Skip to content

Support GPT5 and OpenAI new params #56

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
RandyZ wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Support GPT5 and OpenAI new params #56

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
44 changes: 42 additions & 2 deletions server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -554,6 +554,32 @@ def convert_anthropic_to_litellm(anthropic_request: MessagesRequest) -> Dict[str
"stream": anthropic_request.stream,
}

# For OpenAI Responses-style models (e.g., gpt-5, gpt-4.1, gpt-4o, o3/o4), use `max_completion_tokens`
try:
model_with_prefix = anthropic_request.model
clean_model_name = model_with_prefix.split("/", 1)[-1]
responses_prefixes = (
"gpt-5",
"gpt-4.1",
"gpt-4o",
"o3",
"o4",
"omni",
"chatgpt-4o",
"gpt-4.5",
)
if (model_with_prefix.startswith("openai/") or not any(model_with_prefix.startswith(p + "/") for p in ("anthropic", "gemini"))) \
and any(clean_model_name.startswith(p) for p in responses_prefixes):
# Move value from max_tokens -> max_completion_tokens to satisfy OpenAI Responses API
mt_value = litellm_request.pop("max_tokens", max_tokens)
litellm_request["max_completion_tokens"] = mt_value
logger.debug(
f"Using max_completion_tokens for OpenAI Responses model: {clean_model_name} (value={mt_value})"
)
except Exception as _e:
# Do not fail conversion if detection logic errors; fallback to existing behavior
pass

# Only include thinking field for Anthropic models
if anthropic_request.thinking and anthropic_request.model.startswith("anthropic/"):
litellm_request["thinking"] = anthropic_request.thinking
Expand Down Expand Up @@ -1116,7 +1142,21 @@ async def create_message(
litellm_request = convert_anthropic_to_litellm(request)

# Determine which API key to use based on the model
if request.model.startswith("openai/"):
# Also treat unprefixed OpenAI Responses models (e.g., gpt-5, gpt-4.1, gpt-4o, o3/o4) as OpenAI
responses_prefixes = (
"gpt-5",
"gpt-4.1",
"gpt-4o",
"o3",
"o4",
"omni",
"chatgpt-4o",
"gpt-4.5",
)
clean_model_name_for_key = request.model.split("/", 1)[-1]
is_openai_response_model = any(clean_model_name_for_key.startswith(p) for p in responses_prefixes)

if request.model.startswith("openai/") or (not request.model.startswith(("anthropic/", "gemini/")) and is_openai_response_model):
litellm_request["api_key"] = OPENAI_API_KEY
# Use custom OpenAI base URL if configured
if OPENAI_BASE_URL:
Expand All @@ -1132,7 +1172,7 @@ async def create_message(
logger.debug(f"Using Anthropic API key for model: {request.model}")

# For OpenAI models - modify request format to work with limitations
if "openai" in litellm_request["model"] and "messages" in litellm_request:
if (("openai" in litellm_request["model"]) or is_openai_response_model) and ("messages" in litellm_request):
logger.debug(f"Processing OpenAI model request: {litellm_request['model']}")

# For OpenAI models, we need to convert content blocks to simple strings
Expand Down