scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-05 02:41:56 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 49

fix(chatgpt): select live transport from model metadata

Browse source
This commit is contained in:
jibanez-staticduo 2026-09-09 11:17:09 +02:00
parent 93285edba1
commit 3686f6a005
No known key found for this signature in database
4 changed files with 54 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
litellm/llms/chatgpt/realtime.py
View file

@ -9,6 +9,7 @@ from litellm.llms.openai.realtime.handler import OpenAIRealtime
from litellm.llms.openai.realtime.http_transformation import OpenAIRealtimeHTTPConfig
from litellm.types.realtime import RealtimeQueryParams
from litellm.types.router import GenericLiteLLMParams
from litellm.utils import get_model_info
from .common_utils import CHATGPT_API_BASE
from .responses.transformation import ChatGPTResponsesAPIConfig
@ -34,6 +35,14 @@ def realtime_headers(
}
def realtime_endpoint(model: str) -> str:
try:
model_info: Final = get_model_info(model, custom_llm_provider="chatgpt")
except Exception:
return "realtime"
return "live" if "/v1/live" in (model_info.get("supported_endpoints") or ()) else "realtime"
class ChatGPTRealtime(OpenAIRealtime):
def __init__(self, params: GenericLiteLLMParams, headers: Mapping[str, str]) -> None:
super().__init__()
@ -50,7 +59,7 @@ class ChatGPTRealtime(OpenAIRealtime):
def _construct_url(self, api_base: str, query_params: RealtimeQueryParams) -> str:
base: Final = URL(api_base)
endpoint: Final = "live" if query_params.get("model") == "gpt-live-1-codex" else "realtime"
endpoint: Final = realtime_endpoint(query_params.get("model", ""))
if self._call_id:
return str(
base.copy_with(

10
litellm/model_prices_and_context_window_backup.json
View file

@ -27832,6 +27832,16 @@
"max_tokens": 8191,
"mode": "embedding"
},
"chatgpt/gpt-live-1-codex": {
"litellm_provider": "chatgpt",
"mode": "realtime",
"supported_endpoints": [
"/v1/realtime/calls",
"/v1/live"
],
"supports_audio_input": true,
"supports_audio_output": true
},
"chatgpt/gpt-5.4": {
"litellm_provider": "chatgpt",
"max_input_tokens": 1050000,

10
model_prices_and_context_window.json
View file

@ -27832,6 +27832,16 @@
"max_tokens": 8191,
"mode": "embedding"
},
"chatgpt/gpt-live-1-codex": {
"litellm_provider": "chatgpt",
"mode": "realtime",
"supported_endpoints": [
"/v1/realtime/calls",
"/v1/live"
],
"supports_audio_input": true,
"supports_audio_output": true
},
"chatgpt/gpt-5.4": {
"litellm_provider": "chatgpt",
"max_input_tokens": 1050000,

25
tests/test_litellm/llms/chatgpt/test_realtime.py
View file

@ -40,7 +40,7 @@ async def test_chatgpt_call_keeps_oauth_and_frameless_session(chatgpt_tokens):
@pytest.mark.parametrize("model,endpoint", [("gpt-realtime-1.5", "realtime"), ("gpt-live-1-codex", "live")])
def test_realtime_uses_platform_endpoint_with_oauth_headers(model, endpoint, chatgpt_tokens):
def test_realtime_uses_platform_endpoint_with_oauth_headers(model, endpoint, chatgpt_tokens, local_model_cost_map):
handler = ChatGPTRealtime(
GenericLiteLLMParams(),
{
@ -55,3 +55,26 @@ def test_realtime_uses_platform_endpoint_with_oauth_headers(model, endpoint, cha
assert headers["Authorization"] == "Bearer test-token-default"
assert "authorization" not in headers
assert headers["openai-alpha"] == "quicksilver=v2"
@pytest.mark.parametrize("endpoint", ["live", "realtime"])
@pytest.mark.parametrize("call_id", [None, "rtc_metadata"])
def test_realtime_routes_new_models_using_registered_metadata(endpoint, call_id, chatgpt_tokens, local_model_cost_map):
model = "metadata-voice-model"
litellm.register_model({f"chatgpt/{model}": {
"litellm_provider": "chatgpt", "mode": "realtime", "supported_endpoints": [f"/v1/{endpoint}"]
}})
handler = ChatGPTRealtime(GenericLiteLLMParams(chatgpt_realtime_call_id=call_id), {})
expected = (
f"wss://api.openai.com/v1/{endpoint}?model={model}" if call_id is None
else f"wss://api.openai.com/v1/live/{call_id}" if endpoint == "live"
else f"wss://api.openai.com/v1/realtime?call_id={call_id}"
)
assert handler._construct_url("https://api.openai.com/v1", {"model": model}) == expected
def test_realtime_unknown_model_keeps_standard_endpoint(chatgpt_tokens, local_model_cost_map):
handler = ChatGPTRealtime(GenericLiteLLMParams(), {})
assert handler._construct_url("https://api.openai.com/v1", {"model": "unknown-voice-model"}) == (
"wss://api.openai.com/v1/realtime?model=unknown-voice-model"
)