A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 11, 21:15 UTC
[ez] move harmony utils to parser folder (#30117)
Signed-off-by: Andrew Xia <axia@fb.com> Co-authored-by: Andrew Xia <axia@fb.com>
Andrew Xia committed
Dec 6, 2025 at 14:34 UTC
421125d03a110df7d49f84c7cf8ee9fa089d1dff
9 files changed
+9
-9
tests/entrypoints/openai/parser/__init__.py
tests/entrypoints/openai/parser/test_harmony_utils.py
renamed
+1
-1
@@ -4,7 +4,7 @@
4
from openai.types.responses import ResponseFunctionToolCall, ResponseReasoningItem
5
from openai_harmony import Author, Message, Role, TextContent
6
7
-from vllm.entrypoints.harmony_utils import (
7
+from vllm.entrypoints.openai.parser.harmony_utils import (
8
has_custom_tools,
9
parse_input_to_harmony_message,
10
parse_output_message,
tests/entrypoints/openai/test_response_api_with_harmony.py
+1
-1
@@ -726,7 +726,7 @@ async def test_function_calling_required(client: OpenAI, model_name: str):
726
@pytest.mark.asyncio
727
@pytest.mark.parametrize("model_name", [MODEL_NAME])
728
async def test_system_message_with_tools(client: OpenAI, model_name: str):
729
- from vllm.entrypoints.harmony_utils import get_system_message
729
+ from vllm.entrypoints.openai.parser.harmony_utils import get_system_message
730
731
# Test with custom tools enabled - commentary channel should be available
732
sys_msg = get_system_message(with_custom_tools=True)
vllm/entrypoints/context.py
+1
-1
@@ -19,7 +19,7 @@ from vllm import envs
19
from vllm.entrypoints.chat_utils import (
20
ChatTemplateContentFormatOption,
21
)
22
-from vllm.entrypoints.harmony_utils import (
22
+from vllm.entrypoints.openai.parser.harmony_utils import (
23
get_encoding,
24
get_streamable_parser_for_assistant,
25
render_for_completion,
vllm/entrypoints/openai/parser/harmony_utils.py
renamed
vllm/entrypoints/openai/serving_chat.py
+2
-2
@@ -21,7 +21,8 @@ from vllm.entrypoints.chat_utils import (
21
get_history_tool_calls_cnt,
22
make_tool_call_id,
23
)
24
-from vllm.entrypoints.harmony_utils import (
24
+from vllm.entrypoints.logger import RequestLogger
25
+from vllm.entrypoints.openai.parser.harmony_utils import (
26
get_developer_message,
27
get_stop_tokens_for_assistant_actions,
28
get_streamable_parser_for_assistant,
@@ -30,7 +31,6 @@ from vllm.entrypoints.harmony_utils import (
31
parse_input_to_harmony_message,
32
render_for_completion,
33
)
33
-from vllm.entrypoints.logger import RequestLogger
34
from vllm.entrypoints.openai.protocol import (
35
ChatCompletionLogProb,
36
ChatCompletionLogProbs,
vllm/entrypoints/openai/serving_responses.py
+2
-2
@@ -64,7 +64,8 @@ from vllm.entrypoints.context import (
64
SimpleContext,
65
StreamingHarmonyContext,
66
)
67
-from vllm.entrypoints.harmony_utils import (
67
+from vllm.entrypoints.logger import RequestLogger
68
+from vllm.entrypoints.openai.parser.harmony_utils import (
69
construct_harmony_previous_input_messages,
70
get_developer_message,
71
get_stop_tokens_for_assistant_actions,
@@ -76,7 +77,6 @@ from vllm.entrypoints.harmony_utils import (
77
parse_response_input,
78
render_for_completion,
79
)
79
-from vllm.entrypoints.logger import RequestLogger
80
from vllm.entrypoints.openai.protocol import (
81
DeltaMessage,
82
ErrorResponse,
vllm/entrypoints/openai/tool_parsers/openai_tool_parser.py
+1
-1
@@ -4,7 +4,7 @@ import json
4
from collections.abc import Sequence
5
from typing import TYPE_CHECKING
6
7
-from vllm.entrypoints.harmony_utils import parse_output_into_messages
7
+from vllm.entrypoints.openai.parser.harmony_utils import parse_output_into_messages
8
from vllm.entrypoints.openai.protocol import (
9
ChatCompletionRequest,
10
DeltaMessage,
vllm/reasoning/gptoss_reasoning_parser.py
+1
-1
@@ -5,7 +5,7 @@ from collections.abc import Sequence
5
6
from transformers import PreTrainedTokenizerBase
7
8
-from vllm.entrypoints.harmony_utils import parse_chat_output
8
+from vllm.entrypoints.openai.parser.harmony_utils import parse_chat_output
9
from vllm.entrypoints.openai.protocol import ChatCompletionRequest, DeltaMessage
10
from vllm.entrypoints.tool_server import ToolServer
11
from vllm.logger import init_logger