Working with a Databricks GPT OSS 120B yield API provider's servers are down or overloaded when it is a LiteLLm/Pydantic error

[srggrs]

Issue

I connected a Databricks hosted LLM, GPT OSS 120B using the open ai connector. Seems working no worries using these lines,

import os
import httpx

from openai import OpenAI
from dotenv import load_dotenv

load_dotenv()
DATABRICKS_TOKEN = os.getenv("DATABRICKS_TOKEN")
if not DATABRICKS_TOKEN:
    raise ValueError("DATABRICKS_TOKEN environment variable is not set.")

http_client = httpx.Client(
    timeout=60.0,
    trust_env=False,
    follow_redirects=True,
)

client = OpenAI(
    api_key=DATABRICKS_TOKEN,
    base_url="< my databricks URL >",
    http_client=http_client
)

model: str = "databricks-gpt-oss-120b"
prompt: str = "who are u"

response = client.chat.completions.create(
    model=model,
    messages=[
        {"role": "user", "content": prompt}
    ],
    max_tokens=5000
)

print(response.choices[0].message.content)

so I gave it a go with aider but I get an issue with LiteLLM and pydantic, tho I can see the output from the LLM from the stacktrace.

> who are u

    ░litellm.InternalServerError: InternalServerError: OpenAIException - Invalid response object Traceback (most recent call last):
  File "C:\<path\to>\uv\tools\aider-chat\Lib\site-packages\litellm\litellm_core_utils\llm_response_utils\convert_dict_to_response.py", 
line 564, in convert_to_model_response_object
    message = Message(
              ^^^^^^^^
  File "C:\<path\to>\uv\tools\aider-chat\Lib\site-packages\litellm\types\utils.py", line 633, in __init__
    super(Message, self).__init__(
  File "C:\<path\to>\uv\tools\aider-chat\Lib\site-packages\pydantic\main.py", line 253, in __init__
    validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
pydantic_core._pydantic_core.ValidationError: 1 validation error for Message
content
  Input should be a valid string [type=string_type, input_value=[{'type': 'reasoning', 's...ll be happy to help.'}], input_type=list]
    For further information visit https://errors.pydantic.dev/2.11/v/string_type


received_args={'response_object': {'id': 'chatcmpl_< a random id >', 'choices': [{'finish_reason': 'stop', 'index': 0, 'logprobs': None,   
'message': {'content': [{'type': 'reasoning', 'summary': [{'type': 'summary_text', 'text': 'The user asks "who are u". They want a response. According to system
instructions, we must act as an expert software developer, respond in English_Australia. The user is not requesting code changes. So we should answer the       
question about who we are. We should not propose any file changes. Just answer.'}]}, {'type': 'text', 'text': 'I’m ChatGPT, an AI language model created by     
OpenAI. I’m here to help you with software development questions, code reviews, debugging, and any other programming‑related tasks you might have. If you need  
assistance with your codebase, just let me know which files you’d like me to look at, and I’ll be happy to help.'}], 'refusal': None, 'role': 'assistant',      
'annotations': None, 'audio': None, 'function_call': None, 'tool_calls': None}}], 'created': 11111111, 'model': 'gpt-oss-120b-080525', 'object':
'chat.completion', 'service_tier': None, 'system_fingerprint': None, 'usage': {'completion_tokens': 146, 'prompt_tokens': 1711, 'total_tokens': 1857,
'completion_tokens_details': None, 'prompt_tokens_details': None}}, 'model_response_object': ModelResponse(id='chatcmpl-< a random id>',  
created=111111111, model=None, object='chat.completion', system_fingerprint=None, choices=[Choices(finish_reason='stop', index=0, message=Message(content=None,
role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None))], usage=Usage(completion_tokens=0, prompt_tokens=0, total_tokens=0,      
completion_tokens_details=None, prompt_tokens_details=None)), 'response_type': 'completion', 'stream': False, 'start_time': None, 'end_time': None,
'hidden_params': None, '_response_headers': {'content-type': 'application/json; charset=utf-8', 'server': 'databricks', 'date': 'Thu, 13 Nov 2025 00:25:41 GMT',
'x-databricks-org-id': '< my org id >', 'strict-transport-security': 'max-age=31536000; includeSubDomains; preload, max-age=31536000; includeSubDomains;     
preload', 'x-content-type-options': 'nosniff, nosniff', 'x-request-id': '< a rnd id >', 'content-encoding': 'gzip', 'vary':
'Accept-Encoding', 'alt-svc': 'clear', 'server-timing': 'request_id;dur=0;desc="< a rnd id >", client_protocol;dur=0;desc="HTTP/1.1"',  
'transfer-encoding': 'chunked'}, 'convert_tool_call_to_json_mode': None}
The API provider's servers are down or overloaded.

Anyone can help me understand how to fix this? Just bump the version of litellm? Cheers! 🙏🏼

Version and model info

Tested on Aider 0.86.1 (installed by uv) and current main (see below)

Aider v0.86.2.dev+less
Main model: openai/databricks-gpt-oss-120b with whole edit format
Weak model: openai/databricks-gpt-oss-20b
Git repo: .git with 26 files
Repo-map: using 1024 tokens, auto refresh