[BUG] ModuleNotFoundError: No module named 'litellm' After Upgrading to CrewAI 1.0.0

[mustafaelshobaky]

Description

Describe the bug
After upgrading to CrewAI 1.0.0, attempts to convert text to a Pydantic model fail with the following error:

Failed to convert text into a Pydantic model due to error: No module named 'litellm'

This occurs during the model conversion process inside crewai/utilities/converter.py, where the code tries to import litellm from within internal_instructor.py.

⚙️ Environment Details

CrewAI version: 1.0.0

Python version: 3.11

OS: (e.g., Ubuntu 22.04 / macOS / Windows WSL2)

Installation method: pip install crewai --upgrade

Steps to Reproduce

🧩 Steps to Reproduce

Upgrade CrewAI to version 1.0.0:

pip install -U crewai

Run any code that uses Crew().convert_to_model() or .to_pydantic().

Observe that it fails with ModuleNotFoundError: No module named 'litellm'.

Expected behavior

✅ Expected Behavior

The conversion to a Pydantic model should work without requiring manual installation of litellm, or the dependency should be automatically installed as part of the CrewAI package.

🚑 Actual Behavior

The call fails with ModuleNotFoundError, halting all conversions that rely on InternalInstructor.

Screenshots/Code snippets

🔍 Full Traceback
Traceback (most recent call last):
File "/usr/local/lib/python3.11/site-packages/crewai/utilities/converter.py", line 167, in convert_to_model
escaped_result = json.dumps(json.loads(result, strict=False))
...
json.decoder.JSONDecodeError: Unterminated string starting at: line 447 column 27 (char 24590)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/usr/local/lib/python3.11/site-packages/crewai/utilities/converter.py", line 56, in to_pydantic
result = self._create_instructor().to_pydantic()
File "/usr/local/lib/python3.11/site-packages/crewai/utilities/converter.py", line 137, in _create_instructor
return InternalInstructor(
File "/usr/local/lib/python3.11/site-packages/crewai/utilities/internal_instructor.py", line 65, in init
from litellm import completion
ModuleNotFoundError: No module named 'litellm'

Operating System

Windows 11

Python Version

3.11

crewAI Version

1.0.0

crewAI Tools Version

1.0.0

Virtual Environment

Venv

Evidence

Possible Solution

💡 Suggested Fix

Either:

Add litellm to the CrewAI install_requires in setup.py or pyproject.toml,
or

Handle the missing dependency gracefully with an internal fallback or optional import check.

Additional context

📎 Additional Context

This issue started immediately after updating to version 1.0.0, indicating it might be a packaging or dependency oversight in the new release.

[lorenzejay]

can you share how you are setting up llmand your agent?

LLM(...)

Agent(llm=...)

[mustafaelshobaky]

@crewbase
class DocumentsCrew:
agents_config = "config/agents.yaml"
tasks_config = "config/tasks.yaml"

def _get_llm_for_agent(self) -> LLM:
    return LLM(
        model="openai/gpt-4o",
        temperature=0.2,
        max_tokens=8000,
    )

@agent
def comprehensive_processor(self) -> Agent:
    return Agent(
        config=self.agents_config["comprehensive_processor"],
        tools=[...],
        llm=self._get_llm_for_agent(),
        verbose=False
    )

@task

...

@crew
def crew(self) -> Crew:
    return Crew(
        agents=self.agents,
        tasks=self.tasks,
        process=Process.sequential,
        verbose=False
    )

[jamieredpen]

We have had exactly the same issue, running on azure functions. Nothing has changed other than crew ai version.

[lorenzejay]

@jamieredpen for azure try adding this:

uv add "crewai[azure-ai-inference]"

usage:

llm = LLM(
    model="azure/gpt-4o-mini",
    api_key=os.getenv("AZURE_API_KEY"),
    endpoint=os.getenv("AZURE_API_BASE_URL"),
)

docs: https://docs.crewai.com/en/concepts/llms#azure

[lorenzejay]

@mustafaelshobaky we addressed your issue #3767 can you try upgrading to uv add crewai==1.1.0

[lorenzejay]

@mustafaelshobaky when you have a second can you confirm it working?

[mustafaelshobaky]

@lorenzejay It is working fine now. Thank you.