LiteLLM: Crew stuck in Loop

CrewAI Community Support
,
1

When I kick off a crew, it gets stuck in a loop. Is there any parameter I can use to avoid the loop?
If the problem is my OpenAI API Key then this should definitely stop after some retries, but my crew does not.

llm = LLM(
    model='gpt-4o-mini',
    api_key=OPENAI_API_KEY
    )

Agent setting:

-  allow_delegation=True,
-  verbose=True,
-  max_iter=3,

Stuck in Loop

  • ERROR:root:LiteLLM call failed: litellm.AuthenticationError: AuthenticationError: OpenAIException - Error code: 401 - {‘error’: {‘message’: ‘Incorrect API key provided: sk-proj-*********************************************. You can find your API key at https://platform.openai.com/account/api-keys.’, ‘type’: ‘invalid_request_error’, ‘param’: None, ‘code’: ‘invalid_api_key’}}

  • LiteLLM.Info: If you need to debug this error, use `litellm.set_verbose=True’.

.
.
.

2 Likes
2

I faced on same issue.
I’m using ollama as below.

llm=LLM(model="ollama/llama3.1", base_url="http://localhost:11434")

	@crew
	def crew(self) -> Crew:
		return Crew(
			agents=self.agents,
			tasks=self.tasks,
			process=Process.sequential,
			verbose=True,
			model="ollama/llama3.1",
			chat_llm=llm
		)
3

I’m struck in similar situation. Instead of llm=LLM(…), I am using llma=Ollama(…) as calling LLM was giving an error. No error on execution but goes in loop even though I’ve given max_iterations and max_time. Any help will be appreciated as cannot run open source on local laptop

4

Which model are you using with Ollama? I’ve seen weaker models often go into loops and just perform badly. Test with gpt-4o to see if the problem persists

5

Well, it’s actually gpt-4o.

6

Same problem. Does anyone find the solution for this?