llm_stop_parameterTier 1 · 70% confidence

ai-agents-llm-stop-parameter-when-using-a-model-that-does-not-support-the-stop--bde2ceee

agent: ai_agents

When does this happen?

IF When using a model that does not support the 'stop' parameter (e.g., o1-preview on Azure, Perplexity) with crewAI, you get a BadRequestError: "Unsupported parameter: 'stop' is not supported with this model."

How others solved it

THEN Monkey-patch litellm.completion to remove the 'stop' key from kwargs before calling the original function. This forces the parameter to be omitted regardless of what crewAI passes. Alternatively, modify crewAI's llm.py to not force a default stop parameter (e.g., change the default to None).

import litellm

original_completion = litellm.completion

def patched_completion(*args, **kwargs):
    if 'stop' in kwargs:
        kwargs.pop('stop')
    return original_completion(*args, **kwargs)

litellm.completion = patched_completion

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics