AttributeError: '_GeneratorContextManager' object has no attribute 'get_next_version'

[smanna123]

when i am trying to langgraph customer support example (https://langchain-ai.github.io/langgraph/tutorials/customer-support/customer-support/) , it is facing below error---

Traceback (most recent call last):
File "/home/broadifi/Desktop/ofc_project/lang_graph_bot/agentic_bot.py", line 406, in
for event in events:
File "/home/broadifi/Desktop/ofc_project/lang_graph_bot/venv/lib/python3.10/site-packages/langgraph/pregel/init.py", line 1190, in stream
with SyncPregelLoop(
File "/home/broadifi/Desktop/ofc_project/lang_graph_bot/venv/lib/python3.10/site-packages/langgraph/pregel/loop.py", line 562, in init
self.checkpointer_get_next_version = checkpointer.get_next_version
AttributeError: '_GeneratorContextManager' object has no attribute 'get_next_version'

with the older version it was working fine but after upgradation it is facing above error

[vbarda]

@smanna123 it looks like you're using SqliteSaver.from_conn_string() as a checkpointer, but it actually is a context manager. the right way to use it is

with SqliteSaver.from_conn_string(...) as checkpointer:
    app = graph.compile(checkpointer=checkpointer)
    app.invoke(...)

hope this helps!

[trex55]

I'm getting the similar errors. What version of the packages can I use to make this work as per what is in the course without going down dependency hell?

[Vineetk20-git]

prompt = """You are a smart research assistant. Use the search engine to look up information.
You are allowed to make multiple calls (either together or in sequence).
Only look up information when you are sure of what you want.
If you need to look up some information before asking a follow-up question, you are allowed to do that!
"""
model = ChatGroq(temperature=0, model="Llama-3.3-70b-Specdec")

with SqliteSaver.from_conn_string(f"sqlite:///{db_path}") as memory:
bot = Agent(model, [tool], system=prompt, checkpointer=memory)

messages = [HumanMessage(content="What is the weather in Texas?")]
thread = {"configurable": {"thread_id": "1"}}

for event in bot.graph.stream({"messages": messages}, thread):
    for v in event.values():
        print(v['messages'])

trying this, but getting the following error:

OperationalError Traceback (most recent call last)
Cell In[6], line 8
1 prompt = """You are a smart research assistant. Use the search engine to look up information.
2 You are allowed to make multiple calls (either together or in sequence).
3 Only look up information when you are sure of what you want.
4 If you need to look up some information before asking a follow-up question, you are allowed to do that!
5 """
6 model = ChatGroq(temperature=0, model="Llama-3.3-70b-Specdec")
----> 8 with SqliteSaver.from_conn_string(f"sqlite:///{db_path}") as memory:
9 bot = Agent(model, [tool], system=prompt, checkpointer=memory)
11 messages = [HumanMessage(content="What is the weather in Texas?")]

File c:\Users\vinee\anaconda3\envs\mktest\Lib\contextlib.py:137, in _GeneratorContextManager.enter(self)
135 del self.args, self.kwds, self.func
136 try:
--> 137 return next(self.gen)
138 except StopIteration:
139 raise RuntimeError("generator didn't yield") from None

File c:\Users\vinee\anaconda3\envs\mktest\Lib\site-packages\langgraph\checkpoint\sqlite_init_.py:107, in SqliteSaver.from_conn_string(cls, conn_string)
83 @classmethod
84 @contextmanager
85 def from_conn_string(cls, conn_string: str) -> Iterator["SqliteSaver"]:
86 """Create a new SqliteSaver instance from a connection string.
...
111 )
112 ) as conn:
113 yield cls(conn)

OperationalError: unable to open database file

I have tried creating the file, giving absolute path etc. solutions. But it still doesn't resolve.

[koNoJOJOga]

In Lesson6:Essay Writer of AI Agents in LangGraph also getting similar error.

while executing the below code:

thread = {"configurable": {"thread_id": "1"}} for s in graph.stream({ 'task': "what is the difference between langchain and langsmith", "max_revisions": 2, "revision_number": 1, }, thread): print(s)

 I am getting AttributeError: '_GeneratorContextManager' object has no attribute 'get_next_version'.

The error is as follows:

 ---------------------------------------------------------------------------

AttributeError Traceback (most recent call last) Cell In[52], line 2 1 thread = {"configurable": {"thread_id": "1"}} ----> 2 for s in graph.stream({ 3 'task': "what is the difference between langchain and langsmith", 4 "max_revisions": 2, 5 "revision_number": 1, 6 }, thread): 7 print(s)

File /opt/conda/lib/python3.11/site-packages/langgraph/pregel/init.py:1284, in Pregel.stream(self, input, config, stream_mode, output_keys, interrupt_before, interrupt_after, debug, subgraphs) 1280 if "custom" in stream_modes: 1281 config[CONF][CONFIG_KEY_STREAM_WRITER] = lambda c: stream.put( 1282 ((), "custom", c) 1283 ) -> 1284 with SyncPregelLoop( 1285 input, 1286 stream=StreamProtocol(stream.put, stream_modes), 1287 config=config, 1288 store=store, 1289 checkpointer=checkpointer, 1290 nodes=self.nodes, 1291 specs=self.channels, 1292 output_keys=output_keys, 1293 stream_keys=self.stream_channels_asis, 1294 debug=debug, 1295 ) as loop: 1296 # create runner 1297 runner = PregelRunner( 1298 submit=loop.submit, 1299 put_writes=loop.put_writes, 1300 node_finished=config[CONF].get(CONFIG_KEY_NODE_FINISHED), 1301 ) 1302 # enable subgraph streaming

File /opt/conda/lib/python3.11/site-packages/langgraph/pregel/loop.py:654, in SyncPregelLoop.init(self, input, stream, config, store, checkpointer, nodes, specs, output_keys, stream_keys, check_subgraphs, debug) 652 self.stack = ExitStack() 653 if checkpointer: --> 654 self.checkpointer_get_next_version = checkpointer.get_next_version 655 self.checkpointer_put_writes = checkpointer.put_writes 656 else:

AttributeError: '_GeneratorContextManager' object has no attribute 'get_next_version'

Can you suggest any solution.

you can take all the initialization into 'with' like what I've done, it could be working.

prompt = """You are a smart research assistant. Use the search engine to look up information.  \
You are allowed to make multiple calls (either together or in sequence). \
Only look up information when you are sure of what you want. \
If you need to look up some information before asking a follow up question, you are allowed to do that!
"""
model = ChatOllama(
    model="qwen2.5:7b",
    temperature=0
)
with SqliteSaver.from_conn_string(":memory:") as checkpointer:
    abot = Agent(model, [tool], system=prompt, checkpointer=checkpointer)
    query = "What is the weather in New York?"
    messages = [HumanMessage(content=query)]
    thread = {"configurable": {"thread_id": "1"}}  
    for event in abot.graph.stream({"messages": messages}, thread):
        for v in event.values():
            print(v['messages'])

[koNoJOJOga]

prompt = """You are a smart research assistant. Use the search engine to look up information. You are allowed to make multiple calls (either together or in sequence). Only look up information when you are sure of what you want. If you need to look up some information before asking a follow-up question, you are allowed to do that! """ model = ChatGroq(temperature=0, model="Llama-3.3-70b-Specdec")

with SqliteSaver.from_conn_string(f"sqlite:///{db_path}") as memory: bot = Agent(model, [tool], system=prompt, checkpointer=memory)

messages = [HumanMessage(content="What is the weather in Texas?")]
thread = {"configurable": {"thread_id": "1"}}

for event in bot.graph.stream({"messages": messages}, thread):
    for v in event.values():
        print(v['messages'])

trying this, but getting the following error:

OperationalError Traceback (most recent call last) Cell In[6], line 8 1 prompt = """You are a smart research assistant. Use the search engine to look up information. 2 You are allowed to make multiple calls (either together or in sequence). 3 Only look up information when you are sure of what you want. 4 If you need to look up some information before asking a follow-up question, you are allowed to do that! 5 """ 6 model = ChatGroq(temperature=0, model="Llama-3.3-70b-Specdec") ----> 8 with SqliteSaver.from_conn_string(f"sqlite:///{db_path}") as memory: 9 bot = Agent(model, [tool], system=prompt, checkpointer=memory) 11 messages = [HumanMessage(content="What is the weather in Texas?")]

File c:\Users\vinee\anaconda3\envs\mktest\Lib\contextlib.py:137, in _GeneratorContextManager.enter(self) 135 del self.args, self.kwds, self.func 136 try: --> 137 return next(self.gen) 138 except StopIteration: 139 raise RuntimeError("generator didn't yield") from None

File c:\Users\vinee\anaconda3\envs\mktest\Lib\site-packages\langgraph\checkpoint\sqlite__init__.py:107, in SqliteSaver.from_conn_string(cls, conn_string) 83 @classmethod 84 @contextmanager 85 def from_conn_string(cls, conn_string: str) -> Iterator["SqliteSaver"]: 86 """Create a new SqliteSaver instance from a connection string. ... 111 ) 112 ) as conn: 113 yield cls(conn)

OperationalError: unable to open database file

I have tried creating the file, giving absolute path etc. solutions. But it still doesn't resolve.

you can take all the initialization into 'with' like what I've done, it could be working.

prompt = """You are a smart research assistant. Use the search engine to look up information.  \
You are allowed to make multiple calls (either together or in sequence). \
Only look up information when you are sure of what you want. \
If you need to look up some information before asking a follow up question, you are allowed to do that!
"""
model = ChatOllama(
    model="qwen2.5:7b",
    temperature=0
)
with SqliteSaver.from_conn_string(":memory:") as checkpointer:
    abot = Agent(model, [tool], system=prompt, checkpointer=checkpointer)
    query = "What is the weather in New York?"
    messages = [HumanMessage(content=query)]
    thread = {"configurable": {"thread_id": "1"}}  
    for event in abot.graph.stream({"messages": messages}, thread):
        for v in event.values():
            print(v['messages'])

[jasdeep06]

@vbarda Is this the desired behaviour? Everytime I will need to compile the graph in order to invoke it?

[githubramu]

from langgraph.checkpoint.memory import MemorySaver
memory = MemorySaver()

This fixed the issue.

[vbarda]

This is not a fix to the above issue though, the correct solution for sqlite checkpointer is using it as a context manager as in the selected answer

[Saifullah3711]

This really helped in memory

[trex55]

I'm getting the below error still

---

AttributeError Traceback (most recent call last)
in <cell line: 3>()
1 messages = [HumanMessage(content="Whats the weather in SF?")]
2 thread = {"configurable": {"thread_id": "1"}}
----> 3 for event in abot.graph.stream({"messages": messages}, thread):
4 for v in event.values():
5 print(v)

1 frames
/usr/local/lib/python3.10/dist-packages/langgraph/pregel/loop.py in init(self, input, stream, config, store, checkpointer, nodes, specs, manager, interrupt_after, interrupt_before, output_keys, stream_keys, check_subgraphs, debug)
780 self.stack = ExitStack()
781 if checkpointer:
--> 782 self.checkpointer_get_next_version = checkpointer.get_next_version
783 self.checkpointer_put_writes = checkpointer.put_writes
784 else:

AttributeError: '_GeneratorContextManager' object has no attribute 'get_next_version'

Can you point me in the right direction of what version the different python packages have to be so I can run the code that is in the course please

[JoshChima]

@smanna123, @vbarda @trex55

If anyone is running into this problem while doing the Deeplearning.AI course. I found this implementation of the SqliteSaver to work with the following package versions:

• langgraph==0.2.53
• langgraph-checkpoint==2.0.6
• langgraph-sdk==0.1.36

I have not figured out the proper way to close the SQLite connection after running the streaming. I just close it at the bottom of my notebook.
Note: It is important that you set check_same_thread in the sqlite3 connection to False or you will run into sqlites threading limitations. However, be cautious with this approach as it can lead to other concurrency issues if not managed properly.

I would still recommend @vbarda's answer as that is the proper way to use it.

sqlite_conn = sqlite3.connect("checkpoints.sqlite", check_same_thread=False)
memory = SqliteSaver(sqlite_conn)

class SimpleMemoryAgent(SimpleAgent):

    def __init__(self, model:ChatOpenAI, tools:list, checkpointer:BaseCheckpointSaver, system=""):
        """
        An example of a simple agent with persistent memory and tools provided to it.
        
        Args:
            model (ChatOpenAI): The model to use for generating responses.
            tools (list): A list of tools to use for processing tool calls.
            checkpointer (BaseCheckpointSaver): The checkpointer to use for saving and loading the state of the agent.
            system (str, optional): The system to use for generating the first response. Defaults to "".
        """
        # save the initial system message
        self.system = system

        graph = StateGraph(SimpleAgentState)
        graph.add_node("llm", self.call_openai) # add the agent (LLM) node
        graph.add_node("action", self.take_action) # add the action node

        # add the conditional edge to check if there are any tool calls.
        # If there are, go to the action node, otherwise end.
        graph.add_conditional_edges(
            "llm",
            self.exists_action,
            {True: "action", False: END}
        )
        graph.add_edge("action", "llm") # add the edge to go back to the agent (LLM) node after the action node
        graph.set_entry_point("llm") # set the entry point to the agent (LLM) node
        self.graph = graph.compile(checkpointer=checkpointer) # !!! compile the graph but this time pass in the checkpointer !!!
        self.tools = {t.name: t for t in tools} # save the tools. Map the tool name to the tool object.
        self.model = model.bind_tools(tools) # bind the tools to the model. Lets the model know which tools to use.

[sfc-gh-smisra]

Adding the correct answer and sequence of cells:

......

class AgentState(TypedDict):
    messages: Annotated[list[AnyMessage], operator.add]

Delete this cell entirely:

from langgraph.checkpoint.sqlite import SqliteSaver

memory = SqliteSaver.from_conn_string(":memory:")

The Agent class implementation will be as it is.

Next cell:

prompt = """You are a smart research assistant. Use the search engine to look up information. \
You are allowed to make multiple calls (either together or in sequence). \
Only look up information when you are sure of what you want. \
If you need to look up some information before asking a follow up question, you are allowed to do that!
"""
model = ChatOpenAI(model="")

next cell of messages, thread as is.

Then add this cell:

from langgraph.checkpoint.sqlite import SqliteSaver
with SqliteSaver.from_conn_string(":memory:") as checkpointer:
    abot = Agent(model, [tool], system=prompt, checkpointer=checkpointer)
    for event in abot.graph.stream({"messages": messages}, thread):
        for v in event.values():
            print(v['messages'])

[anandbhaskaran]

How can I ask another message to the same thread at a later point in time?

[BrutalCaeser]

It say Agent object is not defined

[nafshar]

In reference to the SqliteSaver.from_conn_string() being a context manager vs. a checkpointer, there is another way to open this database and keep it open across multiple cells, when working in notebooks, until you finally close it in a downstream cell as follows:

from contextlib import ExitStack

# Use contextlib to manually enter the context manager and keep the object
stack = ExitStack()

# Create the memory context manager and enter the context
memory= stack.enter_context(SqliteSaver.from_conn_string(":memory:"))

# Now the checkpointer can be reused across cells
abot = Agent(model, [tool], system=prompt, checkpointer=memory)

thread = {"configurable": {"thread_id": "1"}}

for event in abot.graph.stream({"messages": messages}, thread):
    for v in event.values():
        print(v['messages'])

In later cells you can close the context by using stack.close() in a cell.

[vbarda]

you can simply do

import sqlite3
conn = sqlite3.connect(...)
checkpointer = SqliteSaver(conn)
...
conn.close()

[nafshar]

This looks even easier !!