langgraph/tutorials/chatbots/information-gather-prompting/

[giscus[bot]]

langgraph/tutorials/chatbots/information-gather-prompting/

Build language agents as graphs

https://langchain-ai.github.io/langgraph/tutorials/chatbots/information-gather-prompting/

[alonsoir]

Input User (q/Q to quit): the schema to extract, and the text to extract it from
Traceback (most recent call last):
File "/Users/aironman/git/reliable-rag-using-langGraph/chat_bot_info_gathering.py", line 96, in
for output in graph.stream([HumanMessage(content=user)], config=config):
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langgraph/pregel/init.py", line 834, in stream
_panic_or_proceed(done, inflight, step)
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langgraph/pregel/init.py", line 1334, in _panic_or_proceed
raise exc
File "/usr/local/Cellar/[email protected]/3.11.9/Frameworks/Python.framework/Versions/3.11/lib/python3.11/concurrent/futures/thread.py", line 58, in run
result = self.fn(*self.args, **self.kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langgraph/pregel/retry.py", line 66, in run_with_retry
task.proc.invoke(task.input, task.config)
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langchain_core/runnables/base.py", line 2368, in invoke
input = step.invoke(
^^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langchain_core/runnables/base.py", line 4396, in invoke
return self.bound.invoke(
^^^^^^^^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langchain_core/language_models/chat_models.py", line 170, in invoke
self.generate_prompt(
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langchain_core/language_models/chat_models.py", line 599, in generate_prompt
return self.generate(prompt_messages, stop=stop, callbacks=callbacks, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langchain_core/language_models/chat_models.py", line 456, in generate
raise e
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langchain_core/language_models/chat_models.py", line 446, in generate
self._generate_with_cache(
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langchain_core/language_models/chat_models.py", line 671, in _generate_with_cache
result = self._generate(
^^^^^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/langchain_openai/chat_models/base.py", line 522, in _generate
response = self.client.create(messages=message_dicts, **params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/openai/_utils/_utils.py", line 277, in wrapper
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/openai/resources/chat/completions.py", line 590, in create
return self._post(
^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/openai/_base_client.py", line 1240, in post
return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/openai/_base_client.py", line 921, in request
return self._request(
^^^^^^^^^^^^^^
File "/Users/aironman/Library/Caches/pypoetry/virtualenvs/reliable-rag-using-langgraph-Jr5kRkNJ-py3.11/lib/python3.11/site-packages/openai/_base_client.py", line 1020, in _request
raise self._make_status_error_from_response(err.response) from None
openai.BadRequestError: Error code: 400 - {'error': {'message': "An assistant message with 'tool_calls' must be followed by tool messages responding to each 'tool_call_id'. The following tool_call_ids did not have response messages: call_ImYUXkktr09rjRrEYsZTkhTe", 'type': 'invalid_request_error', 'param': 'messages.[9].role', 'code': None}}

Process finished with exit code 1

[CARYCHEN04]

try "llm = ChatOpenAI(model="gpt-4",temperature=0)"，it appears work fine.

[hinthornw]

@alonsoir if you continue looping after it's completed then you'll have to add a ToolMessage to the history!

[Klafi11]

@hinthornw do you know if it's possible to kind of reset the llm after it finished the tool_call, that you can query again. Im using the Information gathering chat bot to get User Information for a Project they want to participate. it can be the case that the User may want to requery another project. Currently im using streamlit as application and i need to refresh the page to query again.

[Klafi11]

solved it. Thanks anyway :)

[alonsoir]

Thanks for the updates, you are great people!!

I would like to know your opinion about the use of graphs together with vector databases so that LLMs do not stop scanning certain areas of knowledge and thus stop hallucinating, at least, that is what I understand about this phenomenon , since the LLM does not find the relevant areas in its scan of embedded areas, it has a tendency to be more creative. In theory, if when translating our prompt you are able to go through all the graphs that contain said information, and said information is of quality, the vector database will be able to select the information stored in those nodes using similar cosine. , correct?

I would like to know your expert opinion, because I see above all that regardless of whether this technique is relevant or not with this technology, what has always given me problems is how difficult it is to save and maintain updated information in a graph, especially with neo4j and that peculiar language. I suppose it must be very effective so that they don't deprecate it, because to me it seems horrible and worthy of the hottest flames of one of Dante's infernos.
Ideally, using this technology cries out for an NLP process capable of reading all the information in the text, detecting what a node is, what information about that node is, and what the relationships are with other nodes, so that when you pass it other text, By doing the same thing and finding nodes, information and relationships, come to some agreement on maintaining validity between different versions.

[Nagahemachand]

Is this more like proactive retrieval instead of reactive retreival?

[yuting1214]

Is this article deprecated?
There should be using StateGraph instead of MessageGraph?

[HDO504]

User (q/Q to quit): hello
Traceback (most recent call last):
File "D:\research\LLM_FULL_STACK\LLM_AID\LLM_fullstack\multi_agents.py", line 134, in
for output in graph.stream(
File "D:\anaconda\Lib\site-packages\langgraph\pregel_init_.py", line 997, in stream
panic_or_proceed(done, inflight, loop.step)
File "D:\anaconda\Lib\site-packages\langgraph\pregel_init.py", line 1398, in _panic_or_proceed
raise exc
File "D:\anaconda\Lib\site-packages\langgraph\pregel\executor.py", line 60, in done
task.result()
File "D:\anaconda\Lib\concurrent\futures_base.py", line 449, in result
return self.__get_result()
^^^^^^^^^^^^^^^^^^^
File "D:\anaconda\Lib\concurrent\futures_base.py", line 401, in __get_result
raise self._exception
File "D:\anaconda\Lib\concurrent\futures\thread.py", line 58, in run
result = self.fn(*self.args, **self.kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\anaconda\Lib\site-packages\langgraph\pregel\retry.py", line 25, in run_with_retry
task.proc.invoke(task.input, task.config)
File "D:\anaconda\Lib\site-packages\langgraph\utils.py", line 93, in invoke
ret = self._call_with_config(
^^^^^^^^^^^^^^^^^^^^^^^
File "D:\anaconda\Lib\site-packages\langchain_core\runnables\base.py", line 1785, in _call_with_config
context.run(
File "D:\anaconda\Lib\site-packages\langchain_core\runnables\config.py", line 427, in call_func_with_variable_args
return func(input, **kwargs) # type: ignore[call-arg]
^^^^^^^^^^^^^^^^^^^^^
File "D:\anaconda\Lib\site-packages\langgraph\pregel\write.py", line 97, in _write
values = [
^
File "D:\anaconda\Lib\site-packages\langgraph\pregel\write.py", line 98, in
val if write.mapper is None else write.mapper.invoke(val, config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\anaconda\Lib\site-packages\langgraph\utils.py", line 102, in invoke
ret = context.run(self.func, input, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\anaconda\Lib\site-packages\langgraph\graph\state.py", line 527, in _get_state_key
raise InvalidUpdateError(f"Expected dict, got {input}")
langgraph.errors.InvalidUpdateError: Expected dict, got [HumanMessage(content='hello')]

I just paste this code to my pycharm, but it didn't work

[yuxiaojian]

looks like this article is deprecated. it doesn't work. the "graph.stream" should have "State" type input

[haimh]

I've faced the same issue here!

So, I've tried to put TypeDict into graph as bellow:

user_question = messages[-1].content
inputs = { "messages": [HumanMessage(content=user_question)] }
...
async for event in chatbot.astream_events(
inputs,
config,
stream_mode="updates",
version="v2"
):

and it throw out error:

ValueError: Message dict must contain 'role' and 'content' keys, got {'messages': [HumanMessage(content='What could you do?', id='950572
ce-d259-41fa-9c21-c70a3d88365c')]}

Then, I've tried with convert all mesages into dict with key "role" and "content", but it refuse to run with the same error.

I think it could be a bug with langchain libraries or need to update elsewhere.

Have you got any ideas or suggestion?

Thank in advance.
Richard.

[haimh]

I've just seen latest update from Langchain team. It worked like a champ! Thanks for great effort <3

[deepakjoshi-ekline]

@haimh could you please provide your code or code repo here ?

[haimh]

Here it is:

Link: https://github.com/langchain-ai/langgraph/blob/main/examples/chatbots/information-gather-prompting.ipynb

[deepakjoshi-ekline]

Thanks

[jellis18]

I'm somewhat new to langchain/langgraph. I'm a bit confused here as to why you have the add_tool_message tool. Could we not just have the graph with the info node and prompt node and the conditional edge to route between them.

I'm guessing this is a kind of hack to make things work in the framework of tools but its not clear and maybe some explanation in the sample would clear it up.

Thanks

[RafaelRViana]

Did anybody understand how variables, constraints and requirements inside PromptInstructions were filled?

[haimh]

They were filled through collaboration between Pydantic class interaction with model (bind.tool). For more detail, you should print out the tool_call in function get_prompt_messages0 for further info.

Hope that help,
Richard.

[cris-m]

there is issue with this example. I am getting the following exception:

/usr/local/lib/python3.10/dist-packages/anthropic/_base_client.py in _request(self, cast_to, options, retries_taken, stream, stream_cls)
   1056 
   1057             log.debug("Re-raising status error")
-> 1058             raise self._make_status_error_from_response(err.response) from None
   1059 
   1060         return self._process_response(

BadRequestError: Error code: 400 - {'type': 'error', 'error': {'type': 'invalid_request_error', 'message': 'messages: at least one message is required'}}

this is the tool called from get_prompt_messages:

{'name': 'PromptInstructions', 'args': {'objective': 1, 'variables': ['rag'], 'constraints': [], 'requirements': []}, 'id': 'toolu_011mh6syf4XjdP91GC8oaiv9', 'type': 'tool_call'}