fix(bedrock): propagate timeout to streaming requests

litellm/llms/base_llm/chat/transformation.py

@@ -381,6 +381,7 @@ class BaseConfig(ABC):
         client: AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "CustomStreamWrapper":
         raise NotImplementedError
 
@@ -396,6 +397,7 @@ class BaseConfig(ABC):
         client: HTTPHandler | AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "CustomStreamWrapper":
         raise NotImplementedError
 

litellm/llms/bedrock/chat/agentcore/transformation.py

@@ -640,6 +640,7 @@ class AmazonAgentCoreConfig(BaseConfig, BaseAWSLLM):
         client: Union[HTTPHandler, "AsyncHTTPHandler"] | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "CustomStreamWrapper":
         """
         Simplified sync streaming - returns a generator that yields ModelResponse chunks.
@@ -853,6 +854,7 @@ class AmazonAgentCoreConfig(BaseConfig, BaseAWSLLM):
         client: Optional["AsyncHTTPHandler"] = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "CustomStreamWrapper":
         """
         Simplified async streaming - returns an async generator that yields ModelResponse chunks.

litellm/llms/bedrock/chat/converse_handler.py

@@ -35,6 +35,7 @@ def make_sync_call(
     json_mode: bool | None = False,
     fake_stream: bool = False,
     stream_chunk_size: int | None = None,
+    timeout: float | httpx.Timeout | None = None,
 ) -> tuple[Any, httpx.Headers]:
     if client is None:
         client = _get_httpx_client()  # Create a new client if none provided
@@ -45,6 +46,7 @@ def make_sync_call(
         data=data,
         stream=not fake_stream,
         logging_obj=logging_obj,
+        timeout=timeout,
     )
 
     if response.status_code != 200:
@@ -145,6 +147,7 @@ class BedrockConverseLLM(BaseAWSLLM):
             fake_stream=fake_stream,
             json_mode=json_mode,
             stream_chunk_size=stream_chunk_size,
+            timeout=timeout,
         )
         streaming_response: Final = CustomStreamWrapper(
             completion_stream=completion_stream,
@@ -555,6 +558,7 @@ class BedrockConverseLLM(BaseAWSLLM):
                 json_mode=json_mode,
                 fake_stream=fake_stream,
                 stream_chunk_size=stream_chunk_size,
+                timeout=timeout,
             )
             streaming_response: Final = CustomStreamWrapper(
                 completion_stream=completion_stream,

litellm/llms/bedrock/chat/invoke_handler.py

@@ -163,6 +163,7 @@ async def make_call(
     json_mode: bool | None = False,
     bedrock_invoke_provider: litellm.BEDROCK_INVOKE_PROVIDERS_LITERAL | None = None,
     stream_chunk_size: int | None = None,
+    timeout: float | httpx.Timeout | None = None,
 ) -> tuple[Any, httpx.Headers]:
     try:
         if client is None:
@@ -181,6 +182,7 @@ async def make_call(
             data=data,
             stream=not fake_stream,
             logging_obj=logging_obj,
+            timeout=timeout,
         )
 
         if response.status_code != 200:
@@ -248,6 +250,7 @@ def make_sync_call(
     json_mode: bool | None = False,
     bedrock_invoke_provider: litellm.BEDROCK_INVOKE_PROVIDERS_LITERAL | None = None,
     stream_chunk_size: int | None = None,
+    timeout: float | httpx.Timeout | None = None,
 ) -> tuple[Any, httpx.Headers]:
     try:
         if client is None:
@@ -265,6 +268,7 @@ def make_sync_call(
             data=signed_json_body if signed_json_body is not None else data,
             stream=not fake_stream,
             logging_obj=logging_obj,
+            timeout=timeout,
         )
 
         if response.status_code != 200:

litellm/llms/bedrock/chat/invoke_transformations/base_invoke_transformation.py

@@ -444,6 +444,7 @@ class AmazonInvokeConfig(BaseConfig, BaseAWSLLM):
         client: AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> CustomStreamWrapper:
         completion_stream, response_headers = await make_call(
             client=client,
@@ -456,6 +457,7 @@ class AmazonInvokeConfig(BaseConfig, BaseAWSLLM):
             fake_stream=True if "ai21" in api_base else False,
             bedrock_invoke_provider=self.get_bedrock_invoke_provider(model),
             json_mode=json_mode,
+            timeout=timeout,
         )
         streaming_response: Final = CustomStreamWrapper(
             completion_stream=completion_stream,
@@ -479,6 +481,7 @@ class AmazonInvokeConfig(BaseConfig, BaseAWSLLM):
         client: HTTPHandler | AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> CustomStreamWrapper:
         sync_client: Final = (
             _get_httpx_client(params={}) if client is None or isinstance(client, AsyncHTTPHandler) else client
@@ -495,6 +498,7 @@ class AmazonInvokeConfig(BaseConfig, BaseAWSLLM):
             fake_stream=True if "ai21" in api_base else False,
             bedrock_invoke_provider=self.get_bedrock_invoke_provider(model),
             json_mode=json_mode,
+            timeout=timeout,
         )
         streaming_response: Final = CustomStreamWrapper(
             completion_stream=completion_stream,

litellm/llms/bytez/chat/transformation.py

@@ -257,6 +257,7 @@ class BytezChatConfig(BaseConfig):
         client: HTTPHandler | AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "BytezCustomStreamWrapper":
         if client is None or isinstance(client, AsyncHTTPHandler):
             client = _get_httpx_client(params={})
@@ -299,6 +300,7 @@ class BytezChatConfig(BaseConfig):
         client: HTTPHandler | AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "BytezCustomStreamWrapper":
         if client is None or isinstance(client, HTTPHandler):
             client = get_async_httpx_client(llm_provider=LlmProviders.BYTEZ, params={})

litellm/llms/custom_httpx/llm_http_handler.py

@@ -619,6 +619,7 @@ class BaseLLMHTTPHandler:
                     messages=messages,
                     client=client,
                     json_mode=json_mode,
+                    timeout=timeout,
                 )
             completion_stream, headers = self.make_sync_call(
                 provider_config=provider_config,
@@ -782,6 +783,7 @@ class BaseLLMHTTPHandler:
                 client=client,
                 json_mode=json_mode,
                 signed_json_body=signed_json_body,
+                timeout=timeout,
             )
 
         completion_stream, _response_headers = await self.make_async_call_stream_helper(

litellm/llms/langgraph/chat/transformation.py

@@ -284,6 +284,7 @@ class LangGraphConfig(BaseConfig):
         client: Union[HTTPHandler, "AsyncHTTPHandler"] | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> CustomStreamWrapper:
         """
         Get a CustomStreamWrapper for synchronous streaming.
@@ -343,6 +344,7 @@ class LangGraphConfig(BaseConfig):
         client: Optional["AsyncHTTPHandler"] = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> CustomStreamWrapper:
         """
         Get a CustomStreamWrapper for asynchronous streaming.

litellm/llms/oci/chat/transformation.py

@@ -641,6 +641,7 @@ class OCIChatConfig(BaseConfig):
         client: HTTPHandler | AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "OCIStreamWrapper":
         if client is None or isinstance(client, AsyncHTTPHandler):
             client = _get_httpx_client(params={})
@@ -680,6 +681,7 @@ class OCIChatConfig(BaseConfig):
         client: HTTPHandler | AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "OCIStreamWrapper":
         if client is None or isinstance(client, HTTPHandler):
             client = get_async_httpx_client(llm_provider=LlmProviders.OCI, params={})

litellm/llms/sagemaker/chat/transformation.py

@@ -147,6 +147,7 @@ class SagemakerChatConfig(OpenAIGPTConfig, BaseAWSLLM):
         client: HTTPHandler | AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> CustomStreamWrapper:
         if client is None or isinstance(client, AsyncHTTPHandler):
             client = _get_httpx_client(params={})
@@ -189,6 +190,7 @@ class SagemakerChatConfig(OpenAIGPTConfig, BaseAWSLLM):
         client: HTTPHandler | AsyncHTTPHandler | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> CustomStreamWrapper:
         if client is None or isinstance(client, HTTPHandler):
             try:

litellm/llms/vertex_ai/agent_engine/transformation.py

@@ -364,6 +364,7 @@ class VertexAgentEngineConfig(BaseConfig, VertexBase):
         client: Union[HTTPHandler, "AsyncHTTPHandler"] | None = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "CustomStreamWrapper":
         """Get a CustomStreamWrapper for synchronous streaming."""
         from litellm.llms.custom_httpx.http_handler import (
@@ -422,6 +423,7 @@ class VertexAgentEngineConfig(BaseConfig, VertexBase):
         client: Optional["AsyncHTTPHandler"] = None,
         json_mode: bool | None = None,
         signed_json_body: bytes | None = None,
+        timeout: float | httpx.Timeout | None = None,
     ) -> "CustomStreamWrapper":
         """Get a CustomStreamWrapper for asynchronous streaming."""
         from litellm.llms.custom_httpx.http_handler import (

tests/test_litellm/llms/bedrock/chat/test_bedrock_converse_handler.py

@@ -6,11 +6,12 @@ extension, and AWS credential resolution is stubbed so nothing reaches STS.
 
 from __future__ import annotations
 
-from unittest.mock import MagicMock, patch
+from unittest.mock import AsyncMock, MagicMock, patch
 
 import httpx
 import pytest
 
+import litellm
 from botocore.credentials import Credentials
 from litellm.llms.bedrock.chat.converse_handler import BedrockConverseLLM
 from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler, HTTPHandler
@@ -487,3 +488,66 @@ def test_post_call_is_not_logged_twice_when_the_sync_rust_call_declines():
     assert response.choices[0].message.content == "hi"
     assert len(calls["post_call"]) == 1
     assert "hi" in calls["post_call"][0]["original_response"]
+
+
+@pytest.mark.parametrize(
+    "timeout",
+    [
+        pytest.param(2.25, id="numeric"),
+        pytest.param(httpx.Timeout(2.5), id="httpx-timeout"),
+    ],
+)
+@pytest.mark.asyncio
+async def test_async_converse_streaming_forwards_timeout_to_existing_client(timeout):
+    async def _no_bytes(chunk_size=None):
+        return
+        yield b""
+
+    response = MagicMock()
+    response.status_code = 200
+    response.aiter_bytes = _no_bytes
+    response.headers = httpx.Headers()
+    client = AsyncHTTPHandler()
+    client.post = AsyncMock(return_value=response)
+
+    await litellm.acompletion(
+        model="bedrock/anthropic.claude-haiku-4-5-20251001-v1:0",
+        messages=[{"role": "user", "content": "hi"}],
+        stream=True,
+        timeout=timeout,
+        client=client,
+        aws_access_key_id="fake",
+        aws_secret_access_key="fake",
+        aws_region_name="us-east-1",
+    )
+
+    assert client.post.await_args.kwargs["timeout"] == timeout
+
+
+@pytest.mark.parametrize(
+    "timeout",
+    [
+        pytest.param(3.25, id="numeric"),
+        pytest.param(httpx.Timeout(3.5), id="httpx-timeout"),
+    ],
+)
+def test_sync_converse_streaming_forwards_timeout_to_existing_client(timeout):
+    response = MagicMock()
+    response.status_code = 200
+    response.iter_bytes = MagicMock(return_value=iter(()))
+    response.headers = httpx.Headers()
+    client = HTTPHandler()
+    client.post = MagicMock(return_value=response)
+
+    litellm.completion(
+        model="bedrock/anthropic.claude-haiku-4-5-20251001-v1:0",
+        messages=[{"role": "user", "content": "hi"}],
+        stream=True,
+        timeout=timeout,
+        client=client,
+        aws_access_key_id="fake",
+        aws_secret_access_key="fake",
+        aws_region_name="us-east-1",
+    )
+
+    assert client.post.call_args.kwargs["timeout"] == timeout

tests/test_litellm/llms/bedrock/chat/test_invoke_handler.py

@@ -473,3 +473,70 @@ async def test_async_invoke_streaming_forwards_bedrock_response_headers():
 
     assert stream._hidden_params["additional_headers"]["llm_provider-x-amzn-requestid"] == "req-987"
 
+
+@pytest.mark.parametrize(
+    "timeout",
+    [
+        pytest.param(1.25, id="numeric"),
+        pytest.param(httpx.Timeout(1.5), id="httpx-timeout"),
+    ],
+)
+@pytest.mark.asyncio
+async def test_async_invoke_streaming_forwards_timeout_to_existing_client(timeout):
+    async def _no_bytes(chunk_size=None):
+        return
+        yield b""
+
+    response = MagicMock()
+    response.status_code = 200
+    response.aiter_bytes = _no_bytes
+    response.headers = httpx.Headers()
+    client = AsyncHTTPHandler()
+    client.post = AsyncMock(return_value=response)
+
+    await litellm.acompletion(
+        model="bedrock/invoke/anthropic.claude-haiku-4-5-20251001-v1:0",
+        messages=[{"role": "user", "content": "hi"}],
+        stream=True,
+        timeout=timeout,
+        client=client,
+        aws_access_key_id="fake",
+        aws_secret_access_key="fake",
+        aws_region_name="us-east-1",
+    )
+
+    assert client.post.await_args.kwargs["timeout"] == timeout
+
+
+@pytest.mark.asyncio
+async def test_make_call_sets_timeout_on_httpx_request_transport():
+    transport = MagicMock(return_value=httpx.Response(200, content=b""))
+    timeout = httpx.Timeout(connect=1.0, read=2.0, write=3.0, pool=4.0)
+    client = AsyncHTTPHandler()
+    await client.client.aclose()
+    client.client = httpx.AsyncClient(transport=httpx.MockTransport(transport))
+
+    try:
+        await make_call(
+            client=client,
+            api_base=(
+                "https://bedrock-runtime.us-east-1.amazonaws.com/model/"
+                "anthropic.claude-sonnet-4-6/invoke-with-response-stream"
+            ),
+            headers={},
+            data="{}",
+            model="anthropic.claude-sonnet-4-6",
+            messages=[],
+            logging_obj=MagicMock(),
+            timeout=timeout,
+        )
+    finally:
+        await client.client.aclose()
+
+    assert transport.call_args.args[0].extensions["timeout"] == {
+        "connect": 1.0,
+        "read": 2.0,
+        "write": 3.0,
+        "pool": 4.0,
+    }
+