[BUG] Groq API incompatible with smolagents system messages when using OpenAI endpoint via LiteLLMModel

[RonanKMcGovern]

Describe the bug
A clear and concise description of what the bug is.

Code to reproduce the error

from smolagents import CodeAgent, LiteLLMModel
import os

# Initialize the model with Groq
model = LiteLLMModel(
    "openai/deepseek-r1-distill-llama-70b",
    api_base="https://api.groq.com/openai/v1",
    api_key=os.getenv("GROQ_API_KEY")
)

# Create a minimal agent
agent = CodeAgent(
    tools=[],  # No tools needed to demonstrate the issue
    model=model,
    add_base_tools=False,
    verbosity_level=2
)

# Try to run a simple task - this will fail due to system message issues
try:
    result = agent.run("Say hello!")
    print(result)
except Exception as e:
    print(f"Error: {str(e)}") 

Error logs (if any)

uv run test-scripts/groq_system_message_error.py
/Users/ronanmcgovern/TR/ADVANCED-inference/agentic-rag/.venv/lib/python3.12/site-packages/pydantic/_internal/_config.py:345: UserWarning: Valid config keys have changed in V2:
* 'fields' has been removed
  warnings.warn(message, UserWarning)
╭────────────────────────────────────────────────── New run ──────────────────────────────────────────────────╮
│                                                                                                             │
│ Say hello!                                                                                                  │
│                                                                                                             │
╰─ LiteLLMModel - openai/deepseek-r1-distill-llama-70b ───────────────────────────────────────────────────────╯
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Step 0 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new
LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.

Error in generating model output:
litellm.BadRequestError: OpenAIException - Error code: 400 - {'error': {'message': "'messages.0' : for 
'role:system' the following must be satisfied[('messages.0.content' : value must be a string)]", 'type': 
'invalid_request_error'}}
[Step 0: Duration 0.32 seconds]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Step 1 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new
LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.

Error in generating model output:
litellm.BadRequestError: OpenAIException - Error code: 400 - {'error': {'message': "'messages.0' : for 
'role:system' the following must be satisfied[('messages.0.content' : value must be a string)]", 'type': 
'invalid_request_error'}}
[Step 1: Duration 0.23 seconds]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Step 2 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new
LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.

Error in generating model output:
litellm.BadRequestError: OpenAIException - Error code: 400 - {'error': {'message': "'messages.0' : for 
'role:system' the following must be satisfied[('messages.0.content' : value must be a string)]", 'type': 
'invalid_request_error'}}
[Step 2: Duration 0.62 seconds]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Step 3 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new
LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.

Error in generating model output:
litellm.BadRequestError: OpenAIException - Error code: 400 - {'error': {'message': "'messages.0' : for 
'role:system' the following must be satisfied[('messages.0.content' : value must be a string)]", 'type': 
'invalid_request_error'}}
[Step 3: Duration 0.24 seconds]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Step 4 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new
LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.

Error in generating model output:
litellm.BadRequestError: OpenAIException - Error code: 400 - {'error': {'message': "'messages.0' : for 
'role:system' the following must be satisfied[('messages.0.content' : value must be a string)]", 'type': 
'invalid_request_error'}}
[Step 4: Duration 0.78 seconds]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Step 5 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new
LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.

Error in generating model output:
litellm.BadRequestError: OpenAIException - Error code: 400 - {'error': {'message': "'messages.0' : for 
'role:system' the following must be satisfied[('messages.0.content' : value must be a string)]", 'type': 
'invalid_request_error'}}
[Step 5: Duration 0.25 seconds]
Reached max steps.

Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new
LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.

Final answer: Error in generating final LLM output:
litellm.BadRequestError: OpenAIException - Error code: 400 - {'error': {'message': "'messages.0' : for 
'role:system' the following must be satisfied[('messages.0.content' : value must be a string)]", 'type': 
'invalid_request_error'}}
[Step 6: Duration 0.25 seconds]
Error in generating final LLM output:
litellm.BadRequestError: OpenAIException - Error code: 400 - {'error': {'message': "'messages.0' : for 'role:system' the following must be satisfied[('messages.0.content' : value must be a string)]", 'type': 'invalid_request_error'}}

Expected behavior
Should run as any other litellm model.

Packages version:
Run pip freeze | grep smolagents and paste it here:

smolagents==1.6.0

Additional context
I ran the same tests on groq using only litellm (i.e. not as an agent) and there is no issue handling system prompts.

[elsolo5000-2]

Try this https://github.com/elsolo5000-2/smolagents/blob/main/src/smolagents/models.py
And use flatten_message_as_string=true (or something similar. See in the file) while creating the model

[seanphan]

We have the same issue

[MichaelisTrofficus]

Same here!

[vqndev]

Try this https://github.com/elsolo5000-2/smolagents/blob/main/src/smolagents/models.py And use flatten_message_as_string=true (or something similar. See in the file) while creating the model

I tried this and it didn't work