Error at running crewai using Ollama LLM

CrewAI Community Support
1

File “C:\Users\natha\nine_crew.venv\Lib\site-packages\crewai\agent.py”, line 195, in post_init_setup
self.llm = LLM(**llm_params)
^^^^^^^^^^^^^^^^^
TypeError: LLM.init() got an unexpected keyword argument ‘API_BASE’
An error occurred while running the crew: Command ‘[‘uv’, ‘run’, ‘run_crew’]’ returned non-zero exit status 1.

After resolving the uvloop doesnt support windows error, next I’m stuck with this error when I try to run the project on Windows 11, VS, Ollama LLM,
pls help
thanks

3

I am stuck at the same point.

4

The error occurs because the LLM class in the crewai package expects a BASE_URL key for Ollama’s base URL, but the ENV_VARS in constants.py currently provides API_BASE instead. This mismatch is likely due to a recent update in the LLM class, which changed the expected key from API_BASE to BASE_URL.

Cause:

In the file crewai/cli/constants.py, line 49, the ollama configuration specifies an API_BASE key for the base URL. However, the LLM class being initialized in agent.py (line 195) expects BASE_URL instead. This causes a TypeError when LLM is instantiated with the incorrect keyword argument.

Temporary Workaround:

To fix this without modifying the constants.py file (which might be overridden after an update), you can dynamically override the key in your project by adding the following code in your crew.py file before the @CrewBase class definition:

from crewai.cli.constants import ENV_VARS

# Override the key name dynamically
for entry in ENV_VARS.get("ollama", []):
    if "API_BASE" in entry:
        entry["BASE_URL"] = entry.pop("API_BASE")  # Rename the key

Let me know if it works

5

I can confirm that this workaround does the trick and the crew initial demo script executes

2 Likes