mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-11 08:38:24 +00:00
* fix(vertex_ai): split Gemini SSE lines on CR/LF only so U+2028 in text keeps the stream The Gemini and Vertex AI streaming path read the SSE body with httpx iter_lines()/aiter_lines(). Those follow str.splitlines(), which also breaks on U+0085, U+2028, U+2029, \x0b, \x0c and \x1c-\x1e. JSON leaves U+0085, U+2028 and U+2029 unescaped inside strings, so a data: line whose text holds one of them was cut in two. The tail is not a data: line and the head never parses, and the accumulate-and-retry fallback then carries the unparsed prefix forward, so every later event is appended to it and the stream ends with no more content and no usage. SSE lines end only at CRLF, LF or CR. The new iter_sse_lines/aiter_sse_lines split decoded text on exactly those, treating a CR at the end of one chunk and an LF at the start of the next as one break. The first request and every long-decode hop now read through them. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(streaming): keep merged think tags balanced and split shared SSE lines only on CR/LF Close an open <think> block on the first content or tool-call delta and at stream end, reopen it when reasoning returns, and route the shared chat completions stream through the SSE line splitter so U+2028 and U+2029 inside JSON no longer drop chunks. * fix(streaming): exact-match the SSE [DONE] marker in BaseModelResponseIterator * fix(streaming): split every SSE reader on CR/LF only and keep the logged think block closed Switch the remaining httpx stream readers (Anthropic chat, Google GenAI, Interactions, Codestral, openai_like, Predibase, LangGraph, Vertex Agent Engine, A2A providers, Azure AI agents, sandboxes, the A2A endpoint, and others) from iter_lines/aiter_lines to iter_sse_lines/aiter_sse_lines, so a raw U+2028 or U+2029 inside a JSON string no longer splits an event. Keep the chunk that closes a think block in the logged and cached response when a stream ends without a finish_reason, close the block before a legacy function_call, and make sync Codestral FIM streaming send stream=true and use the text-completion-codestral stream parser. Test fakes that only mocked iter_lines now stream real httpx bytes and assert the content they feed in. * fix(streaming): split buffered SSE parsers only on CR and LF, shard the codestral tests Buffered SSE and JSONL parsers used str.splitlines(), which also breaks on U+2028, U+2029 and U+0085 that JSON allows raw inside strings. They now use split_sse_lines from sse_lines.py, which also gains a str.split fast path. Adds tests/unit/llms/codestral to the unit workflow shard. * fix(streaming): keep U+2028 in restamped Anthropic frames and fold CRLF before splitting SSE lines * test: type the new stream tests, drop a comment, and benchmark the SSE line splitter * test: type the iterator class parameters in the line-separator stream tests * fix(responses): stream tool call arguments opened in a text chunk * fix(responses): hold a tool call that shares a chunk with text until the text ends * test: type the a2a SSE handler helper and the sandbox fake responses --------- Co-authored-by: Tim Vink <vinktim@gmail.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com> Co-authored-by: michelligabriele <gabriele.michelli@icloud.com>
484 lines
18 KiB
Python
484 lines
18 KiB
Python
import json
|
|
from typing import Final
|
|
from unittest.mock import AsyncMock, Mock, patch
|
|
|
|
import aiohttp
|
|
import httpx
|
|
import pytest
|
|
|
|
import litellm
|
|
from litellm.llms.aiohttp_openai.chat.transformation import AiohttpOpenAIChatConfig
|
|
from litellm.llms.custom_httpx.aiohttp_handler import BaseLLMAIOHTTPHandler
|
|
from litellm.llms.custom_httpx.aiohttp_transport import LiteLLMAiohttpTransport
|
|
from litellm.llms.custom_httpx.http_handler import HTTPHandler
|
|
from litellm.types.utils import ModelResponseStream
|
|
|
|
|
|
class TestBaseLLMAIOHTTPHandler:
|
|
"""Test cases for BaseLLMAIOHTTPHandler dependency injection functionality"""
|
|
|
|
def test_init_with_no_client_session(self):
|
|
"""Test handler initialization with no client session"""
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
|
|
assert handler.client_session is None
|
|
assert handler._owns_session is True
|
|
|
|
def test_init_with_provided_client_session(self):
|
|
"""Test handler initialization with provided client session"""
|
|
# Create a mock client session
|
|
mock_session = Mock()
|
|
|
|
handler = BaseLLMAIOHTTPHandler(client_session=mock_session)
|
|
|
|
assert handler.client_session is mock_session
|
|
assert handler._owns_session is False
|
|
|
|
def test_get_async_client_session_with_dynamic_session(self):
|
|
"""Test _get_async_client_session with dynamic session parameter"""
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
|
|
dynamic_session = Mock()
|
|
|
|
result = handler._get_async_client_session(
|
|
dynamic_client_session=dynamic_session
|
|
)
|
|
|
|
assert result is dynamic_session
|
|
|
|
def test_get_async_client_session_with_instance_session(self):
|
|
"""Test _get_async_client_session with instance session"""
|
|
instance_session = Mock()
|
|
handler = BaseLLMAIOHTTPHandler(client_session=instance_session)
|
|
|
|
result = handler._get_async_client_session()
|
|
|
|
assert result is instance_session
|
|
|
|
@patch("aiohttp.ClientSession")
|
|
def test_get_async_client_session_create_new(self, mock_client_session):
|
|
"""Test _get_async_client_session creates new session when none provided"""
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
mock_session_instance = Mock()
|
|
mock_client_session.return_value = mock_session_instance
|
|
|
|
result = handler._get_async_client_session()
|
|
|
|
# Verify new session was created
|
|
mock_client_session.assert_called_once()
|
|
assert handler.client_session is mock_session_instance
|
|
assert handler._owns_session is True
|
|
assert result is mock_session_instance
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_close_with_owned_session(self):
|
|
"""Test close() method with owned session"""
|
|
# Create a mock session that we own
|
|
mock_session = Mock()
|
|
mock_session.closed = False
|
|
mock_session.close = AsyncMock()
|
|
|
|
# Create handler that owns the session
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
handler.client_session = mock_session
|
|
handler._owns_session = True
|
|
|
|
await handler.close()
|
|
|
|
# Verify close was called
|
|
mock_session.close.assert_called_once()
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_close_with_non_owned_session(self):
|
|
"""Test close() method with non-owned session (should not close)"""
|
|
# Create a mock session that we don't own
|
|
mock_session = Mock()
|
|
mock_session.closed = False
|
|
mock_session.close = AsyncMock()
|
|
|
|
handler = BaseLLMAIOHTTPHandler(client_session=mock_session)
|
|
|
|
await handler.close()
|
|
|
|
# Verify close was NOT called since we don't own this session
|
|
mock_session.close.assert_not_called()
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_close_with_already_closed_session(self):
|
|
"""Test close() method with already closed session"""
|
|
mock_session = Mock()
|
|
mock_session.closed = True
|
|
mock_session.close = AsyncMock()
|
|
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
handler.client_session = mock_session
|
|
handler._owns_session = True
|
|
|
|
await handler.close()
|
|
|
|
# Verify close was NOT called since session is already closed
|
|
mock_session.close.assert_not_called()
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_close_with_no_session(self):
|
|
"""Test close() method with no session"""
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
|
|
# Should not raise any exceptions
|
|
await handler.close()
|
|
|
|
def test_session_priority_dynamic_over_instance(self):
|
|
"""Test that dynamic session takes priority over instance session"""
|
|
instance_session = Mock()
|
|
dynamic_session = Mock()
|
|
|
|
handler = BaseLLMAIOHTTPHandler(client_session=instance_session)
|
|
|
|
result = handler._get_async_client_session(
|
|
dynamic_client_session=dynamic_session
|
|
)
|
|
|
|
assert result is dynamic_session
|
|
assert result is not instance_session
|
|
|
|
def test_session_ownership_tracking(self):
|
|
"""Test proper session ownership tracking in various scenarios"""
|
|
# Scenario 1: Provided session - not owned
|
|
provided_session = Mock()
|
|
handler1 = BaseLLMAIOHTTPHandler(client_session=provided_session)
|
|
assert not handler1._owns_session
|
|
|
|
# Scenario 2: No session initially - becomes owned when created
|
|
handler2 = BaseLLMAIOHTTPHandler()
|
|
assert handler2._owns_session
|
|
|
|
with patch("aiohttp.ClientSession") as mock_client_session:
|
|
mock_session_instance = Mock()
|
|
mock_client_session.return_value = mock_session_instance
|
|
|
|
handler2._get_async_client_session()
|
|
assert handler2._owns_session
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_context_manager_pattern_compatibility(self):
|
|
"""Test that the handler works well with context manager pattern"""
|
|
mock_session = Mock()
|
|
mock_session.closed = False
|
|
mock_session.close = AsyncMock()
|
|
|
|
# Test as context manager style usage
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
handler.client_session = mock_session
|
|
handler._owns_session = True
|
|
|
|
try:
|
|
# Simulate some work
|
|
session = handler._get_async_client_session()
|
|
assert session is mock_session
|
|
finally:
|
|
await handler.close()
|
|
|
|
# Verify cleanup happened
|
|
mock_session.close.assert_called_once()
|
|
|
|
@patch("litellm.llms.custom_httpx.aiohttp_handler.aiohttp.ClientSession")
|
|
def test_lazy_session_creation(self, mock_client_session):
|
|
"""Test that session is created lazily only when needed"""
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
|
|
# Session should not be created on init
|
|
mock_client_session.assert_not_called()
|
|
assert handler.client_session is None
|
|
|
|
# Session should be created when requested
|
|
mock_session_instance = Mock()
|
|
mock_client_session.return_value = mock_session_instance
|
|
|
|
session = handler._get_async_client_session()
|
|
|
|
mock_client_session.assert_called_once()
|
|
assert session is mock_session_instance
|
|
assert handler.client_session is mock_session_instance
|
|
|
|
def test_session_reuse(self):
|
|
"""Test that the same session is reused across multiple calls"""
|
|
instance_session = Mock()
|
|
handler = BaseLLMAIOHTTPHandler(client_session=instance_session)
|
|
|
|
# Multiple calls should return the same session
|
|
session1 = handler._get_async_client_session()
|
|
session2 = handler._get_async_client_session()
|
|
session3 = handler._get_async_client_session()
|
|
|
|
assert session1 is session2 is session3 is instance_session
|
|
|
|
# ===============================
|
|
# TRANSPORT INJECTION TESTS
|
|
# ===============================
|
|
|
|
def test_init_with_transport(self):
|
|
"""Test handler initialization with provided transport"""
|
|
mock_transport = Mock(spec=LiteLLMAiohttpTransport)
|
|
|
|
handler = BaseLLMAIOHTTPHandler(transport=mock_transport)
|
|
|
|
assert handler.transport is mock_transport
|
|
assert handler._owns_transport is False
|
|
|
|
def test_init_with_connector(self):
|
|
"""Test handler initialization with provided connector"""
|
|
mock_connector = Mock(spec=aiohttp.BaseConnector)
|
|
|
|
handler = BaseLLMAIOHTTPHandler(connector=mock_connector)
|
|
|
|
assert handler.connector is mock_connector
|
|
assert handler._owns_connector is False
|
|
|
|
def test_init_with_transport_and_session(self):
|
|
"""Test handler initialization with both transport and session"""
|
|
mock_transport = Mock(spec=LiteLLMAiohttpTransport)
|
|
mock_session = Mock()
|
|
|
|
handler = BaseLLMAIOHTTPHandler(
|
|
client_session=mock_session, transport=mock_transport
|
|
)
|
|
|
|
assert handler.transport is mock_transport
|
|
assert handler._owns_transport is False
|
|
assert handler.client_session is mock_session
|
|
assert handler._owns_session is False
|
|
|
|
def test_get_connector_from_provided_connector(self):
|
|
"""Test _get_connector returns provided connector"""
|
|
mock_connector = Mock(spec=aiohttp.BaseConnector)
|
|
handler = BaseLLMAIOHTTPHandler(connector=mock_connector)
|
|
|
|
result = handler._get_connector()
|
|
|
|
assert result is mock_connector
|
|
|
|
def test_get_connector_from_transport(self):
|
|
"""Test _get_connector extracts connector from transport"""
|
|
mock_connector = Mock(spec=aiohttp.BaseConnector)
|
|
|
|
# Use a simple object instead of Mock to avoid callable issues
|
|
class MockSession:
|
|
def __init__(self):
|
|
self.connector = mock_connector
|
|
|
|
mock_session = MockSession()
|
|
|
|
mock_transport = Mock(spec=LiteLLMAiohttpTransport)
|
|
mock_transport.client = mock_session
|
|
|
|
handler = BaseLLMAIOHTTPHandler(transport=mock_transport)
|
|
|
|
result = handler._get_connector()
|
|
|
|
assert result is mock_connector
|
|
|
|
def test_get_connector_from_transport_with_callable_client(self):
|
|
"""Test _get_connector with transport that has callable client"""
|
|
mock_transport = Mock(spec=LiteLLMAiohttpTransport)
|
|
mock_transport.client = lambda: Mock() # Callable client
|
|
|
|
handler = BaseLLMAIOHTTPHandler(transport=mock_transport)
|
|
|
|
result = handler._get_connector()
|
|
|
|
assert result is None
|
|
|
|
@patch("aiohttp.ClientSession")
|
|
def test_create_client_session_with_transport(self, mock_client_session):
|
|
"""Test session creation using transport"""
|
|
mock_session_from_transport = Mock()
|
|
|
|
mock_transport = Mock(spec=LiteLLMAiohttpTransport)
|
|
mock_transport.get_valid_client_session = Mock(return_value=mock_session_from_transport)
|
|
|
|
handler = BaseLLMAIOHTTPHandler(transport=mock_transport)
|
|
|
|
result = handler._create_client_session_with_transport()
|
|
|
|
# Should use transport's session creation method
|
|
mock_transport.get_valid_client_session.assert_called_once()
|
|
assert result is mock_session_from_transport
|
|
|
|
# Should not call aiohttp.ClientSession directly
|
|
mock_client_session.assert_not_called()
|
|
|
|
@patch("aiohttp.ClientSession")
|
|
def test_create_client_session_with_connector(self, mock_client_session):
|
|
"""Test session creation using connector"""
|
|
mock_connector = Mock(spec=aiohttp.BaseConnector)
|
|
mock_session_instance = Mock()
|
|
mock_client_session.return_value = mock_session_instance
|
|
|
|
handler = BaseLLMAIOHTTPHandler(connector=mock_connector)
|
|
|
|
result = handler._create_client_session_with_transport()
|
|
|
|
# Should create session with connector
|
|
mock_client_session.assert_called_once_with(connector=mock_connector)
|
|
assert result is mock_session_instance
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_create_client_session_default_honors_global_ssl_verify_false(
|
|
self, monkeypatch: pytest.MonkeyPatch
|
|
):
|
|
"""Regression test for LIT-3369: `litellm.ssl_verify = False` (set via
|
|
`litellm_settings.ssl_verify: false`) must reach the default session's
|
|
connector instead of being ignored by a bare `aiohttp.ClientSession()`."""
|
|
monkeypatch.setattr(litellm, "ssl_verify", False)
|
|
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
session = handler._create_client_session_with_transport()
|
|
try:
|
|
assert isinstance(session.connector, aiohttp.TCPConnector)
|
|
assert session.connector._ssl is False
|
|
finally:
|
|
await session.close()
|
|
await handler.close()
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_create_client_session_default_keeps_ssl_verification(self):
|
|
"""Default `ssl_verify=True` must not collapse to `ssl=False`."""
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
session = handler._create_client_session_with_transport()
|
|
try:
|
|
assert isinstance(session.connector, aiohttp.TCPConnector)
|
|
assert session.connector._ssl is not False
|
|
finally:
|
|
await session.close()
|
|
await handler.close()
|
|
|
|
def test_get_or_create_transport_resolves_global_ssl_verify(
|
|
self, monkeypatch: pytest.MonkeyPatch
|
|
):
|
|
"""The lazily created transport must carry the resolved global ssl config."""
|
|
monkeypatch.setattr(litellm, "ssl_verify", False)
|
|
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
transport = handler._get_or_create_transport()
|
|
|
|
assert transport is not None
|
|
assert transport._ssl_verify is False
|
|
|
|
def test_get_or_create_transport(self):
|
|
"""Test that _get_or_create_transport creates or returns a transport.
|
|
|
|
When no transport exists, the method should attempt to create one.
|
|
If creation succeeds, it should be stored on the handler.
|
|
If creation fails (e.g. in test environments), None is returned gracefully.
|
|
"""
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
assert handler.transport is None # no transport initially
|
|
|
|
result = handler._get_or_create_transport()
|
|
|
|
# The method should either create a transport or gracefully return None
|
|
if result is not None:
|
|
assert handler.transport is result
|
|
assert handler._owns_transport is True
|
|
assert isinstance(result, LiteLLMAiohttpTransport)
|
|
else:
|
|
# Creation can fail in test environments without full aiohttp setup
|
|
# This is the graceful fallback path
|
|
pass
|
|
|
|
def test_get_or_create_transport_with_existing(self):
|
|
"""Test _get_or_create_transport returns existing transport"""
|
|
mock_transport = Mock(spec=LiteLLMAiohttpTransport)
|
|
handler = BaseLLMAIOHTTPHandler(transport=mock_transport)
|
|
|
|
result = handler._get_or_create_transport()
|
|
|
|
assert result is mock_transport
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_close_with_owned_transport(self):
|
|
"""Test close() method with owned transport"""
|
|
mock_transport = Mock(spec=LiteLLMAiohttpTransport)
|
|
mock_transport.aclose = AsyncMock()
|
|
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
handler.transport = mock_transport
|
|
handler._owns_transport = True
|
|
|
|
await handler.close()
|
|
|
|
# Verify transport close was called
|
|
mock_transport.aclose.assert_called_once()
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_close_with_non_owned_transport(self):
|
|
"""Test close() method with non-owned transport (should not close)"""
|
|
mock_transport = Mock(spec=LiteLLMAiohttpTransport)
|
|
mock_transport.aclose = AsyncMock()
|
|
|
|
handler = BaseLLMAIOHTTPHandler(transport=mock_transport)
|
|
|
|
await handler.close()
|
|
|
|
# Verify transport close was NOT called since we don't own this transport
|
|
mock_transport.aclose.assert_not_called()
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_close_transport_without_aclose_method(self):
|
|
"""Test close() handles transport without aclose method gracefully"""
|
|
mock_transport = Mock() # No aclose method
|
|
|
|
handler = BaseLLMAIOHTTPHandler()
|
|
handler.transport = mock_transport
|
|
handler._owns_transport = True
|
|
|
|
# Should not raise any exceptions
|
|
await handler.close()
|
|
|
|
def test_transport_priority_hierarchy(self):
|
|
"""Test that session creation follows the right priority: transport > connector > default"""
|
|
# Test with transport having get_valid_client_session
|
|
mock_transport = Mock(spec=LiteLLMAiohttpTransport)
|
|
mock_session_from_transport = Mock()
|
|
mock_transport.get_valid_client_session = Mock(return_value=mock_session_from_transport)
|
|
|
|
mock_connector = Mock(spec=aiohttp.BaseConnector)
|
|
|
|
handler = BaseLLMAIOHTTPHandler(transport=mock_transport, connector=mock_connector)
|
|
|
|
result = handler._create_client_session_with_transport()
|
|
|
|
# Should use transport, not connector
|
|
mock_transport.get_valid_client_session.assert_called_once()
|
|
assert result is mock_session_from_transport
|
|
|
|
|
|
def test_make_sync_call_stream_keeps_unicode_line_separators_inside_a_delta() -> None:
|
|
text: Final = "before\u2028between\u2029after"
|
|
chunk: Final = {
|
|
"id": "c1",
|
|
"object": "chat.completion.chunk",
|
|
"created": 1,
|
|
"model": "gpt-5.5",
|
|
"choices": [{"index": 0, "delta": {"content": text}, "finish_reason": None}],
|
|
}
|
|
body: Final = f"data: {json.dumps(chunk, ensure_ascii=False)}\n\ndata: [DONE]\n\n".encode()
|
|
transport: Final = httpx.MockTransport(lambda request: httpx.Response(200, stream=httpx.ByteStream(body)))
|
|
|
|
completion_stream, _ = BaseLLMAIOHTTPHandler().make_sync_call(
|
|
provider_config=AiohttpOpenAIChatConfig(),
|
|
api_base="https://example.com/v1/chat/completions",
|
|
headers={},
|
|
data={"model": "gpt-5.5", "messages": []},
|
|
model="gpt-5.5",
|
|
messages=[],
|
|
logging_obj=Mock(),
|
|
litellm_params={},
|
|
timeout=5.0,
|
|
client=HTTPHandler(client=httpx.Client(transport=transport)),
|
|
)
|
|
|
|
contents: Final = [
|
|
chunk.choices[0].delta.content for chunk in completion_stream if isinstance(chunk, ModelResponseStream)
|
|
]
|
|
assert contents == [text]
|