Skip to content

The fastest, litest AI Gateway. Rust core with Python SDK. Call 100+ LLM APIs in OpenAI (or native) format with cost tracking, guardrails, load balancing, and logging [Bedrock, Azure, OpenAI, Anthropic, OpenAI, VertexAI, vLLM, Nvidia NIM]

14.5 KB
Raw Blame
import asyncio
import importlib
import logging
import os
import time
from collections.abc import Iterator
from typing import Awaitable, Final, TypeVar
from unittest.mock import AsyncMock, Mock

import pytest

import litellm
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.logging_worker import GLOBAL_LOGGING_WORKER
from litellm.utils import _invalidate_model_cost_lowercase_map
from tests._vcr_conftest_common import install_live_call_probe, record_vcr_outcome


_T = TypeVar("_T")


async def _await_with_spawned_tasks(call: Awaitable[_T]) -> _T:
    before: Final = asyncio.all_tasks()
    result: Final = await call
    spawned: Final = asyncio.all_tasks() - before - {asyncio.current_task()}
    await asyncio.wait_for(asyncio.gather(*spawned), timeout=5)
    return result


@pytest.fixture(autouse=True)
def isolate_opik_logging_state(monkeypatch: pytest.MonkeyPatch) -> Iterator[None]:
    original_level = verbose_logger.level
    verbose_logger.setLevel(logging.DEBUG)
    monkeypatch.setattr(litellm, "set_verbose", True)
    yield
    verbose_logger.setLevel(original_level)


INTERVAL_TOO_LONG_TO_FIRE_DURING_THIS_TEST = 3600


@pytest.mark.asyncio
async def test_opik_logging_http_request():
    """
    - Test that HTTP requests are made to Opik
    - Traces and spans are batched correctly
    """
    from litellm.integrations.opik.opik import OpikLogger

    os.environ["OPIK_URL_OVERRIDE"] = "https://fake.comet.com/opik/api"
    os.environ["OPIK_API_KEY"] = "anything"
    os.environ["OPIK_WORKSPACE"] = "anything"

    test_opik_logger = OpikLogger()
    test_opik_logger.flush_interval = INTERVAL_TOO_LONG_TO_FIRE_DURING_THIS_TEST
    test_opik_logger.batch_size = 12

    litellm.callbacks = [test_opik_logger]

    mock_post = AsyncMock(return_value=Mock(status_code=202, text="Accepted"))
    test_opik_logger.async_httpx_client.post = mock_post

    def opik_batch_calls():
        return [
            call for call in mock_post.call_args_list if "/traces/batch" in str(call) or "/spans/batch" in str(call)
        ]

    for _ in range(5):
        await _await_with_spawned_tasks(
            litellm.acompletion(
                model="gpt-3.5-turbo",
                messages=[{"role": "user", "content": "Test message"}],
                max_tokens=10,
                temperature=0.2,
                mock_response="This is a mock response",
            )
        )
    await asyncio.wait_for(GLOBAL_LOGGING_WORKER.flush(), timeout=10)

    assert opik_batch_calls() == [], "events below batch_size must stay queued"
    assert len(test_opik_logger.log_queue) == 10

    for _ in range(3):
        await _await_with_spawned_tasks(
            litellm.acompletion(
                model="gpt-3.5-turbo",
                messages=[{"role": "user", "content": "Test message"}],
                max_tokens=10,
                temperature=0.2,
                mock_response="This is a mock response",
            )
        )
    await asyncio.wait_for(GLOBAL_LOGGING_WORKER.flush(), timeout=10)

    assert opik_batch_calls(), "crossing batch_size must flush the queue"
    events_left_over_after_the_size_triggered_flush = len(test_opik_logger.log_queue)
    assert 0 < events_left_over_after_the_size_triggered_flush < test_opik_logger.batch_size

    calls_before_periodic_flush = len(opik_batch_calls())
    await test_opik_logger.flush_queue()

    assert len(opik_batch_calls()) > calls_before_periodic_flush
    assert len(test_opik_logger.log_queue) == 0


def test_sync_opik_logging_http_request():
    """
    - Test that HTTP requests are made to Opik
    - Traces and spans are batched correctly
    """
    try:
        from litellm.integrations.opik.opik import OpikLogger

        os.environ["OPIK_URL_OVERRIDE"] = "https://fake.comet.com/opik/api"
        os.environ["OPIK_API_KEY"] = "anything"
        os.environ["OPIK_WORKSPACE"] = "anything"

        # Initialize OpikLogger
        test_opik_logger = OpikLogger()

        litellm.callbacks = [test_opik_logger]
        litellm.set_verbose = True

        # Create a mock for the clients's post method
        mock_post = Mock()
        mock_post.return_value.status_code = 204
        mock_post.return_value.text = "Accepted"
        test_opik_logger.sync_httpx_client.post = mock_post

        # Make multiple calls to ensure we don't hit the batch size
        for _ in range(5):
            response = litellm.completion(
                model="gpt-3.5-turbo",
                messages=[{"role": "user", "content": "Test message"}],
                max_tokens=10,
                temperature=0.2,
                mock_response="This is a mock response",
            )

        # Need to wait for a short amount of time as the log_success callback is called in a different thread. One or two seconds is often not enough.
        time.sleep(3)

        # Check that 5 spans and 5 traces were sent
        assert mock_post.call_count == 10, f"Expected 10 HTTP requests, but got {mock_post.call_count}"

    except Exception as e:
        pytest.fail(f"Error occurred: {e}")


def test_opik_attach_to_existing_trace():
    """
    Test attaching spans to existing trace (regression fix for PR #14888)

    - When trace_id is provided via current_span_data, only create a span
    - Do NOT create a new trace (this was the bug)
    - Verify span has correct trace_id and parent_span_id
    """
    try:
        from litellm.integrations.opik.opik import OpikLogger

        os.environ["OPIK_URL_OVERRIDE"] = "https://fake.comet.com/opik/api"
        os.environ["OPIK_API_KEY"] = "anything"
        os.environ["OPIK_WORKSPACE"] = "anything"

        # Initialize OpikLogger
        test_opik_logger = OpikLogger()
        litellm.callbacks = [test_opik_logger]
        litellm.set_verbose = True

        # Create a mock for the sync client's post method
        mock_post = Mock()
        mock_post.return_value.status_code = 204
        mock_post.return_value.text = "Accepted"
        test_opik_logger.sync_httpx_client.post = mock_post

        # Simulate an existing trace and parent span
        existing_trace_id = "existing-trace-12345"
        existing_parent_span_id = "existing-span-67890"

        # Make a completion call with existing trace_id
        response = litellm.completion(
            model="gpt-3.5-turbo",
            messages=[{"role": "user", "content": "Test message"}],
            max_tokens=10,
            temperature=0.2,
            mock_response="This is a mock response",
            metadata={
                "opik": {
                    "current_span_data": {
                        "trace_id": existing_trace_id,
                        "id": existing_parent_span_id,
                    },
                    "tags": ["test-attach-span"],
                }
            },
        )

        # Need to wait for a short amount of time as the log_success callback is called in a different thread
        time.sleep(3)

        # Check the calls made to the mock
        calls_made = mock_post.call_args_list
        trace_calls = [call for call in calls_made if "/traces/batch" in str(call)]
        span_calls = [call for call in calls_made if "/spans/batch" in str(call)]

        # With the fix, when trace_id is provided, we should NOT create a new trace
        assert len(trace_calls) == 0, (
            f"Expected 0 trace calls when attaching to existing trace, but got {len(trace_calls)}"
        )
        assert len(span_calls) == 1, f"Expected exactly 1 span call, but got {len(span_calls)}"

        # Verify span has correct trace_id and parent_span_id
        span_payload = span_calls[0][1]["json"]["spans"][0]
        assert span_payload["trace_id"] == existing_trace_id, (
            f"Expected trace_id to be {existing_trace_id}, but got {span_payload['trace_id']}"
        )
        assert span_payload["parent_span_id"] == existing_parent_span_id, (
            f"Expected parent_span_id to be {existing_parent_span_id}, but got {span_payload['parent_span_id']}"
        )
        assert "test-attach-span" in span_payload["tags"], f"Expected 'test-attach-span' tag in {span_payload['tags']}"

    except Exception as e:
        pytest.fail(f"Error occurred: {e}")


def test_opik_create_new_trace():
    """
    Test normal trace creation when no trace_id is provided

    - When NO trace_id is provided, create both a new trace and a new span
    - Verify the span references the created trace
    - Verify tags are included in both trace and span
    """
    try:
        from litellm.integrations.opik.opik import OpikLogger

        os.environ["OPIK_URL_OVERRIDE"] = "https://fake.comet.com/opik/api"
        os.environ["OPIK_API_KEY"] = "anything"
        os.environ["OPIK_WORKSPACE"] = "anything"

        # Initialize OpikLogger
        test_opik_logger = OpikLogger()
        litellm.callbacks = [test_opik_logger]
        litellm.set_verbose = True

        # Create a mock for the sync client's post method
        mock_post = Mock()
        mock_post.return_value.status_code = 204
        mock_post.return_value.text = "Accepted"
        test_opik_logger.sync_httpx_client.post = mock_post

        # Make a completion call WITHOUT providing trace_id
        response = litellm.completion(
            model="gpt-3.5-turbo",
            messages=[{"role": "user", "content": "Test message"}],
            max_tokens=10,
            temperature=0.2,
            mock_response="This is a mock response",
            metadata={"opik": {"tags": ["test-new-trace"]}},
        )

        # Need to wait for a short amount of time as the log_success callback is called in a different thread
        time.sleep(3)

        # Check the calls made to the mock
        calls_made = mock_post.call_args_list
        trace_calls = [call for call in calls_made if "/traces/batch" in str(call)]
        span_calls = [call for call in calls_made if "/spans/batch" in str(call)]

        # Without trace_id provided, we should create both a new trace and a new span
        assert len(trace_calls) == 1, f"Expected exactly 1 trace call, but got {len(trace_calls)}"
        assert len(span_calls) == 1, f"Expected exactly 1 span call, but got {len(span_calls)}"

        # Verify the span references the created trace
        trace_payload = trace_calls[0][1]["json"]["traces"][0]
        span_payload = span_calls[0][1]["json"]["spans"][0]
        assert span_payload["trace_id"] == trace_payload["id"], "Span should reference the created trace"

        # Verify tags are included in both trace and span
        assert "test-new-trace" in trace_payload["tags"], f"Expected 'test-new-trace' tag in trace tags"
        assert "test-new-trace" in span_payload["tags"], f"Expected 'test-new-trace' tag in span tags"

    except Exception as e:
        pytest.fail(f"Error occurred: {e}")


@pytest.fixture(autouse=True)
def _vcr_outcome_gate(request, vcr):
    install_live_call_probe(request, vcr)
    yield
    record_vcr_outcome(request, vcr)


@pytest.fixture(scope="function", autouse=True)
def isolate_litellm_state():
    """
    Per-function isolation fixture.

    Resets litellm globals to their true defaults before each test and
    restores them afterward, so tests don't leak side effects.
    Works safely under pytest-xdist parallel execution.
    """
    original_state = {}
    for attr in (
        "callbacks",
        "success_callback",
        "failure_callback",
        "_async_success_callback",
        "_async_failure_callback",
    ):
        if hasattr(litellm, attr):
            val = getattr(litellm, attr)
            original_state[attr] = val.copy() if val else []
    for attr in ("pre_call_rules", "post_call_rules"):
        if hasattr(litellm, attr):
            val = getattr(litellm, attr)
            original_state[attr] = val.copy() if val else []
    for attr in _SCALAR_DEFAULTS:
        if hasattr(litellm, attr):
            original_state[attr] = getattr(litellm, attr)
    if hasattr(litellm, "in_memory_llm_clients_cache"):
        litellm.in_memory_llm_clients_cache.flush_cache()
    for attr in (
        "callbacks",
        "success_callback",
        "failure_callback",
        "_async_success_callback",
        "_async_failure_callback",
        "pre_call_rules",
        "post_call_rules",
    ):
        if hasattr(litellm, attr):
            setattr(litellm, attr, [])
    for attr, default_val in _SCALAR_DEFAULTS.items():
        if hasattr(litellm, attr):
            setattr(litellm, attr, default_val)
    yield
    asyncio.run(GLOBAL_LOGGING_WORKER.clear_queue())
    if hasattr(litellm, "in_memory_llm_clients_cache"):
        litellm.in_memory_llm_clients_cache.flush_cache()
    for attr, original_value in original_state.items():
        if hasattr(litellm, attr):
            setattr(litellm, attr, original_value)
    _invalidate_model_cost_lowercase_map()


_SCALAR_DEFAULTS = {
    "num_retries": getattr(litellm, "num_retries", None),
    "num_retries_per_request": getattr(litellm, "num_retries_per_request", None),
    "request_timeout": getattr(litellm, "request_timeout", None),
    "set_verbose": getattr(litellm, "set_verbose", False),
    "cache": getattr(litellm, "cache", None),
    "allowed_fails": getattr(litellm, "allowed_fails", 3),
    "default_fallbacks": getattr(litellm, "default_fallbacks", None),
    "enable_azure_ad_token_refresh": getattr(litellm, "enable_azure_ad_token_refresh", None),
    "tag_budget_config": getattr(litellm, "tag_budget_config", None),
    "model_cost": getattr(litellm, "model_cost", None),
    "token_counter": getattr(litellm, "token_counter", None),
    "disable_aiohttp_transport": getattr(litellm, "disable_aiohttp_transport", False),
    "force_ipv4": getattr(litellm, "force_ipv4", False),
    "drop_params": getattr(litellm, "drop_params", None),
    "modify_params": getattr(litellm, "modify_params", False),
    "api_base": getattr(litellm, "api_base", None),
    "api_key": getattr(litellm, "api_key", None),
}


@pytest.fixture(scope="module", autouse=True)
def setup_and_teardown():
    """
    Module-scoped setup. Reloads litellm only in single-process mode
    (skipped under xdist to avoid cross-worker interference).
    """
    worker_id = os.environ.get("PYTEST_XDIST_WORKER", None)
    if worker_id is None:
        importlib.reload(litellm)
        try:
            if hasattr(litellm, "proxy") and hasattr(litellm.proxy, "proxy_server"):
                import litellm.proxy.proxy_server as proxy_server

                importlib.reload(proxy_server)
        except Exception as e:
            print(f"Error reloading litellm.proxy.proxy_server: {e}")
        if hasattr(litellm, "in_memory_llm_clients_cache"):
            litellm.in_memory_llm_clients_cache.flush_cache()
    yield