Files
litellm/tests/llm_responses_api_testing/base_responses_api.py
Mateo Wang 2c733c00f5 chore(ci): modernize model references in tests and configs (#27856)
* test: modernize models used in CircleCI e2e test suites

Replaces obsolete models (gpt-4o, gpt-4o-mini, gpt-3.5-turbo,
claude-3-5-sonnet-20240620, claude-sonnet-4-20250514) with current
equivalents across the e2e_openai_endpoints and
proxy_e2e_anthropic_messages_tests CircleCI jobs.

- gpt-4o -> gpt-5.5 (responses API e2e tests)
- gpt-4o-mini -> gpt-5-mini (websocket responses, oai_misc_config)
- gpt-4o-mini-2024-07-18 -> gpt-4.1-mini-2025-04-14 (fine-tuning,
  still actively fine-tunable)
- gpt-4 / gpt-3.5-turbo target_model_names example -> gpt-5.5 /
  gpt-5-mini
- bedrock claude-3-5-sonnet-20240620 batch entry -> haiku-4-5-20251001
  (also aligning oai_misc_config model_name with what
  test_bedrock_batches_api.py actually requests)
- bedrock claude-sonnet-4-20250514 (deprecated, retires 2026-06-15)
  -> claude-sonnet-4-5-20250929

* test: point bedrock-claude-sonnet-4 alias at Sonnet 4.6, not 4.5

Greptile/Cursor flagged that after the previous commit, the
bedrock-claude-sonnet-4 alias collided with bedrock-claude-sonnet-4.5
(both pointed to claude-sonnet-4-5-20250929). Rename to
bedrock-claude-sonnet-4.6 and point it at the Sonnet 4.6 Bedrock ID
(us.anthropic.claude-sonnet-4-6, already in the litellm model
registry) so the alias name matches the underlying model version.

* test: modernize models across remaining CI-mounted configs & tests

Expands the modernization sweep to all CircleCI-mounted proxy configs
and to test directories where the model literal is a fixture/route key
(not the test's subject).

Config changes:
- proxy_server_config.yaml: bump gpt-3.5-turbo / gpt-3.5-turbo-1106 /
  gpt-4o / gemini-1.5-flash / dall-e-3 underlying models; rename
  gpt-3.5-turbo-end-user-test alias to gpt-5-mini-end-user-test; bump
  text-embedding-ada-002 underlying to text-embedding-3-small. User-
  facing aliases (gpt-3.5-turbo, gpt-4, text-embedding-ada-002, etc.)
  preserved for backward compatibility with tests.
- simple_config.yaml, otel_test_config.yaml, spend_tracking_config.yaml:
  bump gpt-3.5-turbo underlying to gpt-5-mini.
- pass_through_config.yaml: claude-3-5-sonnet / claude-3-7-sonnet /
  claude-3-haiku entries replaced with claude-sonnet-4-5 / claude-
  haiku-4-5 / claude-opus-4-7.
- oai_misc_config.yaml: align alias name with the gpt-5-mini rename.

Test changes (proactive: claude-sonnet-4-20250514 / claude-opus-4-
20250514 retire 2026-06-15):
- tests/llm_translation/test_anthropic_completion.py: bump 3 references
  + paired Vertex AI ID to claude-sonnet-4-5.
- tests/llm_translation/test_optional_params.py: bump 2 references.
- tests/pass_through_unit_tests/test_anthropic_messages_passthrough.py
  and test_bedrock_anthropic_messages_test.py: bump router fixtures
  using the deprecated model IDs.
- tests/pass_through_unit_tests/base_anthropic_messages_tool_search_test.py:
  modernize docstring examples.
- tests/test_end_users.py: update references to renamed alias.

* test: modernize placeholder model literals in router_unit_tests

Mass replace_all on fixture/placeholder model literals across the
router_unit_tests/ suite (model name is a routing key / label, not the
test subject). Sub-agent sweep so far — additional commits will follow
for logging_callback_tests/, enterprise/, top-level tests/test_*.py,
and other CI-mounted dirs.

Mappings applied:
- gpt-3.5-turbo -> gpt-5-mini
- gpt-4 (bare) -> gpt-5.5
- gpt-4o (bare) -> gpt-5
- text-embedding-ada-002 -> text-embedding-3-small
- claude-3-sonnet-20240229 / claude-3-opus-20240229 /
  claude-3-haiku-20240307 / claude-3-5-sonnet-20240620 ->
  claude-sonnet-4-5-20250929 / claude-opus-4-7 /
  claude-haiku-4-5-20251001 as appropriate

Explicitly preserved:
- gpt-4o-mini-* variants (transcribe, tts, etc.) where they're current
- gpt-4-turbo / gpt-4-vision-preview / gpt-4-0613 (subject literals)
- JSONL batch body literals
- Mock LLM response model fields (must match upstream)
- Fake/mock identifiers

* test: modernize placeholder model literals across remaining CI suites

Sub-agent sweep across logging_callback_tests/, guardrails_tests/,
enterprise/, pass_through_unit_tests/, otel_tests/,
llm_responses_api_testing/, batches_tests/, spend_tracking_tests/,
litellm_utils_tests/, unified_google_tests/, and a few top-level
tests/test_*.py files where the model literal is a fixture or
placeholder (router model_list, mock standard logging payload, mock
callback data) rather than the test's subject.

Mappings applied (see scope notes below):
- gpt-3.5-turbo -> gpt-5-mini
- gpt-4 (bare) -> gpt-5.5
- gpt-4o (bare) -> gpt-5.5 (corrected from initial gpt-5 — bare gpt-5
  is not a valid OpenAI alias; only gpt-5.5 / gpt-5.4 / gpt-5.2-codex
  / gpt-5-mini exist)
- gpt-4o-mini (bare) -> gpt-5-mini
- text-embedding-ada-002 -> text-embedding-3-small
- claude-3-sonnet-20240229 -> claude-sonnet-4-5-20250929
- claude-3-opus-20240229 -> claude-opus-4-7
- claude-3-haiku-20240307 -> claude-haiku-4-5-20251001
- claude-3-5-sonnet-20240620/20241022 -> claude-sonnet-4-5-20250929
- claude-3-7-sonnet-20250219 -> claude-sonnet-4-6
- gemini-1.5-flash -> gemini-2.5-flash
- gemini-1.5-pro -> gemini-2.5-pro

Explicitly preserved (not modernized):
- llm_translation/ tests where model is the SUBJECT (provider-specific
  translation/transformation logic). Only the deprecated 20250514
  references were already bumped in a prior commit.
- Cost-calc / tokenizer subject tests in test_utils.py (skip-ranges
  documented by the sub-agent).
- Bedrock model IDs in test_health_check.py path-stripping tests.
- JSONL batch request bodies and mock LLM response bodies (must match
  upstream literal).
- Langfuse expected-request-body JSON fixtures (cost values are exact-
  match-asserted; changing the model would shift response_cost).
- gpt-3.5-turbo-instruct (text-completion endpoint; no modern OpenAI
  equivalent).
- Top-level tests calling the proxy through user-facing aliases
  (gpt-3.5-turbo, gpt-4, text-embedding-ada-002, dall-e-3) — aliases
  in proxy_server_config.yaml stay; only the underlying model was
  bumped.
- tests/test_gpt5_azure_temperature_support.py (the test's whole point
  is model-name handling).
- Fake / mock / openai/fake identifiers.

Notable side fixes:
- test_spend_accuracy_tests.py: UPSTREAM_MODEL now matches what
  spend_tracking_config.yaml's proxy actually routes to (gpt-5-mini),
  resolving a latent inconsistency.
- proxy_server_config.yaml: bare `gpt-5` alias renamed to `gpt-5.5`
  (bare gpt-5 is not a valid OpenAI alias).
- test_batches_logging_unit_tests.py: explicit_models list entries
  kept distinct (gpt-5-mini + gpt-5.5) after bulk rename.

* test: fix CI failures from model modernization sweep

CI surfaced 4 categories of regression from the bulk modernization:

1. Azure deployment names are customer-specific. Reverted:
   - tests/litellm_utils_tests/test_health_check.py: azure/text-
     embedding-3-small -> azure/text-embedding-ada-002 (the CI Azure
     account does not have a text-embedding-3-small deployment).
   - tests/logging_callback_tests/test_custom_callback_router.py:
     same revert for two router fixtures driving aembedding.

2. gpt-5 family does not accept temperature != 1. Tests that pass a
   custom temperature swapped from gpt-5-mini to gpt-4.1-mini (modern
   non-reasoning OpenAI mini that still accepts temperature/logprobs):
   - tests/logging_callback_tests/test_datadog.py
   - tests/logging_callback_tests/test_langsmith_unit_test.py
   - tests/logging_callback_tests/test_otel_logging.py

3. proxy_server_config.yaml's gpt-3.5-turbo-large alias was routing to
   gpt-5.5 (a reasoning model that rejects logprobs). The proxy test
   tests/test_openai_endpoints.py::test_chat_completion_streaming
   exercises logprobs/top_logprobs through that alias. Bumped the
   underlying model to gpt-4.1 (non-reasoning, still modern).

4. tests/logging_callback_tests/test_gcs_pub_sub.py asserts against a
   pinned JSON fixture (gcs_pub_sub_body/spend_logs_payload.json) with
   hardcoded model="gpt-4o" and a model-specific spend value. Reverted
   the litellm.acompletion calls in the test to model="gpt-4o" so the
   fixture's exact-match assertions still hold.

5. tests/pass_through_unit_tests/test_anthropic_messages_passthrough.py:
   anthropic.messages.create routing to openai/gpt-5-mini returned an
   empty content[0] with max_tokens=100 (reasoning-token consumption).
   Swapped to openai/gpt-4.1-mini.

* test: fix Assistants API model + 2 cursor[bot] review nits

1. pass_through_unit_tests/test_custom_logger_passthrough.py: gpt-5.5
   isn't accepted by the /v1/assistants endpoint
   ("unsupported_model"). Switch to gpt-4.1-mini (modern, Assistants-
   API-supported, non-reasoning).

2. example_config_yaml/pass_through_config.yaml: the previous sweep
   bumped the claude-3-7-sonnet alias to claude-opus-4-7, which is a
   tier change (Sonnet -> Opus). Map to claude-sonnet-4-6 to keep the
   Sonnet tier intact. (Cursor bugbot review.)

3. example_config_yaml/simple_config.yaml: model_name was left as
   gpt-3.5-turbo while the underlying was bumped to gpt-5-mini, which
   muddles the "simple" example. Make both sides gpt-5-mini so the
   most basic example is a straight 1:1 mapping again. (Cursor bugbot
   review.)

* fix: revert gpt-4/gpt-3.5-turbo alias underlying to non-reasoning models

tests/test_openai_endpoints.py::test_completion calls the proxy alias
"gpt-4" with temperature=0, and other tests call gpt-3.5-turbo with
custom temperature / logprobs / the legacy /v1/completions endpoint.
The earlier modernization mapped both aliases to gpt-5.5 / gpt-5-mini,
which are reasoning models that reject temperature != 1 and don't
expose /v1/completions. Map the aliases to gpt-4.1 / gpt-4.1-mini
(modern non-reasoning OpenAI models) instead — keeps user-facing
aliases preserved while picking a current underlying that still
supports the parameters/endpoints the tests exercise.
2026-05-15 15:44:28 -07:00

845 lines
34 KiB
Python

import httpx
import json
import pytest
import sys
from typing import Any, Dict, List, Optional
from unittest.mock import MagicMock, Mock, patch
import os
from litellm._uuid import uuid
import time
import base64
sys.path.insert(
0, os.path.abspath("../..")
) # Adds the parent directory to the system path
import litellm
from abc import ABC, abstractmethod
from litellm.integrations.custom_logger import CustomLogger
import json
from litellm.types.utils import StandardLoggingPayload
from litellm.types.llms.openai import (
ResponseCompletedEvent,
ResponsesAPIResponse,
ResponseAPIUsage,
IncompleteDetails,
)
from openai.types.responses.response_create_params import (
ResponseInputParam,
)
from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler
def validate_responses_api_response(response, final_chunk: bool = False):
"""
Validate that a response from litellm.responses() or litellm.aresponses()
conforms to the expected ResponsesAPIResponse structure.
Args:
response: The response object to validate
Raises:
AssertionError: If the response doesn't match the expected structure
"""
# Validate response structure
print("response=", json.dumps(response, indent=4, default=str))
assert isinstance(
response, ResponsesAPIResponse
), "Response should be an instance of ResponsesAPIResponse"
# Required fields
assert "id" in response and isinstance(
response["id"], str
), "Response should have a string 'id' field"
assert "created_at" in response and isinstance(
response["created_at"], int
), "Response should have an integer 'created_at' field"
if response.get("status") == "completed":
assert "output" in response and isinstance(
response["output"], list
), "Response should have a list 'output' field"
# Optional fields with their expected types
optional_fields = {
"error": (dict, type(None)), # error can be dict or None
"incomplete_details": (IncompleteDetails, type(None)),
"instructions": (str, type(None)),
"metadata": dict,
"model": str,
"object": str,
"parallel_tool_calls": (bool, type(None)),
"temperature": (int, float, type(None)),
"tool_choice": (dict, str, type(None)),
"tools": (list, type(None)),
"top_p": (int, float, type(None)),
"max_output_tokens": (int, type(None)),
"previous_response_id": (str, type(None)),
"reasoning": (dict, type(None)),
"status": str,
"text": dict,
"truncation": (str, type(None)),
"usage": ResponseAPIUsage,
"user": (str, type(None)),
"store": (bool, type(None)),
}
if final_chunk is False:
optional_fields["usage"] = type(None)
for field, expected_type in optional_fields.items():
if field in response:
assert isinstance(
response[field], expected_type
), f"Field '{field}' should be of type {expected_type}, but got {type(response[field])}"
# Check if output has at least one item
if final_chunk is True and response.get("status") == "completed":
assert (
len(response["output"]) > 0
), "Response 'output' field should have at least one item"
return True # Return True if validation passes
class BaseResponsesAPITest(ABC):
"""
Abstract base test class that enforces a common test across all test classes.
"""
@abstractmethod
def get_base_completion_call_args(self) -> dict:
"""Must return the base completion call args"""
pass
def get_base_completion_reasoning_call_args(self) -> dict:
"""Must return the base completion reasoning call args"""
return None
def get_advanced_model_for_shell_tool(self) -> Optional[str]:
"""If specified, overrides the model used by test_responses_api_shell_tool_streaming_sees_shell_output (e.g. openai/gpt-5.2 for shell support)."""
return None
@pytest.mark.parametrize("sync_mode", [True, False])
@pytest.mark.asyncio
async def test_basic_openai_responses_api(self, sync_mode):
litellm._turn_on_debug()
litellm.set_verbose = True
base_completion_call_args = self.get_base_completion_call_args()
try:
if sync_mode:
response = litellm.responses(
input="Basic ping",
max_output_tokens=20,
**base_completion_call_args,
)
else:
response = await litellm.aresponses(
input="Basic ping",
max_output_tokens=20,
**base_completion_call_args,
)
except litellm.InternalServerError:
pytest.skip("Skipping test due to litellm.InternalServerError")
print("litellm response=", json.dumps(response, indent=4, default=str))
# Use the helper function to validate the response
validate_responses_api_response(response, final_chunk=True)
@pytest.mark.parametrize("sync_mode", [True, False])
@pytest.mark.asyncio
@pytest.mark.flaky(retries=3, delay=2)
async def test_basic_openai_responses_api_streaming(self, sync_mode):
litellm._turn_on_debug()
# Enable cost calculation for streaming usage
litellm.include_cost_in_streaming_usage = True
base_completion_call_args = self.get_base_completion_call_args()
collected_content_string = ""
response_completed_event = None
if sync_mode:
response = litellm.responses(
input="Basic ping", stream=True, **base_completion_call_args
)
for event in response:
print("litellm response=", json.dumps(event, indent=4, default=str))
if event.type == "response.output_text.delta":
collected_content_string += event.delta
elif event.type == "response.completed":
response_completed_event = event
else:
response = await litellm.aresponses(
input="Basic ping", stream=True, **base_completion_call_args
)
async for event in response:
print("litellm response=", json.dumps(event, indent=4, default=str))
if event.type == "response.output_text.delta":
collected_content_string += event.delta
elif event.type == "response.completed":
response_completed_event = event
# assert the response completed event is not None
assert response_completed_event is not None
# assert the response completed event has a response
assert response_completed_event.response is not None
# For async agent APIs (like Manus), the response may be in 'running' state
# without content yet - this is valid behavior
response_status = response_completed_event.response.status
if response_status in ["running", "pending"]:
# Running/pending state is acceptable - task started successfully
print(
f"Response is in '{response_status}' state - async agent API behavior"
)
assert response_completed_event.response.id is not None
else:
# For completed responses, validate content and usage
# assert the delta chunks content had len(collected_content_string) > 0
# this content is typically rendered on chat ui's
assert len(collected_content_string) > 0
# assert the response completed event includes the usage
assert response_completed_event.response.usage is not None
# basic test assert the usage seems reasonable
print(
"response_completed_event.response.usage=",
response_completed_event.response.usage,
)
assert (
response_completed_event.response.usage.input_tokens > 0
and response_completed_event.response.usage.input_tokens < 100
)
assert (
response_completed_event.response.usage.output_tokens > 0
and response_completed_event.response.usage.output_tokens < 2000
)
assert (
response_completed_event.response.usage.total_tokens > 0
and response_completed_event.response.usage.total_tokens < 2000
)
# total tokens should be the sum of input and output tokens
assert (
response_completed_event.response.usage.total_tokens
== response_completed_event.response.usage.input_tokens
+ response_completed_event.response.usage.output_tokens
)
# assert the response completed event includes cost when include_cost_in_streaming_usage is True
assert hasattr(
response_completed_event.response.usage, "cost"
), "Cost should be included in streaming responses API usage object"
assert (
response_completed_event.response.usage.cost > 0
), "Cost should be greater than 0"
print(
f"Cost found in streaming response: {response_completed_event.response.usage.cost}"
)
# Reset the setting
litellm.include_cost_in_streaming_usage = False
@pytest.mark.parametrize("sync_mode", [False, True])
@pytest.mark.asyncio
async def test_basic_openai_responses_delete_endpoint(self, sync_mode):
litellm._turn_on_debug()
litellm.set_verbose = True
base_completion_call_args = self.get_base_completion_call_args()
if sync_mode:
response = litellm.responses(
input="Basic ping", max_output_tokens=20, **base_completion_call_args
)
# delete the response
if isinstance(response, ResponsesAPIResponse):
litellm.delete_responses(
response_id=response.id, **base_completion_call_args
)
else:
raise ValueError("response is not a ResponsesAPIResponse")
else:
response = await litellm.aresponses(
input="Basic ping", max_output_tokens=20, **base_completion_call_args
)
# async delete the response
if isinstance(response, ResponsesAPIResponse):
await litellm.adelete_responses(
response_id=response.id, **base_completion_call_args
)
else:
raise ValueError("response is not a ResponsesAPIResponse")
@pytest.mark.parametrize("sync_mode", [True, False])
@pytest.mark.flaky(retries=3, delay=2)
@pytest.mark.asyncio
async def test_basic_openai_responses_streaming_delete_endpoint(self, sync_mode):
# litellm._turn_on_debug()
# litellm.set_verbose = True
base_completion_call_args = self.get_base_completion_call_args()
response_id = None
if sync_mode:
response_id = None
response = litellm.responses(
input="Basic ping",
max_output_tokens=20,
stream=True,
**base_completion_call_args,
)
for event in response:
print("litellm response=", json.dumps(event, indent=4, default=str))
if "response" in event:
response_obj = event.get("response")
if response_obj is not None:
response_id = response_obj.get("id")
print("got response_id=", response_id)
# delete the response
assert response_id is not None
litellm.delete_responses(
response_id=response_id, **base_completion_call_args
)
else:
response = await litellm.aresponses(
input="Basic ping",
max_output_tokens=20,
stream=True,
**base_completion_call_args,
)
async for event in response:
print("litellm response=", json.dumps(event, indent=4, default=str))
if "response" in event:
response_obj = event.get("response")
if response_obj is not None:
response_id = response_obj.get("id")
print("got response_id=", response_id)
# delete the response
assert response_id is not None
await litellm.adelete_responses(
response_id=response_id, **base_completion_call_args
)
@pytest.mark.parametrize("sync_mode", [False, True])
@pytest.mark.flaky(retries=3, delay=2)
@pytest.mark.asyncio
async def test_basic_openai_responses_get_endpoint(self, sync_mode):
litellm._turn_on_debug()
litellm.set_verbose = True
base_completion_call_args = self.get_base_completion_call_args()
if sync_mode:
response = litellm.responses(
input="Basic ping", max_output_tokens=20, **base_completion_call_args
)
# get the response
if isinstance(response, ResponsesAPIResponse):
result = litellm.get_responses(
response_id=response.id, **base_completion_call_args
)
assert result is not None
assert result.id == response.id
assert result.output == response.output
else:
raise ValueError("response is not a ResponsesAPIResponse")
else:
response = await litellm.aresponses(
input="Basic ping", max_output_tokens=20, **base_completion_call_args
)
# async get the response
if isinstance(response, ResponsesAPIResponse):
result = await litellm.aget_responses(
response_id=response.id, **base_completion_call_args
)
assert result is not None
assert result.id == response.id
assert result.output == response.output
else:
raise ValueError("response is not a ResponsesAPIResponse")
@pytest.mark.asyncio
@pytest.mark.flaky(retries=3, delay=2)
async def test_basic_openai_list_input_items_endpoint(self):
"""Test that calls the OpenAI List Input Items endpoint"""
litellm._turn_on_debug()
response = await litellm.aresponses(
model="gpt-5.5",
input="Tell me a three sentence bedtime story about a unicorn.",
)
print("Initial response=", json.dumps(response, indent=4, default=str))
response_id = response.get("id")
assert response_id is not None, "Response should have an ID"
print(f"Got response_id: {response_id}")
list_items_response = await litellm.alist_input_items(
response_id=response_id,
limit=20,
order="desc",
)
print(
"List items response=",
json.dumps(list_items_response, indent=4, default=str),
)
@pytest.mark.asyncio
async def test_multiturn_responses_api(self):
litellm._turn_on_debug()
litellm.set_verbose = True
try:
base_completion_call_args = self.get_base_completion_call_args()
response_1 = await litellm.aresponses(
input="Basic ping", max_output_tokens=20, **base_completion_call_args
)
# follow up with a second request
response_1_id = response_1.id
response_2 = await litellm.aresponses(
input="Basic ping",
max_output_tokens=20,
previous_response_id=response_1_id,
**base_completion_call_args,
)
# assert the response is not None
assert response_1 is not None
assert response_2 is not None
except litellm.InternalServerError:
pytest.skip("Skipping test due to litellm.InternalServerError")
@pytest.mark.asyncio
async def test_responses_api_with_tool_calls(self):
"""Test that calls the Responses API with tool calls including function call and output"""
litellm._turn_on_debug()
litellm.set_verbose = True
base_completion_call_args = self.get_base_completion_call_args()
# Define the input with message, function call, and function call output
input_data: ResponseInputParam = [
{
"type": "message",
"role": "user",
"content": "How is the weather in São Paulo today ?",
},
{
"type": "function_call",
"arguments": '{"location": "São Paulo, Brazil"}',
"call_id": "fc_1fe70e2a-a596-45ef-b72c-9b8567c460e5",
"name": "get_weather",
"id": "fc_1fe70e2a-a596-45ef-b72c-9b8567c460e5",
"status": "completed",
},
{
"type": "function_call_output",
"call_id": "fc_1fe70e2a-a596-45ef-b72c-9b8567c460e5",
"output": "Rainy",
},
]
# Define the tools
tools = [
{
"type": "function",
"name": "get_weather",
"description": "Get current temperature for a given location.",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "City and country e.g. Bogotá, Colombia",
}
},
"required": ["location"],
"additionalProperties": False,
},
}
]
try:
# Make the responses API call
response = await litellm.aresponses(
input=input_data, store=False, tools=tools, **base_completion_call_args
)
except litellm.InternalServerError:
pytest.skip("Skipping test due to litellm.InternalServerError")
print("litellm response=", json.dumps(response, indent=4, default=str))
# Validate the response structure
validate_responses_api_response(response, final_chunk=True)
# Additional assertions specific to tool calls
assert response is not None
assert "output" in response
# For async agent APIs (like Manus), the response may be in 'running' state
# without output yet - this is valid behavior
if response.get("status") in ["running", "pending"]:
print(
f"Response is in '{response.get('status')}' state - async agent API behavior"
)
assert response.get("id") is not None
else:
assert len(response["output"]) > 0
@pytest.mark.asyncio
async def test_responses_api_multi_turn_with_reasoning_and_structured_output(self):
"""
Test multi-turn conversation with reasoning, structured output, and tool calls.
This test validates:
- First call: Model uses reasoning to process a question and makes a tool call
- Tool call handling: Function call output is properly processed
- Second call: Model produces structured output incorporating tool results
- Structured output: Response conforms to defined Pydantic model schema
"""
from pydantic import BaseModel
litellm._turn_on_debug()
litellm.set_verbose = True
base_completion_call_args = self.get_base_completion_reasoning_call_args()
if base_completion_call_args is None:
pytest.skip("Skipping test due to no base completion reasoning call args")
# Define tools for the conversation
tools = [{"type": "function", "name": "get_today"}]
# Define structured output schema
class Output(BaseModel):
today: str
number_of_r: str
# Initial conversation input
input_messages = [
{
"role": "user",
"content": "How many r in strrawberrry? While you're thinking, you should call tool get_today. Then you output the today and number of r",
}
]
# First call - should trigger reasoning and tool call
response = await litellm.aresponses(
input=input_messages,
tools=tools,
reasoning={"effort": "low", "summary": "detailed"},
text_format=Output,
**base_completion_call_args,
)
print("First call output:")
print(json.dumps(response.output, indent=4, default=str))
# Validate first response structure
validate_responses_api_response(response, final_chunk=True)
assert response.output is not None
assert len(response.output) > 0
# Extend input with first response output
input_messages.extend(response.output)
# Process any tool calls and add function outputs
function_outputs = []
for item in response.output:
if hasattr(item, "type") and item.type in [
"function_call",
"custom_tool_call",
]:
if hasattr(item, "name") and item.name == "get_today":
function_outputs.append(
{
"type": "function_call_output",
"call_id": item.call_id,
"output": "2025-01-15",
}
)
# Add function outputs to conversation
input_messages.extend(function_outputs)
print("Second call input:")
print(json.dumps(input_messages, indent=4, default=str))
# Second call - should produce structured output
final_response = await litellm.aresponses(
input=input_messages,
tools=tools,
reasoning={"effort": "low", "summary": "detailed"},
text_format=Output,
**base_completion_call_args,
)
print("Second call output:")
print(json.dumps(final_response.output, indent=4, default=str))
# Validate final response structure
validate_responses_api_response(final_response, final_chunk=True)
assert final_response.output is not None
def test_openai_responses_api_dict_input_filtering(self):
"""
Test that regular dict inputs with status fields are properly filtered
to replicate exclude_unset=True behavior for non-Pydantic objects.
"""
from litellm.llms.openai.responses.transformation import (
OpenAIResponsesAPIConfig,
)
# Test input with regular dict objects (like from JSON)
test_input = [
{"role": "user", "content": "test"},
{
"id": "rs_123",
"summary": [{"text": "test", "type": "summary_text"}],
"type": "reasoning",
"content": None, # Should be filtered out
"encrypted_content": None, # Should be filtered out
"status": None, # Should be filtered out
},
{
"arguments": "{}",
"call_id": "call_123",
"name": "get_today",
"type": "function_call",
"id": "fc_123",
"status": "completed", # Should be preserved (not a default field)
},
]
config = OpenAIResponsesAPIConfig()
validated_input = config._validate_input_param(test_input)
# Verify the results
assert len(validated_input) == 3
# Check reasoning item (index 1)
reasoning_item = validated_input[1]
assert reasoning_item["type"] == "reasoning"
assert (
"status" not in reasoning_item
), "status field should be filtered out from reasoning item"
assert (
"content" not in reasoning_item
), "content field should be filtered out from reasoning item"
assert (
"encrypted_content" not in reasoning_item
), "encrypted_content field should be filtered out from reasoning item"
# Note: ID auto-generation was disabled, so reasoning items may not have IDs
# Only check for ID if it was present in the original input
if "id" in reasoning_item:
assert reasoning_item["id"] == "rs_123", "ID should be preserved if present"
assert "summary" in reasoning_item, "summary field should be preserved"
# Check function call item (index 2)
function_call_item = validated_input[2]
assert function_call_item["type"] == "function_call"
assert (
"status" in function_call_item
), "status field should be preserved in function call item"
assert (
function_call_item["status"] == "completed"
), "status value should be preserved"
print("✅ OpenAI Responses API dict input filtering test passed")
@pytest.mark.parametrize("sync_mode", [False, True])
@pytest.mark.flaky(retries=3, delay=2)
@pytest.mark.asyncio
async def test_basic_openai_responses_cancel_endpoint(self, sync_mode):
try:
litellm._turn_on_debug()
litellm.set_verbose = True
base_completion_call_args = self.get_base_completion_call_args()
if sync_mode:
response = litellm.responses(
input="Basic ping",
max_output_tokens=20,
background=True,
**base_completion_call_args,
)
# cancel the response
if isinstance(response, ResponsesAPIResponse):
cancel_result = litellm.cancel_responses(
response_id=response.id, **base_completion_call_args
)
assert cancel_result is not None
assert hasattr(cancel_result, "id")
# The actual response structure depends on the provider implementation
assert isinstance(cancel_result, ResponsesAPIResponse)
else:
raise ValueError("response is not a ResponsesAPIResponse")
else:
response = await litellm.aresponses(
input="Basic ping",
max_output_tokens=20,
background=True,
**base_completion_call_args,
)
# async cancel the response
if isinstance(response, ResponsesAPIResponse):
cancel_result = await litellm.acancel_responses(
response_id=response.id, **base_completion_call_args
)
assert cancel_result is not None
assert hasattr(cancel_result, "id")
# The actual response structure depends on the provider implementation
assert isinstance(cancel_result, ResponsesAPIResponse)
else:
raise ValueError("response is not a ResponsesAPIResponse")
except Exception as e:
if "Cannot cancel a completed response" in str(e):
pass
else:
raise e
@pytest.mark.parametrize("sync_mode", [False, True])
@pytest.mark.asyncio
async def test_cancel_responses_invalid_response_id(self, sync_mode):
"""Test cancel_responses with invalid response ID should raise appropriate error"""
base_completion_call_args = self.get_base_completion_call_args()
if sync_mode:
with pytest.raises(Exception):
litellm.cancel_responses(
response_id="invalid_response_id_12345", **base_completion_call_args
)
else:
with pytest.raises(Exception):
await litellm.acancel_responses(
response_id="invalid_response_id_12345", **base_completion_call_args
)
@pytest.mark.asyncio
async def test_responses_api_context_management_server_side_compaction(self):
"""
E2E test for server-side compaction (context_management) on OpenAI Responses API.
Passes context_management with compact_threshold; validates that the request is
accepted and returns a valid response. Compaction may not run for short inputs.
"""
base_completion_call_args = self.get_base_completion_call_args()
model = base_completion_call_args.get("model") or ""
# Azure does not support compaction context_management (only clear_tool_results)
if "azure/" in str(model):
pytest.skip("context_management compaction is not supported on Azure")
if "openai/" not in str(model):
pytest.skip(
"context_management server-side compaction e2e is only run for OpenAI"
)
context_management = [{"type": "compaction", "compact_threshold": 200000}]
try:
response = await litellm.aresponses(
input="Short ping to verify context_management is accepted.",
max_output_tokens=20,
context_management=context_management,
**base_completion_call_args,
)
except litellm.InternalServerError:
pytest.skip("Skipping test due to litellm.InternalServerError")
validate_responses_api_response(response, final_chunk=True)
assert response.get("id") is not None
assert response.get("status") is not None
@pytest.mark.asyncio
async def test_responses_api_shell_tool(self):
"""
E2E test for Shell tool on OpenAI Responses API.
Passes tools=[{"type": "shell", "environment": {"type": "container_auto"}}];
validates that the request is accepted and returns a valid response.
Only runs for OpenAI/Azure (Responses API with shell support).
"""
base_completion_call_args = self.get_base_completion_call_args()
model = (
self.get_advanced_model_for_shell_tool()
or base_completion_call_args.get("model")
or ""
)
if "openai/" not in str(model) and "azure/" not in str(model):
pytest.skip("Shell tool e2e is only run for OpenAI/Azure Responses API")
tools = [{"type": "shell", "environment": {"type": "container_auto"}}]
input_msg = "List files in /mnt/data and show python --version."
try:
response = await litellm.aresponses(
**{**base_completion_call_args, "model": model},
input=input_msg,
max_output_tokens=256,
tools=tools,
tool_choice="auto",
)
except litellm.InternalServerError:
pytest.skip("Skipping test due to litellm.InternalServerError")
except litellm.BadRequestError as e:
if "shell" in str(e).lower() and "not supported" in str(e).lower():
pytest.skip(
"Shell tool is not supported for this model (e.g. gpt-5.5); use a model that supports shell"
)
raise
validate_responses_api_response(response, final_chunk=True)
assert response.get("id") is not None
assert response.get("status") is not None
@pytest.mark.asyncio
async def test_responses_api_shell_tool_streaming_sees_shell_output(self):
"""
E2E streaming call with Shell tool; validate we can see shell output in the stream.
Calls aresponses(..., tools=[shell], stream=True), then iterates the stream and
asserts at least one event is shell-related or response output contains shell_call.
Skips when model does not support shell (e.g. gpt-5.5).
"""
base_completion_call_args = self.get_base_completion_call_args()
model = (
self.get_advanced_model_for_shell_tool()
or base_completion_call_args.get("model")
or "openai/gpt-5.2"
)
if "openai/" not in str(model):
pytest.skip(
"Shell tool streaming e2e is only run for OpenAI/Azure Responses API"
)
tools = [{"type": "shell", "environment": {"type": "container_auto"}}]
input_msg = "List files in /mnt/data and run python --version."
stream = await litellm.aresponses(
**{**base_completion_call_args, "model": model},
input=input_msg,
max_output_tokens=512,
tools=tools,
tool_choice="auto",
stream=True,
)
event_types_seen = []
output_items_with_shell = []
async for event in stream:
print("event=", json.dumps(event, indent=4, default=str))
event_type = getattr(event, "type", None) or (
event.get("type") if isinstance(event, dict) else None
)
if event_type is not None:
event_types_seen.append(str(event_type))
if "shell" in str(event_type or "").lower():
output_items_with_shell.append(event_type)
response_obj = getattr(event, "response", None) or (
event.get("response") if isinstance(event, dict) else None
)
if response_obj is not None:
output = getattr(response_obj, "output", None) or (
response_obj.get("output")
if isinstance(response_obj, dict)
else None
)
if isinstance(output, list):
for item in output:
item_type = getattr(item, "type", None) or (
item.get("type") if isinstance(item, dict) else None
)
if item_type and "shell" in str(item_type).lower():
output_items_with_shell.append(item_type)
assert len(event_types_seen) > 0, "Expected at least one stream event"
assert (
len(output_items_with_shell) > 0
), f"Expected to see shell output in stream; event types seen: {event_types_seen!r}"