# Lesson 4: Persistence and Streaming -\> AttributeError: '\_GeneratorContextManager' object has no attribute 'get\_next\_version' **URL:** \ **Category:** AI Agents in LangGraph\ **Created:** [September 24, 2024, 4:20pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391 "2024-09-24T16:20:43Z")\ **Posts on this page:** 14\ **Page:** 1
**Author:** ![Jordi\_Lucas\_Rial](https://avatars.discourse-cdn.com/v4/letter/j/45deac/32.png) [@Jordi\_Lucas\_Rial](https://community.deeplearning.ai/u/Jordi_Lucas_Rial)\ **Post date:** [September 24, 2024, 4:20pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/1 "2024-09-24T16:20:43Z")
Hi everyone! I have the error bellow after that I execute this code: for event in abot.graph.stream({“messages”: messages}, thread): for v in event.values(): print(v[‘messages’]) * * * → 663 self.checkpointer\_get\_next\_version = checkpointer.get\_next\_version 664 self.checkpointer\_put\_writes = checkpointer.put\_writes 665 else: AttributeError: ‘\_GeneratorContextManager’ object has no attribute ‘get\_next\_version’ Someone with the same problem? Thanks! ---
**Author:** ![Jordi\_Lucas\_Rial](https://avatars.discourse-cdn.com/v4/letter/j/45deac/32.png) [@Jordi\_Lucas\_Rial](https://community.deeplearning.ai/u/Jordi_Lucas_Rial)\ **Post date:** [September 24, 2024, 4:52pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/2 "2024-09-24T16:52:33Z")
It works if you put all code that depens on “memory” inside a **with** declaration: ``` with SqliteSaver.from_conn_string(":memory:") as memory: abot = Agent(model, [tool], system=prompt, checkpointer=memory) messages = [HumanMessage(content="What is the weather in sf?")] thread = {"configurable": {"thread_id": "1"}} for event in abot.graph.stream({"messages": messages}, thread): for v in event.values(): print(v['messages']) ``` Someone with a solution without refactoring? Thanks! ---
**Author:** ![wnakano](https://avatars.discourse-cdn.com/v4/letter/w/0ea827/32.png) [@wnakano](https://community.deeplearning.ai/u/wnakano)\ **Post date:** [October 22, 2024, 2:20am UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/3 "2024-10-22T02:20:52Z")
There is a quite forward solution. Create a .db file (e.g checkpoints.db). If you created this .db file in the same level as the script/notebook, you can go straight: ```python from langgraph.checkpoint.sqlite import SqliteSaver db_path = 'checkpoints.db' conn = sqlite3.connect(db_path, check_same_thread=False) memory = SqliteSaver(conn) ``` Remark: the addition of `check_same_thread=False` comes from the requirement of the db connection and its writer being available in the same thread. ---
**Author:** ![jxstanford](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/jxstanford/32/98500_2.png) [@jxstanford](https://community.deeplearning.ai/u/jxstanford)\ **Post date:** [October 27, 2024, 12:52am UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/4 "2024-10-27T00:52:53Z")
This works for the synchronous example. The code can be spread across cells while still using the context. ```auto from contextlib import ExitStack stack = ExitStack() memory = stack.enter_context(SqliteSaver.from_conn_string(":memory:")) abot = Agent(model, [tool], system=prompt, checkpointer=memory) messages = [HumanMessage(content="What is the weather in Issaquah?")] thread = {"configurable": {"thread_id": "1"}} for event in abot.graph.stream({"messages": messages}, thread): for v in event.values(): print(v['messages']) messages = [HumanMessage(content="What about in la?")] thread = {"configurable": {"thread_id": "1"}} for event in abot.graph.stream({"messages": messages}, thread): for v in event.values(): print(v) messages = [HumanMessage(content="Which one is warmer?")] thread = {"configurable": {"thread_id": "1"}} for event in abot.graph.stream({"messages": messages}, thread): for v in event.values(): print(v) messages = [HumanMessage(content="Which one is warmer?")] thread = {"configurable": {"thread_id": "2"}} for event in abot.graph.stream({"messages": messages}, thread): for v in event.values(): print(v) # Exit all contexts stack.close() ``` Here is a similar solution for the async version: ```auto import asyncio from contextlib import AsyncExitStack from langgraph.checkpoint.sqlite.aio import AsyncSqliteSaver stack = AsyncExitStack() memory = await stack.enter_async_context(AsyncSqliteSaver.from_conn_string(":memory:")) abot = Agent(model, [tool], system=prompt, checkpointer=memory) messages = [HumanMessage(content="What is the weather in SF?")] thread = {"configurable": {"thread_id": "4"}} async for event in abot.graph.astream_events({"messages": messages}, thread, version="v1"): kind = event["event"] if kind == "on_chat_model_stream": content = event["data"]["chunk"].content if content: # Empty content in the context of OpenAI means # that the model is asking for a tool to be invoked. # So we only print non-empty content print(content, end="|") await stack.aclose() ``` ---
**Author:** ![linux2ram](https://avatars.discourse-cdn.com/v4/letter/l/f14d63/32.png) [@linux2ram](https://community.deeplearning.ai/u/linux2ram)\ **Post date:** [October 31, 2024, 8:46pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/5 "2024-10-31T20:46:55Z")
from langgraph.checkpoint.memory import MemorySaver memory = MemorySaver() fixed the issue ---
**Author:** ![aifaber](https://avatars.discourse-cdn.com/v4/letter/a/9de053/32.png) [@aifaber](https://community.deeplearning.ai/u/aifaber)\ **Post date:** [November 11, 2024, 11:54am UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/6 "2024-11-11T11:54:31Z")
memory = MemorySaver() It works. !! Then What about “AsyncSqliteSaver”? ---
**Author:** ![milcio79](https://avatars.discourse-cdn.com/v4/letter/m/919ad9/32.png) [@milcio79](https://community.deeplearning.ai/u/milcio79)\ **Post date:** [December 2, 2024, 4:45pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/7 "2024-12-02T16:45:46Z")
I changed one function from this: ```auto def call_openai(self, state: AgentState): messages = model.invoke(state['messages']) if self.system: messages = [SystemMessage(content=self.system)] + messages message = self.model.invoke(messages) return {'messages': [message]} ``` to this: ```auto def call_openai(self, state): messages = state['messages'] if self.system: messages = [SystemMessage(content=self.system)] + messages response = self.model.invoke(messages) return {'messages': [response]} ``` and changed every interaction and call to the model, that were like this: ```auto for event in abot.graph.stream({"messages": messages}, thread): for v in event.values(): print(v['messages']) ``` to this: ```auto with SqliteSaver.from_conn_string(":memory:") as checkpointer: agent = Agent(model, [tool], checkpointer, system=prompt) # Mensajes de prueba messages = [HumanMessage(content="What is the weather in SF?")] thread = {"configurable": {"thread_id": "1"}} # Streaming de eventos for event in agent.graph.stream({"messages": messages}, thread): for value in event.values(): print(value['messages']) ``` And now is working well!!! ---
**Author:** ![Jenys](https://avatars.discourse-cdn.com/v4/letter/j/eb8c5e/32.png) [@Jenys](https://community.deeplearning.ai/u/Jenys)\ **Post date:** [January 3, 2025, 1:58pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/8 "2025-01-03T13:58:42Z")
Hi, Have you figured out how to fix it for this part: from langgraph.checkpoint.aiosqlite import AsyncSqliteSaver memory = AsyncSqliteSaver.from\_conn\_string(“:memory:”) It is clear that from langgraph.checkpoint.sqlite.aio import AsyncSqliteSaver should be used instead but this is still not working: memory = AsyncSqliteSaver.from\_conn\_string(“:memory:”) and if I use: memory = MemorySaver() I don’t get the tokens ☹ I implemented 5 different versions with the suggestions from here and also from gpt-o1 but none returns tokens, even though tokenstreaming is set to true for the model. According to gpt-o1 debugging: Based on your debug output, your model is sending _only a single chunk_ with either an empty content or the full final content—i.e. **no incremental tokens** at all. It would be good if this lesson is updated to reflect the current state. ---
**Author:** ![fanzhou0111](https://avatars.discourse-cdn.com/v4/letter/f/6bbea6/32.png) [@fanzhou0111](https://community.deeplearning.ai/u/fanzhou0111)\ **Post date:** [January 9, 2025, 5:16pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/9 "2025-01-09T17:16:32Z")
Thank you, this works for me. Appreciate! ---
**Author:** ![fanzhou0111](https://avatars.discourse-cdn.com/v4/letter/f/6bbea6/32.png) [@fanzhou0111](https://community.deeplearning.ai/u/fanzhou0111)\ **Post date:** [January 9, 2025, 5:48pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/10 "2025-01-09T17:48:11Z")
this helped me. Thank you ! ---
**Author:** ![cde](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/cde/32/65928_2.png) [@cde](https://community.deeplearning.ai/u/cde)\ **Post date:** [April 21, 2025, 7:22pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/11 "2025-04-21T19:22:22Z")
I know it’s been many months since your reply. Thank you BTW. I just wondering if for lesson 5 you also yse `MemorySaver`, and if you did, could you please share your implementation? It was working for me at first, but after I clear all outputs and restart, I keep getting `IndexError: list index out of range` I would really appreciate your help! ---
**Author:** ![cde](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/cde/32/65928_2.png) [@cde](https://community.deeplearning.ai/u/cde)\ **Post date:** [April 21, 2025, 8:05pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/12 "2025-04-21T20:05:30Z")
NVM, I realized that there was a mistake in the `reduce` function. The IDE re-arranged the indentation around the “for” loop. Thanks! ---
**Author:** ![emilio\_cobog](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/emilio_cobog/32/136785_2.png) [@emilio\_cobog](https://community.deeplearning.ai/u/emilio_cobog)\ **Post date:** [May 15, 2025, 10:06pm UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/13 "2025-05-15T22:06:05Z")
It’s true, ‘from langgraph.checkpoint.sqlite.aio import AsyncSqliteSaver’ works, but then, when executing the streaming tokens funcionality I got this error: AttributeError: ‘\_AsyncGeneratorContextManager’ object has no attribute ‘get\_next\_version’ Have you solved it by any chance? ---
**Author:** ![yinhe](https://avatars.discourse-cdn.com/v4/letter/y/f05b48/32.png) [@yinhe](https://community.deeplearning.ai/u/yinhe)\ **Post date:** [June 13, 2025, 8:41am UTC](https://community.deeplearning.ai/t/lesson-4-persistence-and-streaming-attributeerror-generatorcontextmanager-object-has-no-attribute-get-next-version/697391/14 "2025-06-13T08:41:33Z")
```auto import asyncio from langgraph.checkpoint.sqlite.aio import AsyncSqliteSaver from langgraph.prebuilt import create_react_agent from common.llm_models import gpt_4o_mini async def main(): # Create an AsyncSqliteSaver asynchronous context manager saver = AsyncSqliteSaver.from_conn_string("checkpoints.db") # Enter the asynchronous context manager to get the actual checkpointer instance checkpointer = await saver. __aenter__ () agent = create_react_agent(gpt_4o_mini, tools=[], checkpointer=checkpointer) response = agent.astream({"messages": "who are you?", "remaining_steps": 3}, stream_mode="messages", config={"configurable": {"thread_id": "123"}} ) async for chunk in response: print(chunk) await saver. __aexit__ (None, None, None) if __name__ == ' __main__': asyncio.run(main()) ``` If `async with` is not used, the above approach can be used to manage the context manually. The `async with` statement automatically calls ` __aenter__ ()` to obtain the `checkpointer` object.