What happened?

When calling LiteLLM API functions with missing or invalid parameters, error handling is inconsistent in certain cases:

1. Core Library (Direct Usage): When required parameters are completely omitted (not passed at all), TypeError is raised instead of BadRequestError
2. Proxy Server (Wrong Parameter Names): When using incorrect parameter names (e.g., messages instead of input for responses API), the proxy returns HTTP 500 Internal Server Error instead of HTTP 400 Bad Request

Note: When model parameter is simply missing (FastAPI provides model=None), the proxy correctly returns HTTP 400 with ProxyModelNotFoundError. This case works fine.

Expected Behavior

• Missing or invalid parameters should raise BadRequestError with HTTP 400 status code
• Error message should clearly indicate what went wrong
• No HTTP 500 errors for client-side parameter mistakes

Current Behavior

Case 1: Core Library - Completely omitted required parameter

import litellm

# Omitting 'model' parameter completely
try:
    litellm.completion(messages=[{"role": "user", "content": "test"}])
except TypeError as e:  # Should be BadRequestError
    print(f"TypeError: {e}")
    # Output: TypeError: completion() missing 1 required positional argument: 'model'

Case 2: Proxy - Wrong parameter name causes HTTP 500

# Using 'input' instead of 'messages' for chat/completions endpoint
curl -X POST http://localhost:4000/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-3.5-turbo",
    "input": [{"role": "user", "content": "test"}]
  }'

# Returns: HTTP 500 Internal Server Error
# Should return: HTTP 400 Bad Request

Relevant log output

Core Library:

TypeError: completion() missing 1 required positional argument: 'model'

Proxy Server (Wrong parameter name):

litellm.proxy.proxy_server._handle_llm_api_exception(): Exception occured - Router.acompletion() missing 1 required positional argument: 'messages'
TypeError: Router.acompletion() missing 1 required positional argument: 'messages'

Are you a ML Ops Team?

No

What LiteLLM version are you on?

v1.80.5

Impact

• API consumers cannot distinguish between client and server errors
• Violates HTTP status code conventions (4xx for client errors, 5xx for server errors)
• Makes debugging more difficult for users who provide wrong parameter names
• Core library raises TypeError instead of user-friendly BadRequestError

Solution

I have a fix ready that adds TypeError to BadRequestError conversion in:

1. @client decorator in litellm/utils.py (for core library)
2. route_request function in litellm/proxy/route_llm_request.py (for proxy layer)

This ensures parameter validation errors return HTTP 400 with clear error messages.