyword arguments impact.

If none of these solutions apply to your situation, could you please provide more details about what you mean by "clearing the state in LangGraph"? This will help me provide a more accurate solution.

Sources

• libs/langchain/tests/integration_tests/memory/test_momento.py
• libs/core/langchain_core/runnables/graph.py
• libs/community/langchain_community/chat_message_histories/singlestoredb.py
• libs/core/langchain_core/caches.py
• libs/langchain/langchain/indexes/graph.py

About Dosu

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.

[yassineselmi]

Ideally, we should call a function in the checkpointer (ex: checkpointer.clear(config)).

In the meantime, I call the put function to update the DB entry with an empty checkpoint record (calling the utility function empty_checkpoint(). This is what I'm actually using and it works for my case:

def clear_memory(memory: BaseCheckpointSaver, thread_id: str) -> None:
    checkpoint = empty_checkpoint()
    memory.put(config={"configurable": {"thread_id": thread_id}}, checkpoint=checkpoint, metadata={})

# Calling the function

memory = SqliteSaver.from_conn_string("checkpoints.sqlite")
app = graph.compile(checkpointer=memory)

clear_memory(memory=memory, thread_id="123456")

[jaymangukiya0001]

Hi @yassineselmi

do you have idea that how do i clear the memory when flow reaches to the END node. I am creating the thread id using user's email as I am having multiple user's conversations going on. How do I clear the memory when flow reaches to the end.

[shubhankar1477]

Hi @yassineselmi

So i am serving langgprah as fastapi and clearing memory after each request , but i am getting error pasted below.
And using MemorySaver() not Sqllite saver

ERROR : self.step = self.checkpoint_metadata["step"] + 1
KeyError: 'step'

[Ishankoradia]

memory.put looks like an internal function, i dont like the idea of using it. I hope they take it as a feature/enhancement request & add a clear method.

i am using a PostgresSaver and i did this . Simply delete all refernces of that thread

 try:
        with conn.cursor() as cursor:
            # Start a transaction block
            cursor.execute("BEGIN;")

            # Execute multiple delete queries
            delete_query_1 = "DELETE FROM checkpoint_blobs WHERE thread_id = %s"
            delete_query_2 = "DELETE FROM checkpoint_writes WHERE thread_id = %s"
            delete_query_3 = "DELETE FROM checkpoints WHERE thread_id = %s"
            cursor.execute(delete_query_1, (thread_id,))
            cursor.execute(delete_query_2, (thread_id,))
            cursor.execute(delete_query_3, (thread_id,))

            # Commit the transaction
            cursor.execute("COMMIT;")
            print(f"Deleted all references of thread_id: {thread_id}")
    except Exception as e:
        # Rollback the transaction in case of error
        cursor.execute("ROLLBACK;")
        print(f"Error deleting thread_id {thread_id}: {e}")

[vigneshmj1997]

It there a way to clear one variable of the state ..? Rather reset the whole state

[arunzone]

Preferably Annotated ones

[daher928]

Hello, any updates on this matter?

[heitoralthmann]

Here's my current implementation of clear_memory function presented above – works with InMemorySaver (using LangGraph v0.2.72):

from langgraph.checkpoint.base import BaseCheckpointSaver

def clear_memory(memory: BaseCheckpointSaver, thread_id: str) -> None:
    """ Clear the memory for a given thread_id. """
    try:
        # If it's an InMemorySaver (which MemorySaver is an alias for),
        # we can directly clear the storage and writes
        if hasattr(memory, 'storage') and hasattr(memory, 'writes'):
            # Clear all checkpoints for this thread_id (all namespaces)
            memory.storage.pop(thread_id, None)

            # Clear all writes for this thread_id (for all namespaces)
            keys_to_remove = [key for key in memory.writes.keys() if key[0] == thread_id]
            for key in keys_to_remove:
                memory.writes.pop(key, None)

            print(f"Memory cleared for thread_id: {thread_id}")
            return

    except Exception as e:
        print(f"Error clearing InMemorySaver storage for thread_id {thread_id}: {e}")

[juanmadlg]

I don't want to make noise in the thread, but thinking about a use case with a chatbot where the thread_id represents each conversation, it would be very useful to have a method that clears all data associated with a thread_id that hasn't had any data saved or accessed for a certain period of time.