[LangGraph + Ollama] Agent using local model (qwen2.5) returns AIMessage(content='') even when tool responds correctly

[matiasdev30]

[LangGraph + Ollama] Agent using local model (qwen2.5) returns AIMessage(content='') even when tool responds correctly

I’m using create_react_agent from langgraph.prebuilt with a local model served via Ollama (qwen2.5), and the agent consistently returns an AIMessage with an empty content field — even though the tool returns a valid string.

Code

from langgraph.prebuilt import create_react_agent
from langchain_ollama import ChatOllama

model = ChatOllama(model="qwen2.5")

def search(query: str):
    """Call to surf the web."""
    if "sf" in query.lower() or "san francisco" in query.lower():
        return "It's 60 degrees and foggy."
    return "It's 90 degrees and sunny."

agent = create_react_agent(model=model, tools=[search])

response = agent.invoke(
    {},
    {"messages": [{"role": "user", "content": "what is the weather in sf"}]}
)
print(response)

Output

{
  'messages': [
    AIMessage(
      content='', 
      additional_kwargs={}, 
      response_metadata={
        'model': 'qwen2.5',
        'created_at': '2025-04-24T09:13:29.983043Z',
        'done': True,
        'done_reason': 'load',
        'total_duration': None,
        'load_duration': None,
        'prompt_eval_count': None,
        'prompt_eval_duration': None,
        'eval_count': None,
        'eval_duration': None,
        'model_name': 'qwen2.5'
      },
      id='run-6a897b3a-1971-437b-8a98-95f06bef3f56-0'
    )
  ]
}

As shown above, the agent responds with an empty string, even though the search() tool clearly returns "It's 60 degrees and foggy.".

Has anyone seen this behavior? Could it be an issue with qwen2.5, langgraph.prebuilt, the Ollama config, or maybe a mismatch somewhere between them?

Any insight appreciated.
``

[dosubot[bot]]

The issue you're encountering, where the AIMessage has an empty content field despite the tool returning a valid response, could be related to how the AIMessage is structured in LangChain. The content field is often set to None when the message is used to represent a tool call or function call with specific arguments and function names, rather than conveying a textual message. This is indicated by the presence of additional_kwargs that specify the tool's function and arguments, such as function_call or tool_calls ^[1][2].

In your case, it seems that the tool's response is not being directly set as the content of the AIMessage. Instead, the response might be captured in a different part of the message structure, possibly within additional_kwargs or another metadata field. You might want to check if the tool's output is being stored elsewhere in the message object.

Additionally, ensure that the integration between the create_react_agent function and the ChatOllama model is correctly configured to handle tool outputs. There might be a need to adjust how the agent processes and formats the tool's response to ensure it appears in the content field of the AIMessage.

If the issue persists, consider reviewing the implementation details of the create_react_agent function and the ChatOllama class to ensure they are correctly handling and passing tool outputs.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}