agent_executor with history on Azure issues

[youdar]

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

# Test function - to show the issue
def eval_basic_lanchain_azure_agent():
    print(f'\n- - -  run: {eval_basic_lanchain_azure_agent.__name__}() - - - ')
    t0 = time()
    from langchain.agents import AgentExecutor, Tool, create_openai_functions_agent
    from langchain.agents.openai_functions_agent.base import OpenAIFunctionsAgent
    from langchain_community.chat_message_histories import ChatMessageHistory
    from langchain_core.runnables.history import RunnableWithMessageHistory
    from langchain.schema.messages import SystemMessage
    from langchain_openai import AzureChatOpenAI
    
    # When using ChatOpenAI instead of AzureChatOpenAI, there is not such issue.  
    from langchain_openai import ChatOpenAI
    from utils import get_current_time

    tools = [get_current_time]

    llm = AzureChatOpenAI(
        azure_deployment=AZURE_OPENAI_CHAT_DEPLOYMENT_NAME,
        openai_api_version=OPENAI_API_VERSION,
        azure_endpoint=OPENAI_API_BASE,
        temperature=0,
    )

    system_message = SystemMessage(content="You're a helpful assistant")
    prompt = OpenAIFunctionsAgent.create_prompt(system_message=system_message, )
    agent = create_openai_functions_agent(llm, tools, prompt)
    _agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=False, return_intermediate_steps=False)

    message_history = ChatMessageHistory()
    agent_executor = RunnableWithMessageHistory(
        _agent_executor,
        lambda session_id: message_history,
        input_messages_key="input",
        history_messages_key="chat_history",
    )

    query = 'in 10 words tell me about LangChain'
    result = agent_executor.invoke(
        {'input': query},
        config={"configurable": {"session_id": "test-session_id"}},
    )
    response = result['output']
    print("query:", query)
    print(response)
    print(f"Response time: {round(time() - t0)} sec")

Description

I am working on an agent_executor that worked with OpenAI tools and models and that has chat history.

When I was working directly with OpenAI credentials (from langchain_openai import ChatOpenAI) I had no issue.
Switching to OpenAI via Azure (from langchain_openai import AzureChatOpenAI), when interacting with the agent I get something like:
Parent run b134b735-...2 not found for run 9ce0ed27-... Treating as a root run.

While looking for solutions I found reference to solutions of the form

from azureml.core import Experiment
from azureml.core import Workspace
ws = Workspace.from_config()
experiment = Experiment(workspace=ws, name='MyExperiment')
run = experiment.start_logging()
# run the agent / queries
run.complete()

but it looks a bit overkill to start Azure workspace and experiments every time we run something.
I also tried adding session_id to ChatMessageHistory(). It did not help

Note that when calling the llm model without the agent (llm.invoke(...)), there are not such issues

Any thought on what am I missing?

System Info

Platform: Mac (Sonoma 14.3)

% pip freeze | grep lang
langchain==0.1.17
langchain-community==0.0.36
langchain-core==0.1.50
langchain-google-vertexai==1.0.3
langchain-openai==0.1.6
langchain-text-splitters==0.0.1
langchainhub==0.1.15
langsmith==0.1.53
openinference-instrumentation-langchain==0.1.14

% pip freeze | grep openai
langchain-openai==0.1.6
openai==1.25.1
openinference-instrumentation-openai==0.1.4

% python3 --version
Python 3.11.8

[dosubot[bot]]

Hello @youdar! 👋 I'm Dosu, a bot here to help you solve bugs, answer questions, and guide you on contributing while we wait for a human maintainer. I'm here to assist you! Let me take a look at your issue and I'll get back to you with a full answer shortly.

[cbull-81]

Having a similar issue using DynamoDBChatMessageHistory. Any input would be welcomed

[dosubot[bot]]

To address the "Parent run not found for run... Treating as a root run" error, it's crucial to manage run_id and thread_id properly when using AgentExecutor with AzureChatOpenAI. This error usually indicates a mismatch or absence of these identifiers, which are essential for linking execution instances and maintaining conversation threads within the LangChain framework. Ensure that every run or tool output submission is correctly associated with its parent thread by managing these identifiers accurately.

For the issue with the compas_fab.robots module not being found, this indicates that the compas_fab package might not be installed or is incorrectly installed in your environment. You should install it using pip install compas_fab and verify that it's accessible in your Python environment, especially if you're using a virtual environment.

Regarding the use of session_id with ChatMessageHistory(), ensure that its implementation aligns with the specific requirements of your application and the libraries in use. Incorrect usage might not directly cause the issues you're facing but could lead to other problems if not handled properly.

Make sure all dependencies, including openai, langchain, and any Azure-specific libraries, are correctly installed and up to date. Consider increasing the logging level to debug for more detailed error messages, which can be helpful in diagnosing and resolving the issue.

If you continue to face difficulties, reviewing the documentation of the services and libraries you're using for correct implementation of session_id and other configurations might provide further insights.

Details

Sources

• libs/partners/openai/langchain_openai/llms/azure.py
• libs/community/langchain_community/chat_models/azure_openai.py
• libs/community/langchain_community/chat_message_histories/sql.py
• libs/community/langchain_community/chat_message_histories/dynamodb.py
• libs/community/langchain_community/chat_message_histories/cassandra.py
• libs/partners/openai/langchain_openai/chat_models/azure.py
• libs/core/langchain_core/runnables/history.py
• libs/langchain/langchain/agents/openai_assistant/base.py
• libs/langchain/langchain/agents/agent.py
• libs/experimental/langchain_experimental/plan_and_execute/executors/agent_executor.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.

To continue the conversation, mention @dosu.

[youdar]

Did not help solving the issue.
It looks like it is running, but I wonder if there is some impact on performance and what is it