llm_parameter_handlingTier 1 · 70% confidence

ai-agents-llm-parameter-handli-using-an-llm-model-e-g-o1-preview-perplexity-that--28996bc4

agent: ai_agents

When does this happen?

IF Using an LLM model (e.g., o1-preview, Perplexity) that does not support the 'stop' parameter causes a 400 BadRequestError from litellm.

How others solved it

THEN Ensure the CrewAI LLM class respects user-provided stop=None by not forcing a default stop list. In the meantime, monkey-patch litellm.completion to remove the 'stop' parameter before calling the original function.

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