Agent executor with Ollama backend always terminates with 'done_reason': 'load' (empty content)

[FieteO]

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 👆

Run Ollama

ollama pull qwen2.5:7b
ollama serve

Example Code

from langchain_ollama import ChatOllama
from langchain_core.messages import HumanMessage
from langgraph.prebuilt import create_react_agent
from langchain_core.runnables.config import RunnableConfig

llm = ChatOllama(model='qwen2.5:7b',
                 temperature=0,
                 base_url='http://localhost:11434')
tools = []
agent = create_react_agent(
    model=llm,
    tools = tools
)
config: RunnableConfig = {
    "configurable": {
        "thread_id":"1"
    }
}

result = agent.invoke(HumanMessage('Hello'), config=config)
print(result["messages"])

Description

I am using create_react_agent to create an agent executor that is supposed to call tools.
In the given example, I am not defining any tools, but just want to get back a 'normal' text response from the agent.

When running the script, I get the following output (manually formatted for better readability):

$ python demo.py 
[
  AIMessage(
    content='',
    additional_kwargs={},
    response_metadata={
      'model': 'qwen2.5:7b',
      'created_at': '2025-05-21T09:37:56.153539017Z',
      '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': 'qwen2.5:7b'
    },
    id='run--b46ba0b2-e87b-494c-bb95-20db8db39139-0')
]

Based on the example output given in How to migrate from legacy LangChain agents to LangGraph, I would expect the content key to be populated with anything other than '' (an empty string).
In other places I have seen that for a correct response, the done_reason was 'stop', in my example it is 'load'.
Is this kind of an intermediate, streamed result? And if so, what can I do to wait for the full response?
What can I do to make this work?

System Info

Dependency versions:

langchain-core==0.3.60
langchain-ollama==0.3.3
langgraph==0.3.24

$ python --version
Python 3.10.12