Skip to content

[BUGFIX] Parse or convert thinking chunks given content format #41822

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
juliendenize wants to merge 2 commits into
base: main
Choose a base branch
from
+377 −70
Open

[BUGFIX] Parse or convert thinking chunks given content format #41822

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
65db272
[BUGFIX] Parse or convert thinking chunks given content format
juliendenize May 6, 2026
a6ea18c
Apply comments
juliendenize May 6, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
336 changes: 279 additions & 57 deletions tests/entrypoints/test_chat_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,7 @@
parse_chat_messages,
parse_chat_messages_async,
)
from vllm.exceptions import VLLMValidationError
from vllm.inputs import MultiModalDataDict, MultiModalUUIDDict
from vllm.multimodal.utils import (
encode_audio_url,
Expand Down Expand Up @@ -2086,65 +2087,286 @@ def test_parse_chat_messages_multiple_images_interleave_with_placeholders(
)


def test_parse_chat_messages_include_thinking_chunk(mistral_model_config):
messages = [
{
"role": "system",
"content": [
{"type": "text", "text": "You are a helpful assistant."},
{
"type": "thinking",
"closed": True,
"thinking": "Only return the answer when you are confident.",
},
],
},
{"role": "user", "content": "What is 2+2?"},
{
"role": "assistant",
"content": [
{"type": "text", "text": "Let me think about it."},
{"type": "thinking", "closed": True, "thinking": "2+2 = 4"},
{
"type": "text",
"text": "The answer is 4.",
},
],
},
]

conversation_with_thinking, _, _ = parse_chat_messages(
messages,
mistral_model_config,
content_format="openai",
class TestParseChatMessagesThinking:
@pytest.mark.parametrize(
("config_fixture", "messages", "content_format", "expected"),
[
pytest.param(
"mistral_model_config",
[
{
"role": "system",
"content": [
{
"type": "text",
"text": "You are a helpful assistant.",
},
{
"type": "thinking",
"closed": True,
"thinking": "Only return the answer when you are confident.", # noqa: E501
},
],
},
{"role": "user", "content": "What is 2+2?"},
{
"role": "assistant",
"content": [
{"type": "text", "text": "Let me think about it."},
{
"type": "thinking",
"thinking": "2+2 = 4",
},
{"type": "text", "text": "The answer is 4."},
],
},
],
"openai",
[
{
"role": "system",
"content": [
{
"type": "text",
"text": "You are a helpful assistant.",
},
{
"type": "thinking",
"thinking": "Only return the answer when you are confident.", # noqa: E501
},
],
},
{
"role": "user",
"content": [{"type": "text", "text": "What is 2+2?"}],
},
{
"role": "assistant",
"content": [
{"type": "text", "text": "Let me think about it."},
{
"type": "thinking",
"thinking": "2+2 = 4",
},
{"type": "text", "text": "The answer is 4."},
],
},
],
id="mistral_thinking_preserved",
),
pytest.param(
"phi3v_model_config",
[
{
"role": "assistant",
"content": [
{"type": "text", "text": "Let me think about it."},
{
"type": "thinking",
"closed": True,
"thinking": "2+2 = 4",
},
{"type": "text", "text": "The answer is 4."},
],
},
],
"openai",
[
{
"role": "assistant",
"content": [
{"type": "text", "text": "Let me think about it."},
{
"type": "thinking",
"thinking": "2+2 = 4",
},
{"type": "text", "text": "The answer is 4."},
],
},
],
id="openai_thinking_preserved",
),
pytest.param(
"phi3v_model_config",
[
{
"role": "system",
"content": [
{
"type": "text",
"text": "You are a helpful assistant.",
},
{
"type": "thinking",
"thinking": "system reasoning",
},
],
},
{
"role": "assistant",
"content": [
{
"type": "thinking",
"thinking": "assistant reasoning",
},
{"type": "text", "text": "The answer is 4."},
],
},
],
"openai",
[
{
"role": "system",
"content": [
{
"type": "text",
"text": "You are a helpful assistant.",
},
{
"type": "thinking",
"thinking": "system reasoning",
},
],
},
{
"role": "assistant",
"content": [
{
"type": "thinking",
"thinking": "assistant reasoning",
},
{"type": "text", "text": "The answer is 4."},
],
},
],
id="openai_system_and_assistant",
),
pytest.param(
"phi3v_model_config",
[
{
"role": "assistant",
"content": [
{"type": "thinking", "thinking": "first thought"},
{"type": "text", "text": "middle text"},
{"type": "thinking", "thinking": "second thought"},
],
},
],
"openai",
[
{
"role": "assistant",
"content": [
{
"type": "thinking",
"thinking": "first thought",
},
{"type": "text", "text": "middle text"},
{
"type": "thinking",
"thinking": "second thought",
},
],
},
],
id="openai_multiple_thinking_blocks",
),
pytest.param(
"phi3v_model_config",
[
{
"role": "assistant",
"content": "The answer is 4.",
"reasoning": "I computed 2+2",
},
],
"openai",
[
{
"role": "assistant",
"content": [
{"type": "text", "text": "The answer is 4."},
],
"reasoning": "I computed 2+2",
"reasoning_content": "I computed 2+2",
},
],
id="top_level_reasoning_no_thinking",
),
pytest.param(
"mistral_model_config",
[
{
"role": "assistant",
"content": [
{"type": "text", "text": "text content"},
{
"type": "thinking",
"thinking": "thinking content",
},
],
},
],
"string",
[
{
"role": "assistant",
"content": "text content",
"reasoning": "thinking content",
"reasoning_content": "thinking content",
},
],
id="string_thinking_extracted",
),
],
)
def test_thinking_handling(
self,
request: pytest.FixtureRequest,
config_fixture: str,
messages: list[dict[str, object]],
content_format: Literal["openai", "string"],
expected: list[dict[str, object]],
) -> None:
model_config = request.getfixturevalue(config_fixture)
conversation, _, _ = parse_chat_messages(
messages,
model_config,
content_format=content_format,
)
assert conversation == expected

@pytest.mark.parametrize(
"model_config_fixture", ["phi3v_model_config", "mistral_model_config"]
)
def test_reasoning_and_thinking_conflict_raises(
self,
model_config_fixture: str,
request: pytest.FixtureRequest,
) -> None:
model_config = request.getfixturevalue(model_config_fixture)
messages = [
{
"role": "assistant",
"content": [
{"type": "thinking", "thinking": "some thinking"},
{"type": "text", "text": "answer"},
],
"reasoning": "top-level reasoning",
},
]

expected_conversation = [
{
"role": "system",
"content": [
{"type": "text", "text": "You are a helpful assistant."},
{
"type": "text",
"text": "Only return the answer when you are confident.",
},
],
},
{
"role": "user",
"content": [{"type": "text", "text": "What is 2+2?"}],
},
{
"role": "assistant",
"content": [
{"type": "text", "text": "Let me think about it."},
{"type": "text", "text": "2+2 = 4"},
{"type": "text", "text": "The answer is 4."},
],
},
]

assert conversation_with_thinking == expected_conversation
with pytest.raises(
VLLMValidationError,
match="Cannot specify both a top-level 'reasoning' field and "
"'thinking' content blocks",
):
parse_chat_messages(
messages,
model_config,
content_format="openai",
)


def test_parse_chat_messages_single_empty_audio_with_uuid(
Expand Down
Loading
Loading