scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-29 01:42:19 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 100

only override if the global timeout sets timeout to 6000s

Browse source
This commit is contained in:
harish876 2026-04-14 04:25:50 +00:00
parent 0fa9cd1426
commit 2c14d421dc
2 changed files with 27 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

15
litellm/litellm_core_utils/completion_timeout.py
View file

@ -26,9 +26,13 @@ class CompletionTimeout:
``kwargs["timeout"]``, ``kwargs["request_timeout"]``, then ``global_timeout``
(e.g. :attr:`litellm.request_timeout` from proxy ``litellm_settings``), else ``600``.
Coerce :class:`httpx.Timeout` when the provider does not support it; map ``6000``
(:data:`~litellm.constants.DEFAULT_REQUEST_TIMEOUT_SECONDS`) to ``600`` for completion.
Coerce :class:`httpx.Timeout` when the provider does not support it. If the value
came only from ``global_timeout`` (no model/kwargs timeout) and equals ``6000``
(:data:`~litellm.constants.DEFAULT_REQUEST_TIMEOUT_SECONDS`), use ``600`` for
completion so chat calls do not inherit the long package default; explicit
``model_timeout`` / kwargs values of ``6000`` are left unchanged.
"""
timeout_from_global_only = False
if model_timeout is not None:
resolved: Union[float, str, httpx.Timeout] = model_timeout
elif kwargs.get("timeout") is not None:
@ -37,6 +41,7 @@ class CompletionTimeout:
resolved = kwargs["request_timeout"]
else:
resolved = global_timeout if global_timeout is not None else 600
timeout_from_global_only = True
if isinstance(resolved, httpx.Timeout) and not supports_httpx_timeout(
custom_llm_provider
@ -48,8 +53,10 @@ class CompletionTimeout:
elif not isinstance(resolved, httpx.Timeout):
resolved = float(resolved) # type: ignore
if not isinstance(resolved, httpx.Timeout) and float(resolved) == float(
DEFAULT_REQUEST_TIMEOUT_SECONDS
if (
timeout_from_global_only
and not isinstance(resolved, httpx.Timeout)
and float(resolved) == float(DEFAULT_REQUEST_TIMEOUT_SECONDS)
):
resolved = 600.0

19
tests/test_litellm/test_completion_timeout_resolution.py
View file

@ -79,8 +79,8 @@ def test_global_timeout_package_default_coerced_to_600_for_completion():
)
def test_explicit_request_timeout_6000_normalized_to_completion_default():
"""6000 is the package sentinel; completion always uses 600 instead."""
def test_explicit_request_timeout_6000_preserved():
"""Explicit deployment/request timeout must not be truncated by the package sentinel."""
assert (
CompletionTimeout.resolve(
None,
@ -89,7 +89,20 @@ def test_explicit_request_timeout_6000_normalized_to_completion_default():
global_timeout=None,
supports_httpx_timeout=supports_httpx_timeout,
)
== 600.0
== 6000.0
)
def test_explicit_model_timeout_6000_preserved():
assert (
CompletionTimeout.resolve(
6000.0,
{"timeout": 1.0, "request_timeout": 2.0},
"openai",
global_timeout=None,
supports_httpx_timeout=supports_httpx_timeout,
)
== 6000.0
)