Skip to content

feat: add enterprise gateway support for LLM providers #963

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

Closed
wants to merge 10 commits into from
Closed

feat: add enterprise gateway support for LLM providers #963

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
ad9752c
feat: add LLMWithGateway for enterprise OAuth support
ak684 Oct 30, 2025
4cfd5a4
feat: improve token TTL handling and add extended thinking support
ak684 Oct 30, 2025
28eb5a1
feat: pass custom_llm_provider to litellm calls
ak684 Oct 30, 2025
ec34c93
Add ssl_verify support for gateway LLM calls
ak684 Oct 31, 2025
1029d4f
more tls fixes
ak684 Oct 31, 2025
10fdfbc
simply PR
ak684 Oct 31, 2025
00e9ae8
fix pre-commit
ak684 Oct 31, 2025
c52bb3e
nit
ak684 Oct 31, 2025
4b9b99d
hardening gateway header/tls handling
ak684 Nov 1, 2025
aa373aa
remove rendering
ak684 Nov 3, 2025
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
2 changes: 2 additions & 0 deletions openhands-sdk/openhands/sdk/llm/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,7 @@
from openhands.sdk.llm.llm import LLM
from openhands.sdk.llm.llm_registry import LLMRegistry, RegistryEvent
from openhands.sdk.llm.llm_response import LLMResponse
from openhands.sdk.llm.llm_with_gateway import LLMWithGateway
from openhands.sdk.llm.message import (
ImageContent,
Message,
Expand All @@ -23,6 +24,7 @@
__all__ = [
"LLMResponse",
"LLM",
"LLMWithGateway",
"LLMRegistry",
"RouterLLM",
"RegistryEvent",
Expand Down
149 changes: 88 additions & 61 deletions openhands-sdk/openhands/sdk/llm/llm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -162,6 +162,14 @@ class LLM(BaseModel, RetryMixin, NonNativeToolCallingMixin):
)
ollama_base_url: str | None = Field(default=None)

ssl_verify: bool | str | None = Field(
default=None,
description=(
"TLS verification forwarded to LiteLLM; "
"set to False when corporate proxies break certificate chains."
),
)

drop_params: bool = Field(default=True)
modify_params: bool = Field(
default=True,
Expand Down Expand Up @@ -583,33 +591,35 @@ def responses(
def _one_attempt(**retry_kwargs) -> ResponsesAPIResponse:
final_kwargs = {**call_kwargs, **retry_kwargs}
with self._litellm_modify_params_ctx(self.modify_params):
with warnings.catch_warnings():
warnings.filterwarnings("ignore", category=DeprecationWarning)
typed_input: ResponseInputParam | str = (
cast(ResponseInputParam, input_items) if input_items else ""
)
ret = litellm_responses(
model=self.model,
input=typed_input,
instructions=instructions,
tools=resp_tools,
api_key=self.api_key.get_secret_value()
if self.api_key
else None,
api_base=self.base_url,
api_version=self.api_version,
timeout=self.timeout,
drop_params=self.drop_params,
seed=self.seed,
**final_kwargs,
)
assert isinstance(ret, ResponsesAPIResponse), (
f"Expected ResponsesAPIResponse, got {type(ret)}"
)
# telemetry (latency, cost). Token usage mapping we handle after.
assert self._telemetry is not None
self._telemetry.on_response(ret)
return ret
with self._litellm_ssl_verify_ctx():
with warnings.catch_warnings():
warnings.filterwarnings("ignore", category=DeprecationWarning)
typed_input: ResponseInputParam | str = (
cast(ResponseInputParam, input_items) if input_items else ""
)
ret = litellm_responses(
model=self.model,
input=typed_input,
instructions=instructions,
tools=resp_tools,
api_key=self.api_key.get_secret_value()
if self.api_key
else None,
api_base=self.base_url,
api_version=self.api_version,
custom_llm_provider=self.custom_llm_provider,
timeout=self.timeout,
drop_params=self.drop_params,
seed=self.seed,
**final_kwargs,
)
assert isinstance(ret, ResponsesAPIResponse), (
f"Expected ResponsesAPIResponse, got {type(ret)}"
)
# telemetry (latency, cost). Token usage mapping we handle after.
assert self._telemetry is not None
self._telemetry.on_response(ret)
return ret

try:
resp: ResponsesAPIResponse = _one_attempt()
Expand Down Expand Up @@ -642,40 +652,45 @@ def _transport_call(
) -> ModelResponse:
# litellm.modify_params is GLOBAL; guard it for thread-safety
with self._litellm_modify_params_ctx(self.modify_params):
with warnings.catch_warnings():
warnings.filterwarnings(
"ignore", category=DeprecationWarning, module="httpx.*"
)
warnings.filterwarnings(
"ignore",
message=r".*content=.*upload.*",
category=DeprecationWarning,
)
warnings.filterwarnings(
"ignore",
message=r"There is no current event loop",
category=DeprecationWarning,
)
warnings.filterwarnings(
"ignore",
category=UserWarning,
)
# Some providers need renames handled in _normalize_call_kwargs.
ret = litellm_completion(
model=self.model,
api_key=self.api_key.get_secret_value() if self.api_key else None,
base_url=self.base_url,
api_version=self.api_version,
timeout=self.timeout,
drop_params=self.drop_params,
seed=self.seed,
messages=messages,
**kwargs,
)
assert isinstance(ret, ModelResponse), (
f"Expected ModelResponse, got {type(ret)}"
)
return ret
with self._litellm_ssl_verify_ctx():
with warnings.catch_warnings():
warnings.filterwarnings(
"ignore", category=DeprecationWarning, module="httpx.*"
)
warnings.filterwarnings(
"ignore",
message=r".*content=.*upload.*",
category=DeprecationWarning,
)
warnings.filterwarnings(
"ignore",
message=r"There is no current event loop",
category=DeprecationWarning,
)
warnings.filterwarnings(
"ignore",
category=UserWarning,
)
# Some providers need renames handled in _normalize_call_kwargs.
ret = litellm_completion(
model=self.model,
api_key=self.api_key.get_secret_value()
if self.api_key
else None,
base_url=self.base_url,
api_version=self.api_version,
custom_llm_provider=self.custom_llm_provider,
timeout=self.timeout,
ssl_verify=self.ssl_verify,
drop_params=self.drop_params,
seed=self.seed,
messages=messages,
**kwargs,
)
assert isinstance(ret, ModelResponse), (
f"Expected ModelResponse, got {type(ret)}"
)
return ret

@contextmanager
def _litellm_modify_params_ctx(self, flag: bool):
Expand All @@ -686,6 +701,18 @@ def _litellm_modify_params_ctx(self, flag: bool):
finally:
litellm.modify_params = old

@contextmanager
def _litellm_ssl_verify_ctx(self):
if self.ssl_verify is None:
yield
return
old = getattr(litellm, "ssl_verify", None)
try:
litellm.ssl_verify = self.ssl_verify
yield
finally:
litellm.ssl_verify = old

# =========================================================================
# Capabilities, formatting, and info
# =========================================================================
Expand Down
Loading
Loading