biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Refactor] Relocate chat completion and anthropic tests (#36919)

Browse Source 
Signed-off-by: sfeng33 <4florafeng@gmail.com>
This commit is contained in:
Flora Feng
2026-03-14 00:16:16 -04:00
committed by GitHub
parent 236de72e49
commit bcfdadb1bc
14 changed files with 14 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

0
tests/entrypoints/openai/chat_completion/__init__.py Normal file
View File

3
tests/entrypoints/openai/test_chat.py → tests/entrypoints/openai/chat_completion/test_chat.py
View File

@@ -14,13 +14,12 @@ import requests
import torch
from openai import BadRequestError
from tests.utils import RemoteOpenAIServer
from vllm.entrypoints.openai.chat_completion.protocol import (
ChatCompletionRequest,
)
from vllm.sampling_params import SamplingParams
from ...utils import RemoteOpenAIServer
# any model with a chat template should work here
MODEL_NAME = "HuggingFaceH4/zephyr-7b-beta"

3
tests/entrypoints/openai/test_chat_echo.py → tests/entrypoints/openai/chat_completion/test_chat_echo.py
View File

@@ -7,10 +7,9 @@ import openai # use the official client for correctness check
import pytest
import pytest_asyncio
from tests.utils import RemoteOpenAIServer
from vllm.config import ModelConfig
from ...utils import RemoteOpenAIServer
# # any model with a chat template should work here
MODEL_NAME = "Qwen/Qwen2-1.5B-Instruct"

0
tests/entrypoints/openai/test_chat_error.py → tests/entrypoints/openai/chat_completion/test_chat_error.py
View File

3
tests/entrypoints/openai/test_chat_logit_bias_validation.py → tests/entrypoints/openai/chat_completion/test_chat_logit_bias_validation.py
View File

@@ -5,10 +5,9 @@ import openai
import pytest
import pytest_asyncio
from tests.utils import RemoteOpenAIServer
from vllm.config import ModelConfig
from ...utils import RemoteOpenAIServer
MODEL_NAME = "Qwen/Qwen2.5-1.5B-Instruct"

2
tests/entrypoints/openai/test_chat_with_tool_reasoning.py → tests/entrypoints/openai/chat_completion/test_chat_with_tool_reasoning.py
View File

@@ -5,7 +5,7 @@ import openai # use the official client for correctness check
import pytest
import pytest_asyncio
from ...utils import RemoteOpenAIServer
from tests.utils import RemoteOpenAIServer
# a reasoning and tool calling model
MODEL_NAME = "Qwen/QwQ-32B"

2
tests/entrypoints/openai/test_completion_with_function_calling.py → tests/entrypoints/openai/chat_completion/test_completion_with_function_calling.py
View File

@@ -10,7 +10,7 @@ import pytest
import pytest_asyncio
# downloading lora to test lora requests
from ...utils import ROCM_ENV_OVERRIDES, ROCM_EXTRA_ARGS, RemoteOpenAIServer
from tests.utils import ROCM_ENV_OVERRIDES, ROCM_EXTRA_ARGS, RemoteOpenAIServer
# any model with a chat template should work here
MODEL_NAME = "Qwen/Qwen3-0.6B"

2
tests/entrypoints/openai/test_enable_force_include_usage.py → tests/entrypoints/openai/chat_completion/test_enable_force_include_usage.py
View File

@@ -4,7 +4,7 @@ import openai
import pytest
import pytest_asyncio
from ...utils import RemoteOpenAIServer
from tests.utils import RemoteOpenAIServer
@pytest.fixture(scope="module")

13
tests/entrypoints/openai/test_serving_chat.py → tests/entrypoints/openai/chat_completion/test_serving_chat.py
View File

@@ -10,6 +10,12 @@ import pytest
import pytest_asyncio
from openai import OpenAI
from tests.entrypoints.openai.utils import (
accumulate_streaming_response,
verify_chat_response,
verify_harmony_messages,
)
from tests.utils import RemoteOpenAIServer
from vllm._aiter_ops import is_aiter_found_and_supported
from vllm.config import MultiModalConfig
from vllm.entrypoints.openai.chat_completion.protocol import (
@@ -39,13 +45,6 @@ from vllm.tokenizers.registry import tokenizer_args_from_config
from vllm.tool_parsers import ToolParserManager
from vllm.v1.engine.async_llm import AsyncLLM
from ...utils import RemoteOpenAIServer
from .utils import (
accumulate_streaming_response,
verify_chat_response,
verify_harmony_messages,
)
GPT_OSS_MODEL_NAME = "openai/gpt-oss-20b"
GPT_OSS_SPECULATOR_NAME = "RedHatAI/gpt-oss-20b-speculator.eagle3"

0
tests/entrypoints/openai/test_serving_chat_stream_harmony.py → tests/entrypoints/openai/chat_completion/test_serving_chat_stream_harmony.py
View File

375
tests/entrypoints/openai/test_anthropic_messages_conversion.py
View File

@@ -1,375 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
"""Unit tests for Anthropic-to-OpenAI request conversion.
Tests the image source handling and tool_result content parsing in
AnthropicServingMessages._convert_anthropic_to_openai_request().
"""
from vllm.entrypoints.anthropic.protocol import (
AnthropicMessagesRequest,
)
from vllm.entrypoints.anthropic.serving import AnthropicServingMessages
_convert = AnthropicServingMessages._convert_anthropic_to_openai_request
_img_url = AnthropicServingMessages._convert_image_source_to_url
def _make_request(
messages: list[dict],
**kwargs,
) -> AnthropicMessagesRequest:
return AnthropicMessagesRequest(
model="test-model",
max_tokens=128,
messages=messages,
**kwargs,
)
# ======================================================================
# _convert_image_source_to_url
# ======================================================================
class TestConvertImageSourceToUrl:
def test_base64_source(self):
source = {
"type": "base64",
"media_type": "image/jpeg",
"data": "iVBORw0KGgo=",
}
assert _img_url(source) == "data:image/jpeg;base64,iVBORw0KGgo="
def test_base64_png(self):
source = {
"type": "base64",
"media_type": "image/png",
"data": "AAAA",
}
assert _img_url(source) == "data:image/png;base64,AAAA"
def test_url_source(self):
source = {
"type": "url",
"url": "https://example.com/image.jpg",
}
assert _img_url(source) == "https://example.com/image.jpg"
def test_missing_type_defaults_to_base64(self):
"""When 'type' is absent, treat as base64."""
source = {
"media_type": "image/webp",
"data": "UklGR",
}
assert _img_url(source) == "data:image/webp;base64,UklGR"
def test_missing_media_type_defaults_to_jpeg(self):
source = {"type": "base64", "data": "abc123"}
assert _img_url(source) == "data:image/jpeg;base64,abc123"
def test_url_source_missing_url_returns_empty(self):
source = {"type": "url"}
assert _img_url(source) == ""
def test_empty_source_returns_data_uri_shell(self):
source: dict = {}
assert _img_url(source) == "data:image/jpeg;base64,"
# ======================================================================
# Image blocks inside user messages
# ======================================================================
class TestImageContentBlocks:
def test_base64_image_in_user_message(self):
request = _make_request(
[
{
"role": "user",
"content": [
{"type": "text", "text": "Describe this image"},
{
"type": "image",
"source": {
"type": "base64",
"media_type": "image/jpeg",
"data": "iVBORw0KGgo=",
},
},
],
}
]
)
result = _convert(request)
user_msg = result.messages[0]
assert user_msg["role"] == "user"
parts = user_msg["content"]
assert len(parts) == 2
assert parts[0] == {"type": "text", "text": "Describe this image"}
assert parts[1] == {
"type": "image_url",
"image_url": {"url": "data:image/jpeg;base64,iVBORw0KGgo="},
}
def test_url_image_in_user_message(self):
request = _make_request(
[
{
"role": "user",
"content": [
{"type": "text", "text": "What is this?"},
{
"type": "image",
"source": {
"type": "url",
"url": "https://example.com/cat.png",
},
},
],
}
]
)
result = _convert(request)
parts = result.messages[0]["content"]
assert parts[1] == {
"type": "image_url",
"image_url": {"url": "https://example.com/cat.png"},
}
# ======================================================================
# tool_result content handling
# ======================================================================
class TestToolResultContent:
def _make_tool_result_request(
self, tool_result_content
) -> AnthropicMessagesRequest:
"""Build a request with assistant tool_use followed by user
tool_result."""
return _make_request(
[
{
"role": "assistant",
"content": [
{
"type": "tool_use",
"id": "call_001",
"name": "read_file",
"input": {"path": "/tmp/img.png"},
}
],
},
{
"role": "user",
"content": [
{
"type": "tool_result",
"tool_use_id": "call_001",
"content": tool_result_content,
}
],
},
]
)
def test_tool_result_string_content(self):
request = self._make_tool_result_request("file contents here")
result = _convert(request)
tool_msg = [m for m in result.messages if m["role"] == "tool"]
assert len(tool_msg) == 1
assert tool_msg[0]["content"] == "file contents here"
assert tool_msg[0]["tool_call_id"] == "call_001"
def test_tool_result_text_blocks(self):
request = self._make_tool_result_request(
[
{"type": "text", "text": "line 1"},
{"type": "text", "text": "line 2"},
]
)
result = _convert(request)
tool_msg = [m for m in result.messages if m["role"] == "tool"]
assert len(tool_msg) == 1
assert tool_msg[0]["content"] == "line 1\nline 2"
def test_tool_result_with_image(self):
"""Image in tool_result should produce a follow-up user message."""
request = self._make_tool_result_request(
[
{
"type": "image",
"source": {
"type": "base64",
"media_type": "image/png",
"data": "AAAA",
},
}
]
)
result = _convert(request)
tool_msg = [m for m in result.messages if m["role"] == "tool"]
assert len(tool_msg) == 1
assert tool_msg[0]["content"] == ""
# The image should be injected as a follow-up user message
follow_up = [
m
for m in result.messages
if m["role"] == "user" and isinstance(m.get("content"), list)
]
assert len(follow_up) == 1
img_parts = follow_up[0]["content"]
assert len(img_parts) == 1
assert img_parts[0] == {
"type": "image_url",
"image_url": {"url": "data:image/png;base64,AAAA"},
}
def test_tool_result_with_text_and_image(self):
"""Mixed text+image tool_result: text in tool msg, image in user
msg."""
request = self._make_tool_result_request(
[
{"type": "text", "text": "Here is the screenshot"},
{
"type": "image",
"source": {
"type": "base64",
"media_type": "image/jpeg",
"data": "QUFB",
},
},
]
)
result = _convert(request)
tool_msg = [m for m in result.messages if m["role"] == "tool"]
assert len(tool_msg) == 1
assert tool_msg[0]["content"] == "Here is the screenshot"
follow_up = [
m
for m in result.messages
if m["role"] == "user" and isinstance(m.get("content"), list)
]
assert len(follow_up) == 1
assert follow_up[0]["content"][0]["image_url"]["url"] == (
"data:image/jpeg;base64,QUFB"
)
def test_tool_result_with_multiple_images(self):
request = self._make_tool_result_request(
[
{
"type": "image",
"source": {
"type": "base64",
"media_type": "image/png",
"data": "IMG1",
},
},
{
"type": "image",
"source": {
"type": "url",
"url": "https://example.com/img2.jpg",
},
},
]
)
result = _convert(request)
follow_up = [
m
for m in result.messages
if m["role"] == "user" and isinstance(m.get("content"), list)
]
assert len(follow_up) == 1
urls = [p["image_url"]["url"] for p in follow_up[0]["content"]]
assert urls == [
"data:image/png;base64,IMG1",
"https://example.com/img2.jpg",
]
def test_tool_result_none_content(self):
request = self._make_tool_result_request(None)
result = _convert(request)
tool_msg = [m for m in result.messages if m["role"] == "tool"]
assert len(tool_msg) == 1
assert tool_msg[0]["content"] == ""
def test_tool_result_no_follow_up_when_no_images(self):
"""Ensure no extra user message is added when there are no images."""
request = self._make_tool_result_request(
[
{"type": "text", "text": "just text"},
]
)
result = _convert(request)
user_follow_ups = [
m
for m in result.messages
if m["role"] == "user" and isinstance(m.get("content"), list)
]
assert len(user_follow_ups) == 0
# ======================================================================
# Attribution header stripping
# ======================================================================
class TestAttributionHeaderStripping:
def test_billing_header_stripped_from_system(self):
"""Claude Code's x-anthropic-billing-header block should be
stripped to preserve prefix caching."""
request = _make_request(
[{"role": "user", "content": "Hello"}],
system=[
{"type": "text", "text": "You are a helpful assistant."},
{
"type": "text",
"text": "x-anthropic-billing-header: "
"cc_version=2.1.37.abc; cc_entrypoint=cli;",
},
],
)
result = _convert(request)
system_msg = result.messages[0]
assert system_msg["role"] == "system"
assert system_msg["content"] == "You are a helpful assistant."
def test_system_without_billing_header_unchanged(self):
"""Normal system blocks should pass through unchanged."""
request = _make_request(
[{"role": "user", "content": "Hello"}],
system=[
{"type": "text", "text": "You are a helpful assistant."},
{"type": "text", "text": " Be concise."},
],
)
result = _convert(request)
system_msg = result.messages[0]
assert system_msg["content"] == "You are a helpful assistant. Be concise."
def test_system_string_unchanged(self):
"""String system prompts should pass through unchanged."""
request = _make_request(
[{"role": "user", "content": "Hello"}],
system="You are a helpful assistant.",
)
result = _convert(request)
system_msg = result.messages[0]
assert system_msg["content"] == "You are a helpful assistant."